0.983688 -7.53492 as well as 0.58373 -7.53707 et al. 0.449749 -7.96724 based on 0.790833 -8.13204 will be 0.912485 -8.18256 associated with 0.888628 -8.24017 rather than 0.8091 -8.25847 should be 0.636122 -8.3407 according to 0.894521 -8.38248 with respect to 0.704526 -8.38839 may be 0.717234 -8.54142 in terms of 0.6679 -8.5742 machine translation 0.490123 -8.64426 in order to 0.926 -8.69197 must be 0.614181 -8.86223 more than 0.915493 -8.87455 along with 0.631166 -8.87948 could be 0.882222 -8.89289 derived from 0.637312 -8.89405 in this paper, we 0.473369 -8.93415 a single 0.988304 -8.94055 rely on 0.79845 -8.95723 machine learning 0.945504 -8.9575 followed by 0.872038 -8.9803 correspond to 0.643799 -9.00271 did not 0.996785 -9.0182 carried out 0.664286 -9.02018 test set 0.8379 -9.0235 there is no 0.788934 -9.03629 corresponds to 0.585616 -9.04842 would be 0.85049 -9.06358 named entity 0.865482 -9.06405 most frequent 0.9653 -9.06525 at least one 0.868895 -9.07053 et al., 2010) 0.646638 -9.07477 extracted from 0.964052 -9.10255 allows us 0.571098 -9.10935 & 0.573598 -9.11193 at least 0.375251 -9.11411 has been 0.415249 -9.14063 due to 0.774336 -9.15114 on the other hand, 0.768879 -9.19648 less than 0.785024 -9.21184 information about 0.618759 -9.21868 better than 0.835655 -9.22914 table 4: 0.384388 -9.23146 have been 0.873457 -9.24214 produced by 0.830861 -9.3034 depending on 0.670588 -9.31709 in addition to 0.766234 -9.33093 cannot be 0.610649 -9.33993 gold standard 0.527363 -9.34172 instead of 0.521793 -9.36536 c©2011 association for computational linguistics 0.852843 -9.3684 likely to be 0.953975 -9.3712 deal with 0.938525 -9.38251 error rate 0.843854 -9.38551 together with 0.57837 -9.38938 figure 1: 0.813084 -9.3954 data sets 0.706161 -9.40248 word alignment 0.952381 -9.40474 in other words, 0.325207 -9.43583 do not 0.835052 -9.43929 pos tags 0.439163 -9.4404 language model 0.745856 -9.44809 hmm 0.175118 -9.44833 can be 0.398413 -9.45528 natural language 0.870722 -9.45742 coreference resolution 0.724868 -9.459 lead to 0.723684 -9.45935 focused on 0.57596 -9.46027 as follows: 0.748588 -9.46207 a variety of 0.629259 -9.46516 table 3: 0.746479 -9.46615 dependency parsing 0.875486 -9.46774 spans 0.844203 -9.46882 14 0.683333 -9.47365 shown in figure 0.900415 -9.4763 as well. 0.80602 -9.48169 we would like to 0.566445 -9.48934 focus on 0.493622 -9.49978 applied to 0.881633 -9.50283 2 related work 0.952153 -9.50542 figure 4: 0.795302 -9.51224 higher than 0.828467 -9.51313 attempt to 0.485607 -9.51343 related to 0.86166 -9.51486 (such as 0.720222 -9.5178 leads to 0.883333 -9.52001 preprocessing 0.626327 -9.53152 a. 0.789831 -9.53543 large number of 0.933649 -9.53683 most likely 0.592734 -9.53838 data set 0.764331 -9.53953 in this section, 0.660333 -9.54046 structural 0.969072 -9.54767 named entities 0.941463 -9.55007 manually annotated 0.614108 -9.5507 language models 0.499314 -9.55088 table 1: 0.798587 -9.5562 noun phrase 0.629386 -9.55639 training data. 0.6125 -9.55975 generating 0.936275 -9.56381 relationship between 0.958549 -9.57377 sentence-level 0.918269 -9.58567 experimented with 0.522621 -9.58766 german 0.859574 -9.5932 we believe that 0.69337 -9.59377 references 0.604211 -9.59547 ... 0.785714 -9.59853 we want to 0.646489 -9.60209 rich 0.921182 -9.60223 in contrast to 0.920792 -9.60582 penn treebank 0.60084 -9.60608 traditional 0.988506 -9.61601 depend on 0.607843 -9.61974 to build 0.818182 -9.62008 table 5: 0.780576 -9.62045 16 0.762069 -9.6228 a subset of 0.479564 -9.62334 to identify 0.732484 -9.6256 to avoid 0.906404 -9.63436 18 0.890476 -9.63585 look at 0.760417 -9.63667 tests 0.935829 -9.65368 belong to 0.982249 -9.65404 resulted in 0.921875 -9.65449 treated as 0.781955 -9.65811 bleu score 0.926316 -9.6587 comparisons 0.845815 -9.66266 more complex 0.855204 -9.66423 in this case, 0.944444 -9.66979 widely used 0.515807 -9.67598 vs. 0.954286 -9.67778 information retrieval 0.804878 -9.68074 (see section 0.993789 -9.68271 processes 0.701863 -9.6833 verbal 0.846154 -9.68558 ner 0.485163 -9.68672 our method 0.647215 -9.68847 kind of 0.948571 -9.68993 inspired by 0.551257 -9.6965 up to 0.948276 -9.69799 table 6: 0.819742 -9.69848 future work. 0.892857 -9.6996 focuses on 0.942857 -9.70094 in practice, 0.88 -9.70674 cosine similarity 0.300818 -9.70943 a set of 0.603774 -9.71106 generated by 0.947674 -9.71149 how well 0.769231 -9.71436 access to 0.761364 -9.71851 close to 0.969325 -9.71872 domain adaptation 0.290181 -9.71918 does not 0.866337 -9.72808 noun phrases 0.818584 -9.73102 relies on 0.758621 -9.73748 represented by 0.941176 -9.73812 each other. 0.49835 -9.73909 to generate 0.405702 -9.74072 modeling 0.539806 -9.74091 we do not 0.762646 -9.74196 differences between 0.876923 -9.74276 drawn from 0.760618 -9.74305 depends on 0.839623 -9.74571 can be found 0.798283 -9.75038 dialogue act 0.896739 -9.75091 constructing 0.620779 -9.7512 showed that 0.555094 -9.75219 a few 0.476336 -9.75219 refer to 0.734545 -9.75346 r. 0.405345 -9.75813 compared to 0.636364 -9.75963 table 1. 0.623342 -9.76554 consisting of 0.733333 -9.77509 reranking 0.588942 -9.77882 improving 0.815668 -9.77895 21 0.574713 -9.78241 filtering 0.56044 -9.78783 figure 2: 0.888889 -9.79403 solutions 0.718978 -9.79688 can be seen 0.643275 -9.79722 minimal 0.714801 -9.79933 we found that 0.754032 -9.79966 bigrams 0.926829 -9.8004 table 3 shows 0.279623 -9.80226 - 0.700348 -9.80299 smoothing 0.6 -9.80534 might be 0.440608 -9.80772 problems 0.865591 -9.81132 lexicalized 0.925926 -9.81561 pos tag 0.706093 -9.81683 past 0.895954 -9.8196 consist of 0.545064 -9.82161 tend to 0.920245 -9.82322 speech recognition 0.979167 -9.82455 controlled 0.686007 -9.82501 obtained from 0.609164 -9.8258 experimental results 0.833333 -9.82706 table 1 shows 0.648318 -9.82816 lack of 0.909639 -9.82931 leading to 0.93038 -9.83246 objective function 0.333876 -9.8326 part of 0.829146 -9.83387 as a result, 0.636905 -9.83834 regular 0.77193 -9.8396 age 0.992701 -9.84331 logistic regression 0.884393 -9.84556 22 0.69395 -9.8457 test set. 0.73494 -9.84849 phrase pairs 0.541485 -9.85071 syntax 0.90303 -9.8509 divided into 0.689046 -9.85105 my 0.978417 -9.8562 dynamic programming 0.454264 -9.85971 table 2: 0.415803 -9.85971 via 0.856354 -9.86152 17 0.892216 -9.8617 its own 0.511811 -9.86235 generally 0.922581 -9.86463 real-world 0.877907 -9.86552 we can see that 0.728 -9.8659 training data, 0.662252 -9.86732 accuracies 0.985294 -9.86787 so far 0.660066 -9.86823 hybrid 0.350698 -9.86914 detection 0.733607 -9.87355 statistically significant 0.700375 -9.87551 forest 0.922078 -9.87797 capable of 0.625749 -9.87829 all possible 0.885542 -9.8791 word alignments 0.744681 -9.88119 make use 0.992424 -9.88126 etc.) 0.783019 -9.88329 in this work, we 0.699248 -9.88411 upon 0.785714 -9.88539 flat 0.88024 -9.8856 32 0.748918 -9.88898 represented as 0.977778 -9.88997 figure 5: 0.645161 -9.89191 c. 0.732218 -9.89509 we find that 0.763636 -9.89794 conversation 0.89375 -9.89804 experimental setup 0.722449 -9.89877 success 0.6231 -9.89882 employed 0.468268 -9.90048 50 0.702703 -9.90074 constructions 0.112037 -9.90087 et al., 0.827957 -9.90132 0, 0.611111 -9.90274 later 0.6231 -9.90276 why 0.562189 -9.90576 12 0.823529 -9.90583 coarse 0.705882 -9.90724 parallel corpus 0.903226 -9.90818 commonly used 0.851429 -9.9084 compounds 0.8125 -9.9087 cross-lingual 0.970149 -9.90902 repeat 0.581769 -9.91276 11 0.583784 -9.91473 s. 0.496094 -9.91511 future work 0.63961 -9.91622 weighting 0.787129 -9.9181 benefit from 0.913333 -9.92337 word-based 0.682836 -9.92391 oov 0.955882 -9.92403 annotations, 0.8 -9.92422 e. 0.774038 -9.92513 expansion 0.984375 -9.9264 speech act 0.93007 -9.92951 45 0.643333 -9.92982 involving 0.619195 -9.93095 not necessarily 0.612121 -9.93196 et al., 2008). 0.607784 -9.9377 student 0.652778 -9.93925 relations between 0.984252 -9.94054 marginal 0.725322 -9.9429 interactions 0.666667 -9.94374 event extraction 0.865031 -9.9446 randomly selected 0.722222 -9.94535 wsd 0.727273 -9.9456 even if 0.651568 -9.94586 13 0.669118 -9.9468 dictionaries 0.496957 -9.9498 almost 0.825843 -9.95149 first step 0.99187 -9.954 et al.’s 0.821229 -9.95424 tune 0.6875 -9.95517 this paper. 0.99187 -9.95583 intrinsic 0.809783 -9.95803 two types of 0.669145 -9.95817 post 0.916084 -9.95954 led to 0.883117 -9.96238 belonging to 0.739726 -9.96285 targets 0.636364 -9.96355 results show that 0.841176 -9.96355 search space 0.768473 -9.96422 at most 0.385856 -9.96453 our system 0.771144 -9.96525 lower than 0.697959 -9.96586 speed 0.692 -9.96714 it is possible 0.615873 -9.96724 ability to 0.75 -9.9677 czech 0.98374 -9.96879 unable to 0.551282 -9.97605 as follows. 0.909091 -9.97629 mapped to 0.717391 -9.97646 translation quality 0.760976 -9.97649 phrase table 0.568306 -9.97757 total number 0.498947 -9.97819 training set 0.991667 -9.97893 options 0.641115 -9.97914 fine-grained 0.707627 -9.97958 sentence pairs 0.669202 -9.97975 historical 0.843373 -9.98024 h. 0.596386 -9.98075 usage 0.308126 -9.98226 various 0.582133 -9.98548 subjective 0.321869 -9.98616 training data 0.78125 -9.98714 family 0.723214 -9.98798 typed 0.983333 -9.99157 focusing on 0.953125 -9.99219 dialogue acts 0.72 -9.99287 humans 0.691358 -9.9936 maximum entropy 0.436275 -9.99393 in addition, 0.464684 -9.99464 to learn 0.8125 -9.99465 clauses 0.359202 -9.99469 8 0.816092 -9.99532 deterministic 0.462385 -9.99536 so that 0.624161 -9.99558 incorporating 0.859873 -9.99582 supervised learning 0.991525 -9.99678 forum 0.585799 -9.99779 ambiguity 0.816092 -9.99833 eight 0.667954 -10.0005 effectively 0.884354 -10.0032 semantic roles 0.858974 -10.0043 dealing with 0.705628 -10.0047 word order 0.427663 -10.0058 the fact that 0.642599 -10.0096 portion of 0.991379 -10.0119 document-level 0.768041 -10.0125 nominal 0.567797 -10.0126 location 0.883562 -10.013 test set, 0.704348 -10.0132 penalty 0.558904 -10.0141 generic 0.567797 -10.0154 in this paper we 0.381865 -10.0156 to improve 0.991379 -10.0161 informal 0.852564 -10.0171 divergence 0.982906 -10.0191 publicly available 0.432343 -10.02 short 0.991304 -10.0218 structures, 0.991304 -10.0219 filters 0.660232 -10.0222 id 0.526961 -10.0234 information, 0.877551 -10.0235 figure 2 shows 0.866667 -10.0254 closely related 0.434057 -10.0261 to extract 0.709821 -10.027 workers 0.643382 -10.0278 relations, 0.662745 -10.0301 practical 0.748744 -10.033 empirically 0.898551 -10.0333 correlated with 0.462715 -10.0339 smaller 0.892857 -10.0342 in section 2, 0.70852 -10.0342 functional 0.684874 -10.0346 different types 0.506912 -10.0352 15 0.99115 -10.0357 upper bound 0.798851 -10.0369 hold 0.910448 -10.037 replaced by 0.633094 -10.038 oracle 0.982609 -10.0382 there exists 0.966387 -10.0401 names, 0.929688 -10.0413 talk 0.875 -10.0441 smith 0.794286 -10.0449 handling 0.589905 -10.0453 be able to 0.541114 -10.0491 others 0.839744 -10.0498 comparable corpora 0.922481 -10.0524 differs from 0.443649 -10.0529 9 0.957983 -10.0532 ranging from 0.530928 -10.0549 predicting 0.702703 -10.0556 development set 0.413417 -10.0557 statistics 0.973913 -10.0566 as opposed to 0.965812 -10.0567 smoothed 0.808383 -10.0571 tends to 0.950413 -10.0575 their own 0.943089 -10.0576 accepted 0.825 -10.0603 ge 0.489035 -10.0603 hypotheses 0.66129 -10.0631 challenges 0.928571 -10.0635 growing 0.693333 -10.0637 poor 0.852349 -10.0638 soft 0.784091 -10.0638 uncertainty 0.928571 -10.065 table 4 shows 0.847682 -10.0651 26 0.908397 -10.0663 situations 0.990909 -10.0676 selection, 0.990909 -10.0677 benchmark 0.878571 -10.0679 alternatives 0.735 -10.0687 integrating 0.914729 -10.069 problematic 0.61324 -10.0691 and, 0.663934 -10.07 have shown 0.74359 -10.0716 compositional 0.487859 -10.0724 below 0.762162 -10.0729 ilp 0.270014 -10.0743 10 0.883212 -10.0747 determine whether 0.539295 -10.075 speaker 0.914062 -10.0776 dependent on 0.934426 -10.0787 simulated 0.906977 -10.0808 spoken dialogue 0.801205 -10.0814 codes 0.668067 -10.0819 sparse 0.507246 -10.0841 an important 0.594684 -10.0844 25 0.990741 -10.0872 table 7: 0.88806 -10.088 pragmatic 0.990741 -10.0882 package 0.990741 -10.0883 chinese-english 0.360736 -10.0889 basic 0.517766 -10.09 that is, 0.704225 -10.0901 we assume that 0.661157 -10.0901 account for 0.803681 -10.0918 syntax-based 0.730964 -10.0937 et al., 2008) 0.294893 -10.0941 web 0.990654 -10.0944 spurious 0.981651 -10.0953 relying on 0.948718 -10.0955 else 0.990654 -10.0969 31 0.451362 -10.0997 range of 0.384724 -10.1006 in this paper 0.41806 -10.1013 exact 0.880597 -10.1028 political 0.782353 -10.104 weak 0.990566 -10.1056 size, 0.948276 -10.1059 conditional probability 0.892308 -10.1059 ci 0.947826 -10.1071 slots 0.696262 -10.1074 in this section 0.708738 -10.108 aac 0.430357 -10.1084 f-score 0.570978 -10.1107 potentially 0.688073 -10.1117 cases where 0.326132 -10.1136 segmentation 0.836735 -10.1139 preferred 0.73057 -10.1159 normalization 0.339286 -10.1161 trained on 0.972477 -10.1163 well-known 0.808917 -10.1176 how many 0.822368 -10.1183 taken from 0.748634 -10.1183 merging 0.739362 -10.1191 sparsity 0.768786 -10.1198 utilize 0.808917 -10.1207 form, 0.515544 -10.1208 beyond 0.567823 -10.1219 sometimes 0.501229 -10.122 2) 0.867647 -10.1221 outperformed 0.585284 -10.1237 cues 0.611722 -10.1242 to detect 0.92437 -10.1242 lexical items 0.92437 -10.1249 correctness 0.522788 -10.125 grammars 0.672566 -10.1253 information extraction 0.783133 -10.1262 unigrams 0.578947 -10.1283 communication 0.566038 -10.1287 field 0.980952 -10.1288 filter out 0.877863 -10.131 internet 0.634921 -10.1316 specifically 0.85 -10.1332 public 0.770588 -10.1335 specialized 0.16453 -10.1349 such as 0.990291 -10.135 1999) 0.806452 -10.135 this means that 0.938596 -10.1353 subtrees 0.566879 -10.1357 mostly 0.562893 -10.1357 refers to 0.445972 -10.1359 in proceedings of 0.62069 -10.137 structure, 0.946429 -10.1379 relationships between 0.990291 -10.1382 extreme 0.810458 -10.1382 parallel corpora 0.848921 -10.1395 b) 0.576159 -10.1405 effectiveness of 0.451417 -10.1411 previous work 0.555215 -10.1424 kinds of 0.665198 -10.1434 analyzed 0.308129 -10.1436 ranking 0.583051 -10.1439 nature of 0.612782 -10.1443 wrong 0.828767 -10.1452 hits 0.923077 -10.1452 queries, 0.722513 -10.1464 a lot 0.990196 -10.1471 extractor 0.990196 -10.1477 social media 0.609665 -10.1478 writing 0.599278 -10.1486 types, 0.990099 -10.149 cross validation 0.419014 -10.1503 especially 0.570033 -10.1509 samples 0.895161 -10.1512 vector space 0.915254 -10.1512 properly 0.888889 -10.1514 tokenization 0.772455 -10.1518 by means 0.564103 -10.1521 papers 0.980583 -10.1526 directional 0.747191 -10.1532 small number of 0.232824 -10.1541 without 0.971429 -10.1553 error reduction 0.6 -10.1555 gap 0.489104 -10.1562 difference between 0.602941 -10.1578 parse tree 0.760234 -10.1592 deep 0.929825 -10.1593 synonymy 0.836879 -10.1607 recursive 0.869231 -10.161 inter-annotator agreement 0.888 -10.1614 chunking 0.738889 -10.1619 parse trees 0.830986 -10.1626 it is important to 0.962264 -10.1633 of course, 0.953704 -10.1634 leaving 0.426716 -10.1646 statistical machine 0.798701 -10.1647 aims to 0.99 -10.1652 48 0.84058 -10.1653 sequential 0.99 -10.1654 44 0.741573 -10.1658 equally 0.751445 -10.1677 fast 0.893443 -10.1688 annotation scheme 0.399674 -10.169 highly 0.868217 -10.1699 training set, 0.961905 -10.1699 korean 0.533528 -10.1701 prosodic 0.62753 -10.1706 b. 0.713542 -10.1709 sense disambiguation 0.989899 -10.1716 so on. 0.574324 -10.1717 implicit 0.718085 -10.1741 training set. 0.989899 -10.175 content, 0.271624 -10.1759 full 0.886179 -10.1765 chat 0.668203 -10.1771 consistency 0.792208 -10.1776 feature sets 0.542683 -10.1777 semi-supervised 0.519553 -10.1785 punctuation 0.844444 -10.1792 ; 0.582456 -10.1802 bootstrapping 0.521127 -10.1802 popular 0.409722 -10.1811 arabic 0.390205 -10.1826 you 0.989796 -10.1833 ensemble 0.519663 -10.1837 approximately 0.729282 -10.1859 viterbi 0.586331 -10.1893 even though 0.833333 -10.19 regardless of 0.794702 -10.19 without any 0.693467 -10.1914 figure 1 shows 0.552716 -10.1917 figure 1. 0.646288 -10.1917 shallow 0.970297 -10.192 population 0.927928 -10.1921 kl 0.541538 -10.1924 cause 0.808219 -10.193 section 4. 0.885246 -10.1933 this work. 0.550955 -10.1935 figure 3: 0.734463 -10.1942 chunks 0.827338 -10.195 news articles 0.392569 -10.1955 de 0.989691 -10.1955 display 0.866142 -10.1961 role labeling 0.816901 -10.1961 backoff 0.989691 -10.1966 proposal 0.952381 -10.1967 stronger 0.422932 -10.197 online 0.705263 -10.1983 possibly 0.842105 -10.1984 feature space 0.737143 -10.1985 diversity 0.798658 -10.1985 f. 0.497396 -10.1988 40 0.951923 -10.1994 a1 0.723757 -10.2002 retrieved 0.853846 -10.2007 l. 0.678049 -10.2007 overlapping 0.747059 -10.2009 patterns, 0.847328 -10.201 transformed 0.329143 -10.201 chinese 0.842105 -10.2012 transcription 0.811189 -10.2012 facts 0.842105 -10.2041 caused by 0.806897 -10.2043 let us 0.826087 -10.2052 abstracts 0.693878 -10.2056 put 0.71978 -10.2061 19 0.501333 -10.2065 approximate 0.951923 -10.2077 250 0.521739 -10.21 yield 0.718232 -10.2104 meaningful 0.759259 -10.2117 abstractive 0.59176 -10.2121 substantial 0.779221 -10.2124 labels, 0.657407 -10.213 heuristics 0.210826 -10.2132 several 0.393388 -10.2136 typically 0.606299 -10.2151 hindi 0.676471 -10.2168 fluency 0.35 -10.2169 100 0.7625 -10.217 merge 0.910714 -10.2175 dialectal 0.989474 -10.2176 doesn’t 0.630901 -10.2177 defined by 0.481297 -10.2181 modified 0.888889 -10.2194 coordination 0.95098 -10.2195 conditioned on 0.96 -10.22 hypergraph 0.989474 -10.2201 prosody 0.895652 -10.2203 topical 0.647059 -10.2208 k. 0.979167 -10.2211 more sophisticated 0.774194 -10.2217 pos tagging 0.851562 -10.2219 swedish 0.814286 -10.2223 150 0.681818 -10.2244 situation 0.772727 -10.2254 multi-word 0.850394 -10.227 limitations 0.451327 -10.2273 groups 0.868852 -10.2274 physical 0.828358 -10.2279 purely 0.989362 -10.2282 trivial 0.590909 -10.2283 in fact, 0.868852 -10.2287 section 3, 0.989362 -10.2289 agreement, 0.594595 -10.2307 proportion of 0.588679 -10.2317 parts of 0.730994 -10.2328 bleu scores 0.53605 -10.2328 mwes 0.901786 -10.2334 erroneous 0.959596 -10.2335 tweets, 0.405745 -10.2341 textual 0.702703 -10.2344 (%) 0.691099 -10.2345 conventional 0.802817 -10.235 best performing 0.594595 -10.2351 negation 0.678392 -10.2362 figure 2. 0.482143 -10.2371 etc. 0.469734 -10.2375 datasets 0.941748 -10.2388 stands for 0.560554 -10.2391 an additional 0.38843 -10.2392 long 0.909091 -10.2393 bleu, 0.854839 -10.2397 data set, 0.65566 -10.24 fairly 0.643836 -10.2401 combined with 0.576642 -10.2407 frames 0.423228 -10.2411 paraphrases 0.893805 -10.2414 v. 0.860656 -10.2421 k-means 0.978947 -10.2421 regarded as 0.579336 -10.243 we focus on 0.743902 -10.2437 figure 1, 0.754717 -10.2446 phrase pair 0.801418 -10.2447 assumptions 0.6875 -10.2452 test sets 0.886957 -10.2454 } 0.95 -10.2457 diferent 0.872881 -10.246 · 0.98913 -10.246 addressing 0.636771 -10.2462 paraphrasing 0.96875 -10.2472 concerned with 0.220257 -10.2475 . 0.542484 -10.2475 phenomena 0.782313 -10.248 we observe that 0.84252 -10.249 appearing in 0.87931 -10.2494 worse than 0.98913 -10.2495 classifiers, 0.479592 -10.2501 only one 0.98913 -10.2505 hm 0.87931 -10.251 started 0.577778 -10.2513 virtual 0.892857 -10.2522 unaligned 0.718391 -10.2528 by applying 0.685864 -10.2529 experiment with 0.853659 -10.2546 fusion 0.442982 -10.2556 unigram 0.578358 -10.256 baselines 0.777027 -10.2567 emotions 0.808824 -10.2572 das 0.564286 -10.2582 preference 0.737805 -10.2583 targeted 0.829457 -10.259 unary 0.892857 -10.2592 realization 0.649289 -10.2594 extremely 0.989011 -10.2596 verbnet 0.658537 -10.2597 generalization 0.658537 -10.2601 modern 0.350345 -10.2615 questions 0.9 -10.2616 sound 0.931373 -10.2626 no longer 0.576779 -10.2632 brown 0.629464 -10.2634 table 3. 0.46683 -10.2636 the latter 0.846774 -10.2638 assessing 0.663366 -10.2655 relational 0.858333 -10.2657 expressions, 0.94898 -10.2677 inflectional 0.361357 -10.2684 global 0.988889 -10.2684 leveraging 0.931373 -10.2684 concepts, 0.697802 -10.2686 intersection 0.615385 -10.2692 all three 0.59127 -10.2692 consistent with 0.858333 -10.2703 distinguish between 0.75 -10.2716 active learning 0.899083 -10.2725 comes from 0.77027 -10.2729 concrete 0.883929 -10.2736 measures, 0.391304 -10.2739 labeling 0.43129 -10.2746 bigram 0.731707 -10.2748 re 0.761589 -10.2748 their corresponding 0.595142 -10.2757 slot 0.753247 -10.2757 redundant 0.668367 -10.276 chain 0.698324 -10.2786 duration 0.85124 -10.2788 distributional similarity 0.801471 -10.2788 alignment, 0.32801 -10.279 matching 0.71345 -10.2796 (2003) 0.773973 -10.2798 expensive 0.90566 -10.28 constructs 0.939394 -10.2806 undirected 0.54 -10.2807 creating 0.890909 -10.2807 nc 0.394316 -10.2812 names 0.563636 -10.2813 zero 0.393617 -10.2827 entailment 0.988764 -10.2828 connect 0.460048 -10.2829 to create 0.378913 -10.2834 shown in table 0.957895 -10.2839 we hypothesize that 0.318605 -10.2842 clusters 0.869565 -10.2852 player 0.505882 -10.2854 methodology 0.921569 -10.2856 bottom-up 0.706897 -10.2858 feature selection 0.608511 -10.2859 (c) 0.78169 -10.2859 tutoring 0.631336 -10.2865 old 0.862069 -10.2871 german, 0.832 -10.2877 ontological 0.475196 -10.2884 typical 0.599174 -10.2885 specified 0.85 -10.2886 discovery 0.512121 -10.2913 return 0.27758 -10.2913 constraints 0.27383 -10.2913 supervised 0.490251 -10.2919 substitution 0.988636 -10.2926 polynomial 0.868421 -10.2933 stock 0.904762 -10.2935 orders 0.875 -10.2944 fundamental 0.698864 -10.2945 perfect 0.988636 -10.2949 triples 0.479893 -10.2966 particularly 0.938144 -10.2969 maximum likelihood 0.698864 -10.297 generalize 0.440181 -10.2974 scale 0.451306 -10.2974 frequently 0.881818 -10.2976 indirect 0.286948 -10.2991 tag 0.513846 -10.2996 mwe 0.920792 -10.302 encodes 0.601695 -10.3026 detected 0.551601 -10.3036 code 0.255141 -10.3036 random 0.424051 -10.3043 fully 0.761905 -10.3058 (see figure 0.988506 -10.3064 trends 0.967033 -10.3065 averaged over 0.697143 -10.3066 annotating 0.788321 -10.3066 language pairs 0.988506 -10.3073 nested 0.928571 -10.3076 quality, 0.956989 -10.3079 marking 0.988506 -10.3082 linearly 0.988506 -10.3089 apart from 0.988506 -10.3099 rise 0.988506 -10.3099 51 0.719512 -10.3109 synonym 0.327434 -10.3118 summary 0.584677 -10.3119 we obtain 0.873874 -10.3121 interpretations 0.594142 -10.3132 hierarchy 0.834711 -10.314 irrelevant 0.988372 -10.3141 freely available 0.594142 -10.3145 the presence of 0.479564 -10.3153 go 0.846154 -10.3165 fourth 0.988372 -10.3167 alignments, 0.65641 -10.3168 achieved by 0.421053 -10.3169 mining 0.946809 -10.3174 our own 0.9375 -10.3178 mistakes 0.966667 -10.3183 locally 0.36378 -10.3186 coverage 0.364353 -10.3188 temporal 0.539792 -10.3189 distributed 0.946809 -10.3193 affected by 0.988372 -10.3193 interpret 0.552727 -10.3194 decisions 0.977273 -10.3195 encouraging 0.514196 -10.3196 multilingual 0.679558 -10.3199 similarities 0.988372 -10.3201 aimed at 0.87963 -10.3215 5) 0.369106 -10.3226 trigger 0.988235 -10.3236 naive bayes 0.50303 -10.3241 memory 0.746667 -10.3241 self-training 0.791045 -10.3245 survey 0.675824 -10.3261 by adding 0.444444 -10.3262 reviews 0.87156 -10.3272 on average, 0.495575 -10.3274 raw 0.72327 -10.3275 classified as 0.640394 -10.3286 bad 0.506135 -10.3286 θ 0.858407 -10.329 rated 0.738562 -10.3291 expanded 0.733766 -10.3298 neutral 0.584362 -10.3306 relation between 0.988235 -10.3314 only, 0.8 -10.3317 dutch 0.785185 -10.3318 sensitive to 0.826446 -10.3322 coarse-grained 0.988235 -10.3325 settings, 0.988235 -10.3326 2.5 0.927083 -10.333 logical form 0.988235 -10.3335 so-called 0.558491 -10.3339 operations 0.426374 -10.334 predictions 0.804688 -10.3341 first-order 0.607143 -10.3355 versions of 0.988095 -10.336 random walk 0.278351 -10.3366 clustering 0.415449 -10.3374 1) 0.547101 -10.3378 23 0.498489 -10.3385 robust 0.571429 -10.3388 personal 0.491228 -10.339 running 0.885714 -10.3392 second-order 0.732026 -10.3393 rouge 0.607143 -10.3397 sentiment analysis 0.783582 -10.3398 reasoning 0.988095 -10.34 complicated 0.412371 -10.3409 million 0.585774 -10.3416 transformation 0.664865 -10.3419 technical 0.877358 -10.3423 k-best 0.79845 -10.3434 maltparser 0.22454 -10.344 linguistic 0.604444 -10.3441 equal to 0.471545 -10.345 external 0.525253 -10.3453 rule-based 0.80315 -10.3457 redundancy 0.830508 -10.3457 policies 0.43418 -10.3475 explicit 0.672222 -10.3482 involved in 0.328477 -10.3495 f1 0.792308 -10.3495 this paper presents 0.877358 -10.3497 come from 0.632353 -10.3501 mappings 0.387454 -10.3507 surface 0.577869 -10.3508 removing 0.758865 -10.3511 collocations 0.253956 -10.3513 per 0.52 -10.3515 classify 0.987952 -10.3524 distinctions 0.741497 -10.3526 conversations 0.965517 -10.3526 face 0.9 -10.3531 33 0.987952 -10.3532 adjusted 0.781955 -10.3541 (using 0.848214 -10.3541 (pos) 0.606335 -10.3541 lexicons 0.823529 -10.3547 high-level 0.848214 -10.3547 starting from 0.610092 -10.355 efforts 0.908163 -10.3551 di 0.916667 -10.3554 universal 0.9 -10.3556 human annotators 0.71519 -10.3556 lms 0.987952 -10.3559 accounts for 0.925532 -10.3568 bionlp shared task 0.916667 -10.3577 as expected, 0.26899 -10.3584 events 0.372852 -10.3595 strong 0.291099 -10.3597 algorithms 0.97619 -10.3609 41 0.944444 -10.3611 fine 0.806452 -10.3624 prototype 0.987805 -10.363 most probable 0.281219 -10.364 categories 0.965116 -10.3641 excluding 0.3859 -10.3641 — 0.481268 -10.3646 included in 0.891089 -10.3652 36 0.310552 -10.366 pattern 0.854545 -10.3662 nine 0.748252 -10.3666 larger than 0.628713 -10.3671 template 0.834783 -10.3671 snippets 0.518519 -10.3673 feature set 0.678161 -10.3678 clinical 0.482558 -10.368 pairwise 0.371972 -10.3681 scoring 0.965116 -10.3686 take advantage 0.924731 -10.369 polysemy 0.907216 -10.37 refined 0.672316 -10.3706 suffix 0.743056 -10.3708 your 0.793651 -10.3713 satisfy 0.62439 -10.3715 criterion 0.652406 -10.3724 300 0.774436 -10.3724 instances, 0.692771 -10.3731 capturing 0.822034 -10.3732 2007 0.934066 -10.3739 predictor 0.680233 -10.3744 hash 0.652406 -10.3754 established 0.726667 -10.376 rating 0.456693 -10.3768 japanese 0.59292 -10.377 uniform 0.975904 -10.3779 speech acts 0.987654 -10.378 representations, 0.682353 -10.3791 knowledge base 0.721854 -10.3793 constraints, 0.695122 -10.3804 binding 0.987654 -10.3813 syllable 0.457672 -10.3815 hard 0.422122 -10.3815 distinct 0.42369 -10.3821 constituent 0.718954 -10.3823 them, 0.678363 -10.3824 constraints on 0.964706 -10.3825 etc.). 0.987654 -10.3831 37 0.776923 -10.3834 language modeling 0.90625 -10.3834 allow us 0.953488 -10.3835 feature templates 0.513423 -10.3837 as described in 0.914894 -10.3845 un 0.826087 -10.385 wer 0.964286 -10.3851 semantic orientation 0.943182 -10.3857 differently 0.845455 -10.3858 block 0.396 -10.3868 out of 0.826087 -10.3879 claims 0.831858 -10.3879 grid 0.943182 -10.388 entity linking 0.552529 -10.3883 studied 0.471591 -10.3884 control 0.634021 -10.3893 utility 0.9875 -10.3902 logs 0.808333 -10.3902 screen 0.964286 -10.3909 search engine 0.695652 -10.3909 naturally 0.831858 -10.391 sufficiently 0.913978 -10.3914 10, 0.757353 -10.3917 richer 0.28012 -10.392 complex 0.837838 -10.393 feature vectors 0.562753 -10.3933 scientific 0.905263 -10.3936 lsa 0.766917 -10.394 vs 0.648649 -10.3949 extractive 0.49375 -10.3949 49th annual meeting 0.818966 -10.3949 grouping 0.837838 -10.395 chains 0.905263 -10.3953 measure, 0.761194 -10.3954 there are many 0.519031 -10.3957 analysis, 0.9875 -10.3961 43 0.471429 -10.3965 j. 0.830357 -10.3972 contrastive 0.560976 -10.3975 large-scale 0.987342 -10.3975 so far, 0.166489 -10.3979 baseline 0.751825 -10.3981 diverse 0.844037 -10.3983 dependency trees 0.857143 -10.3984 city 0.437346 -10.3987 steps 0.872549 -10.3988 teams 0.932584 -10.4001 atomic 0.904255 -10.4013 care 0.932584 -10.4018 theorem 0.963855 -10.4018 ing 0.632124 -10.4019 sign 0.987342 -10.4019 figure 6: 0.963855 -10.402 difficulties 0.913043 -10.4025 hierarchical phrase-based 0.836364 -10.4025 light verb 0.887755 -10.4026 composite 0.88 -10.4026 carnegie mellon 0.987342 -10.4028 inducing 0.987342 -10.4031 removes 0.987342 -10.4032 pure 0.572034 -10.4034 projection 0.952941 -10.4036 in this study, 0.932584 -10.4037 28 0.987342 -10.4041 qa 0.811966 -10.4044 collaborative 0.987342 -10.4051 increasingly 0.69375 -10.4059 tuples 0.987342 -10.4071 abbreviations 0.492114 -10.4075 expert 0.400835 -10.408 scope 0.849057 -10.4081 discourse relations 0.414798 -10.4082 lda 0.737589 -10.4097 balanced 1 -10.4098 mwes. 0.553785 -10.41 3) 0.718121 -10.4102 resources, 0.535581 -10.4112 parsing, 0.987179 -10.4112 standard deviation 0.506711 -10.4118 completely 0.598131 -10.4121 captures 0.564315 -10.4121 we call 0.676647 -10.4122 domain-specific 0.710526 -10.4128 most common 0.975 -10.413 discount 0.931818 -10.4131 accept 0.357262 -10.4139 french 0.952381 -10.4142 each other, 0.575758 -10.4143 predictive 0.894737 -10.4146 guarantee 0.535581 -10.4147 exist 0.886598 -10.4149 handled 0.855769 -10.415 predictors 0.512027 -10.4154 gains 0.975 -10.4154 in total, 0.855769 -10.4158 new york 0.987179 -10.4164 incomplete 0.544747 -10.4164 arbitrary 0.468391 -10.4165 background 0.834862 -10.4167 consideration 0.931818 -10.4172 opinion mining 0.987179 -10.4175 clustering, 0.334311 -10.4181 require 0.698718 -10.4183 efficiency 0.987179 -10.4184 relaxed 0.798319 -10.4195 released 0.602871 -10.4197 simpler 0.841121 -10.4198 rte 0.651685 -10.4199 feature vector 0.565401 -10.4199 phrasal 0.405172 -10.4199 text, 0.547244 -10.4207 indeed 0.912088 -10.421 in section 4, 0.4573 -10.4211 % 0.827273 -10.422 belongs to 0.525547 -10.4221 β 0.668639 -10.4229 by comparing 0.23508 -10.4232 what 0.893617 -10.4238 1,000 0.815789 -10.4245 prior work 0.377111 -10.4254 unique 0.987013 -10.4257 flexibility 0.321088 -10.4257 requires 0.893617 -10.4259 ssc 0.628272 -10.4263 classifying 0.877551 -10.4264 part-of-speech tags 0.668639 -10.4265 an instance 0.814159 -10.4267 es 0.680982 -10.4267 pivot 0.61194 -10.4268 experts 0.496732 -10.4268 pruning 0.412162 -10.427 efficient 0.459384 -10.4271 issues 0.61194 -10.4279 compared with 0.902174 -10.4288 divide 0.363796 -10.429 summaries 0.987013 -10.4293 candidates, 0.495114 -10.4293 cover 0.962963 -10.4296 discourse connectives 0.911111 -10.4298 predicates, 0.827273 -10.4302 an appropriate 0.987013 -10.4305 rankings 0.714286 -10.4305 estimating 0.711409 -10.4305 quickly 0.35082 -10.4306 lm 0.699346 -10.4306 recognized 0.951807 -10.4307 blogs 0.987013 -10.4309 search, 0.941176 -10.4315 relative frequency 0.780488 -10.432 cfg 0.448925 -10.4328 normalized 0.987013 -10.433 inflection 0.448925 -10.434 variation 0.655172 -10.4348 correlation between 0.709459 -10.4355 organization 0.613065 -10.4355 sampled 0.833333 -10.4361 how much 0.733813 -10.4365 yielded 0.241621 -10.4368 good 0.725352 -10.4373 prefer 0.769841 -10.4377 leaf 0.755725 -10.4379 ideas 0.482866 -10.4379 (5) 0.930233 -10.438 2), 0.231707 -10.438 patterns 0.561181 -10.439 regarding 0.940476 -10.4402 contents 0.585253 -10.4403 largely 0.75969 -10.4404 arguments, 0.81982 -10.4412 29 0.709459 -10.4413 iterative 0.514184 -10.4416 wsj 0.9625 -10.4419 1.5 0.867347 -10.4421 27 0.731884 -10.4421 checking 0.731884 -10.4427 data-driven 0.838095 -10.4434 length, 0.831776 -10.4438 numerous 0.55102 -10.4439 adjacent 0.9625 -10.4442 databases 0.455307 -10.4444 edit 0.774194 -10.4444 addresses 0.95122 -10.445 audio 0.72028 -10.4453 extensions 0.210652 -10.446 multiple 0.986842 -10.4463 morphosyntactic 0.986842 -10.4464 34 0.91954 -10.4469 “i 0.454039 -10.4475 six 0.838095 -10.4482 but rather 0.408578 -10.4484 30 0.60396 -10.4486 pronouns 0.986842 -10.4488 picture 0.715278 -10.4492 sentiment classification 0.891304 -10.4499 parsers, 0.615385 -10.4501 draw 0.910112 -10.4504 transcribed 0.503448 -10.4507 crf 0.534884 -10.4511 graph-based 0.522222 -10.4512 strongly 0.183729 -10.4514 – 0.929412 -10.4514 aa 0.71831 -10.452 shortest 0.867347 -10.4532 consisted of 0.726619 -10.4536 ground 0.986667 -10.4536 disambiguating 0.84466 -10.4537 popularity 0.974026 -10.4548 for this reason, 0.986667 -10.4548 behaviour 0.710345 -10.4554 fraction of 0.986667 -10.4556 discrepancy 0.710345 -10.4557 available, 0.584112 -10.4561 word segmentation 0.710345 -10.4563 separated 0.649425 -10.4565 cognitive 0.721429 -10.4566 exploited 0.9 -10.4573 functions, 0.986667 -10.4573 avoiding 0.950617 -10.4573 ), 0.9 -10.4575 can be easily 0.882979 -10.4576 competing 0.824074 -10.4578 happens 0.939759 -10.4586 text mining 0.950617 -10.4587 compatible with 0.986667 -10.459 demographic 0.742424 -10.4594 mixed 0.610256 -10.4601 unlabeled data 0.702703 -10.4606 modifications 0.793103 -10.461 beneficial 0.440318 -10.4613 segment 0.751938 -10.462 stem 0.803571 -10.4628 (2006), 0.793103 -10.463 heavily 1 -10.4632 module. 0.766129 -10.4637 coding 0.95 -10.4645 verified 0.603015 -10.4648 possibility 0.88172 -10.4655 automatically generated 0.664634 -10.4655 none of 0.95 -10.4662 poorly 0.588517 -10.4665 it does not 0.88172 -10.4665 binary classification 0.857143 -10.4669 f1 score 0.95 -10.4674 hiero 0.716312 -10.4677 an existing 0.679487 -10.4681 processed 0.755906 -10.4684 there has been 0.422222 -10.4692 em 0.88172 -10.4693 reporting 0.986486 -10.4695 accessible 0.344262 -10.4696 classifiers 0.736842 -10.4699 break 0.391121 -10.4703 action 0.588517 -10.4703 networks 0.775 -10.4705 very similar 0.660606 -10.4706 story 0.507143 -10.471 explored 0.961538 -10.4711 descriptive 0.522727 -10.4718 rare 0.986486 -10.4719 coordinate 0.986486 -10.472 classical 0.485342 -10.4721 descriptions 0.986486 -10.4722 worker 0.927711 -10.4723 numeric 0.986486 -10.4728 decomposed 0.633333 -10.473 anaphora 0.445983 -10.4735 units 0.693333 -10.4735 evaluation metrics 0.79646 -10.4737 inherent 0.927711 -10.4738 collapsing 0.731343 -10.4742 causes 0.440541 -10.4746 speakers 0.961538 -10.4747 ds 0.917647 -10.4748 java 0.579439 -10.4749 blog 0.496575 -10.4751 frequencies 0.465257 -10.4754 valid 0.496552 -10.477 added to 0.246414 -10.4771 they are 0.801802 -10.4786 epi 0.522901 -10.4793 roughly 0.880435 -10.4802 na¨ıve 0.773109 -10.4809 log-likelihood 0.986301 -10.481 carry out 0.85567 -10.4811 collective 0.986301 -10.4815 assessed 0.789474 -10.4818 by considering 0.758065 -10.482 serve as 0.357143 -10.4821 written 0.570776 -10.4822 estimates 0.371373 -10.483 previously 0.509091 -10.483 default 0.986301 -10.4843 heterogeneous 0.85567 -10.4844 denoted by 0.961039 -10.4848 reuters 0.729323 -10.485 indicating that 0.693878 -10.4851 relation extraction 0.609375 -10.486 defining 0.949367 -10.4862 38 0.729323 -10.4867 covering 0.606218 -10.487 machine translation, 0.789474 -10.4873 discarded 0.906977 -10.4876 ap 0.863158 -10.4877 combination. 0.664596 -10.4883 chance 0.70922 -10.4893 words) 0.916667 -10.4893 medium 0.973333 -10.4894 typing 0.916667 -10.4903 rewriting 0.532 -10.4903 relative to 0.916667 -10.4904 koehn 0.863158 -10.4907 n-best list 0.766667 -10.4907 relaxation 0.206111 -10.4908 approaches 0.432718 -10.4911 changes 0.8 -10.4916 senses, 0.52549 -10.4922 natural language processing, 0.9375 -10.4923 56 0.972973 -10.4926 wmt 2011 0.916667 -10.4927 (like 0.846939 -10.4927 schema 0.386892 -10.4933 n-grams 0.746032 -10.4935 thesaurus 0.863158 -10.4939 discrete 0.42 -10.4942 hence 0.656442 -10.4945 our experiments. 0.771186 -10.4947 complement 0.948718 -10.4948 strategies. 0.364839 -10.4949 contextual 0.88764 -10.4954 regard to 0.372781 -10.4964 discriminative 0.491409 -10.4971 d. 0.305851 -10.4976 five 0.732824 -10.498 lee 0.732824 -10.4983 motivated by 0.433155 -10.4984 to select 0.285219 -10.4985 wordnet 0.960526 -10.4986 mobile 0.428198 -10.4988 incorrect 0.771186 -10.4991 generative model 0.986111 -10.5 8, 0.547009 -10.5002 accuracy, 0.699301 -10.5003 we adopt 0.986111 -10.5004 computations 0.896552 -10.5007 adaptive 0.861702 -10.5017 stability 0.38961 -10.5019 ambiguous 0.960526 -10.5021 2009, 0.364326 -10.5031 to train 0.694444 -10.5036 values, 0.986111 -10.5039 paradigms 0.446991 -10.5048 little 0.925926 -10.5051 proportional to 0.817308 -10.5051 treelet 0.425974 -10.5053 assigned to 0.330203 -10.5054 complete 0.609626 -10.5054 management 0.670968 -10.5054 other than 0.373253 -10.5055 applications 0.151129 -10.5055 no 0.432796 -10.5057 heuristic 0.432796 -10.5059 annotator 0.948052 -10.5062 expanding 0.491349 -10.5064 context, 0.785714 -10.5072 decision tree 0.744 -10.5072 portuguese 0.47557 -10.5081 rules, 0.355839 -10.5091 (a) 0.627119 -10.5093 the importance of 0.528226 -10.5094 subjectivity 0.869565 -10.5095 france 0.404706 -10.5095 to perform 0.985915 -10.5096 boost 0.373494 -10.5101 links 0.326687 -10.5103 change 0.59 -10.5103 transition-based 0.555556 -10.5104 training corpus 0.985915 -10.5107 slow 0.725191 -10.5109 requiring 0.936709 -10.5118 ibm model 0.530364 -10.512 posts 0.570755 -10.5124 perhaps 0.96 -10.5125 user satisfaction 0.438889 -10.5127 automated 0.936709 -10.5128 53 0.517375 -10.5129 source language 0.734375 -10.5133 word-level 0.583333 -10.5135 published 0.895349 -10.5135 box 0.886364 -10.5136 52 0.481605 -10.5138 medical 0.556054 -10.5144 repeated 0.381053 -10.5147 > 0.904762 -10.5151 out-of-vocabulary 0.457576 -10.5162 we introduce 0.310924 -10.5165 sample 0.628571 -10.5167 triggers 0.96 -10.5171 by contrast, 0.747967 -10.5173 receive 0.336066 -10.5181 combination of 0.494662 -10.5182 architecture 0.620112 -10.5183 merged 0.778761 -10.5189 gaps 0.809524 -10.5189 specifies 0.444126 -10.5194 missing 0.972222 -10.5195 fix 0.708029 -10.521 introduced by 0.985714 -10.5211 as such, 0.620112 -10.5216 scfg 0.389381 -10.5218 primary 0.752066 -10.5218 development set. 0.836735 -10.5222 premise 0.376033 -10.5227 understanding 0.503704 -10.5229 paths 0.836735 -10.5232 there are several 0.559633 -10.5235 notion 0.778761 -10.5237 non-projective 0.297542 -10.524 now 0.261477 -10.5245 processing 0.677852 -10.5245 offer 0.985714 -10.5247 beam-search 0.851064 -10.5249 msa 0.925 -10.5253 searches 0.947368 -10.5259 dependency-based 0.885057 -10.526 adjective-noun 0.851064 -10.5261 stories 0.752066 -10.5267 monotone 0.719697 -10.5273 domains, 0.556561 -10.5273 chart 0.985714 -10.5276 implementations 0.947368 -10.5281 readily 0.821782 -10.5282 64 0.858696 -10.5284 ai 0.807692 -10.5285 teacher 0.695035 -10.5285 35 0.395402 -10.5292 his 0.985714 -10.5294 hypothesized 0.745902 -10.5295 minor 0.947368 -10.5295 95 0.400943 -10.5298 gene 0.903614 -10.5301 ratios 0.367793 -10.5304 to obtain 0.673333 -10.5306 supports 0.337229 -10.5307 lists 0.55157 -10.5308 to construct 0.788991 -10.5308 2000), 0.985714 -10.5311 aims at 0.972222 -10.5313 collection, 0.394077 -10.5314 open 0.924051 -10.5319 projective 0.756303 -10.5319 sibling 0.985714 -10.5327 categorized 0.91358 -10.5333 grammar induction 0.765217 -10.5345 median 0.726562 -10.5346 exploring 0.82 -10.5354 determiners 0.334992 -10.5355 representations 0.602151 -10.5359 to assess 0.528926 -10.5374 500 0.8 -10.5374 service 0.430137 -10.5378 semantically 0.314788 -10.5378 our model 0.892857 -10.539 scoring function 0.658065 -10.5394 discovered 0.971831 -10.5397 (both 0.958904 -10.5405 to this end, 0.985507 -10.5405 language-specific 0.849462 -10.5406 covered by 0.417098 -10.5406 but not 0.902439 -10.5409 ease 0.765217 -10.5415 stated 0.577114 -10.5417 amounts of 0.892857 -10.542 (all 0.453988 -10.5421 we see 0.849462 -10.5422 high quality 0.459119 -10.5424 children 0.55814 -10.5427 optimized 0.743802 -10.5428 identical to 0.902439 -10.5429 indicated by 0.226994 -10.5435 across 0.774775 -10.5436 indexing 0.61236 -10.5436 maps 0.958904 -10.544 twitter, 0.883721 -10.544 to, 0.611111 -10.5442 indicators 0.515873 -10.5446 varying 0.758621 -10.5449 interpolated 0.671141 -10.545 titles 0.892857 -10.545 implementing 0.544248 -10.5452 indicate that 0.934211 -10.5455 personal data 0.641975 -10.5455 an alternative 0.923077 -10.5455 markup 0.534188 -10.5456 efficiently 0.685315 -10.5457 g. 0.934211 -10.5458 np, 0.985507 -10.5463 roles, 0.9125 -10.5464 simple, 0.985507 -10.5464 signals 0.461783 -10.5467 induced 0.695652 -10.5469 phrase, 0.627219 -10.548 modifiers 0.6 -10.548 tokens, 0.985507 -10.5481 blocks 0.9125 -10.5482 feature functions 0.763158 -10.5484 wu 0.292092 -10.5486 prediction 0.580808 -10.549 to ensure 0.453704 -10.5497 early 0.660131 -10.5498 theme 0.857143 -10.5501 39 0.653846 -10.5501 subtree 0.35687 -10.5502 h 0.902439 -10.5505 ad 0.190038 -10.5509 6 0.570732 -10.5512 shift 0.728 -10.5513 75 0.763158 -10.5515 guided 0.171441 -10.553 + 0.855556 -10.5538 broader 0.985294 -10.5538 ssl 0.498127 -10.5541 templates 0.505792 -10.5558 versus 0.737705 -10.556 (chiang, 0.412821 -10.5566 issue 0.464286 -10.5566 target language 0.767857 -10.558 convey 0.934211 -10.558 54 0.911392 -10.5582 i am 0.824742 -10.5582 forms, 0.366397 -10.5582 constraint 0.985294 -10.5584 70% 0.634146 -10.5585 recognizing 0.690647 -10.5586 lin 0.628743 -10.5586 seven 0.745763 -10.5586 intuitive 0.651613 -10.5597 it, 0.713178 -10.5598 insertion 0.745763 -10.5608 quantitative 0.559242 -10.5614 encoding 0.586387 -10.5616 mi 0.901235 -10.5617 initialization 0.761062 -10.5619 crowdsourcing 0.204301 -10.5623 within 0.370833 -10.5627 entries 0.933333 -10.5628 serves as 0.465347 -10.5629 internal 0.25 -10.5633 inference 0.31938 -10.5635 tweets 1 -10.564 vice versa. 0.933333 -10.5642 dependency parsing, 0.639752 -10.5648 doing 0.816327 -10.565 urdu 0.765766 -10.5673 previous work on 0.547945 -10.5675 because they 0.360396 -10.5676 structured 0.783019 -10.5677 slight 0.699248 -10.5681 hpsg 0.503876 -10.5686 approaches. 0.853933 -10.5686 checked 0.911392 -10.5687 phrase tables 0.922078 -10.5689 trials 0.853933 -10.5691 removed. 0.621302 -10.5691 phonetic 0.971014 -10.5702 human judges 0.872093 -10.5704 46 0.9 -10.5707 gestures 0.945205 -10.5711 week 0.777778 -10.5711 ontologies 0.796117 -10.5712 products 0.273455 -10.5714 morphological 0.553991 -10.5717 acquisition 0.520833 -10.5723 constrained 0.346863 -10.5724 studies 0.945205 -10.5729 markable 0.410853 -10.5734 stanford 0.72 -10.5738 associations 0.783019 -10.5747 consensus 0.441441 -10.5748 becomes 0.985075 -10.5753 templates. 0.985075 -10.5755 female 0.862069 -10.5756 young 0.985075 -10.5757 drawing 0.985075 -10.5759 capitalization 0.418919 -10.5767 bayesian 0.348231 -10.5767 transition 0.758929 -10.5772 grounding 0.890244 -10.5781 video 0.668966 -10.5785 spaces 0.788462 -10.5787 goals 0.829787 -10.5792 maximizing 1 -10.5792 construction. 0.910256 -10.5794 cooccurrence 0.540541 -10.58 constituents 0.932432 -10.5804 dependencies, 0.546296 -10.5804 a wide 0.747826 -10.5805 rule extraction 0.551887 -10.5808 spatial 0.560976 -10.5808 intermediate 0.336842 -10.5808 related work 0.53304 -10.5812 john 0.788462 -10.582 factorization 0.546296 -10.5824 suggest that 0.944444 -10.5832 speaker’s 0.710938 -10.5833 representation, 0.944444 -10.5834 web-scale 0.447205 -10.5837 adjectives 0.295148 -10.5838 potential 0.710938 -10.584 transformations 0.63522 -10.584 to classify 0.8 -10.5843 translation quality. 0.653333 -10.5845 store 0.33107 -10.5845 extended 0.984848 -10.5858 eliminating 0.641026 -10.5859 ∼ 0.546296 -10.5869 evaluations 0.821053 -10.5872 differ from 0.888889 -10.5872 4.5 0.510121 -10.5874 mert 0.63522 -10.5874 records 0.624242 -10.5874 achieving 0.22265 -10.5875 errors 0.741379 -10.588 algorithms, 0.860465 -10.5881 pos tagger 0.984848 -10.5883 triggered 0.673759 -10.5883 more general 0.852273 -10.5884 integrated into 0.984848 -10.5888 duplicate 0.984848 -10.5891 frequency, 0.688889 -10.5905 michael 0.508065 -10.5907 directed 0.189568 -10.5907 simple 0.407792 -10.5909 messages 0.957143 -10.5911 principles 0.373102 -10.5912 link 0.835165 -10.5916 post-processing 0.0965678 -10.5921 1 0.433628 -10.5923 parses 0.649007 -10.5928 canonical 0.909091 -10.5929 imply 0.852273 -10.5929 labelling 0.659864 -10.593 syntactically 0.636943 -10.5932 incorporated 0.444444 -10.5935 m. 0.422969 -10.5937 gender 0.279268 -10.5939 base 0.659864 -10.594 restrictions 0.620482 -10.5943 david 0.476868 -10.5943 documents, 1 -10.5947 operations. 1 -10.5947 technique. 0.5 -10.5948 ten 0.703125 -10.595 black 0.572165 -10.5957 γ 0.842697 -10.5958 adverbs 0.943662 -10.5959 peer 0.253266 -10.596 generation 0.909091 -10.5961 coupled 0.575916 -10.5962 extracts 0.827957 -10.5964 variables, 0.761468 -10.5965 possible, 0.536036 -10.5966 word pairs 0.63871 -10.5971 consecutive 0.467354 -10.5971 free 0.984615 -10.5972 worth noting 0.556098 -10.5975 framenet 0.786408 -10.5977 extrinsic 0.878049 -10.5979 exclude 0.731092 -10.5983 judgements 0.984615 -10.5984 sparseness 0.943662 -10.5986 participate in 0.75 -10.5988 human judgments 0.761468 -10.5988 wiki 0.931507 -10.5994 trial 0.544186 -10.5994 gold-standard 0.756757 -10.5996 rarely 0.850575 -10.5997 limits 0.678832 -10.6002 has no 0.773585 -10.6002 contribute to 0.984615 -10.6006 layers 0.388095 -10.6015 as shown in 0.984615 -10.6016 logical forms 0.984615 -10.6017 non-zero 0.441358 -10.6021 english, 0.703125 -10.6021 2000 0.956522 -10.6022 false positives 0.715447 -10.6031 and lapata, 0.73913 -10.6033 click 0.970149 -10.6035 antecedents 0.931507 -10.6036 productive 0.878049 -10.6038 gender, 0.608187 -10.6041 day 0.640523 -10.6056 inferred 0.984615 -10.6063 co-occurrences 0.664336 -10.6064 judgment 0.819149 -10.6067 deeper 0.686567 -10.6068 labelled 0.819149 -10.607 individuals 0.984615 -10.6072 78 0.706349 -10.6073 unified 0.116959 -10.6074 some 0.728814 -10.6074 polysemous 0.743363 -10.6076 sharing 0.339416 -10.6079 twitter 0.678832 -10.6084 phrase-based translation 0.858824 -10.6086 spoken language 0.984615 -10.6086 weakly 0.833333 -10.6086 states, 0.930556 -10.6096 coverage, 0.689394 -10.6096 section 5. 1 -10.6105 criteria. 0.858824 -10.6106 svms 0.766355 -10.6108 sub-word 0.804124 -10.611 orthographic 0.706349 -10.6112 shown to be 0.587912 -10.6119 made by 0.817204 -10.6122 days 0.625 -10.6124 date 0.747748 -10.6138 hidden markov 0.580645 -10.6139 newswire 0.817204 -10.6141 regions 0.907895 -10.6144 storage 0.942857 -10.6145 well-formed 0.676471 -10.6147 collections 0.955882 -10.6147 quantifiers 0.876543 -10.615 ignored 0.502024 -10.6153 somewhat 0.795918 -10.6154 computer science, 0.955882 -10.6155 citations 0.886076 -10.6156 scenarios 0.341418 -10.6156 loss 0.984375 -10.6159 dominated by 0.984375 -10.6161 unambiguous 0.930556 -10.6162 unrestricted 0.776699 -10.6164 die 0.743363 -10.6175 completed 0.695312 -10.6188 plot 0.570681 -10.6188 substantially 0.57754 -10.6189 ie 0.695312 -10.619 force 0.984375 -10.6192 runtime 0.687023 -10.6195 secondary 0.824176 -10.6196 resolved 0.810526 -10.6197 reliably 0.984375 -10.6198 groups, 0.984375 -10.62 arg max 0.984375 -10.6201 1997), 0.984375 -10.6204 well known 0.610778 -10.6205 schemes 0.457912 -10.6208 opinions 0.776699 -10.6209 saw 0.984375 -10.6209 modest 0.655172 -10.6215 as shown in table 0.984375 -10.6219 declarative 0.929577 -10.622 gives us 0.984375 -10.6222 responsible for 0.382075 -10.6223 outputs 0.776699 -10.6232 xml 0.917808 -10.6238 (by 0.906667 -10.6238 fewer than 0.60355 -10.6244 trees, 0.752294 -10.6249 two main 0.657343 -10.6254 rhetorical 0.747748 -10.6255 wikipedia, 0.896104 -10.6255 precision, recall 0.650685 -10.6257 keywords 0.875 -10.6257 defined as: 0.942029 -10.6258 messages. 0.969231 -10.6261 listing 0.984375 -10.6261 probability mass 0.984375 -10.6262 looked at 0.362288 -10.6264 in particular, 0.622642 -10.6265 interactive 0.555 -10.6269 jointly 0.466431 -10.6273 auxiliary 0.456081 -10.6277 1000 0.527027 -10.6278 en 0.698413 -10.6282 we are interested 0.757009 -10.6285 it’s 0.515021 -10.629 would have 0.657343 -10.6291 language model, 0.144697 -10.6294 5 0.370536 -10.6294 these two 0.646259 -10.6295 restricted to 0.745455 -10.63 superior 0.942029 -10.6305 binarization 0.793814 -10.6306 implies that 0.256111 -10.6308 name 0.917808 -10.6313 1.2 0.506276 -10.6316 generalized 0.984127 -10.632 capability 0.793814 -10.6322 an open 0.787879 -10.6323 sentences) 0.917808 -10.6324 fi 0.61875 -10.6325 ngram 0.44918 -10.6325 aspects of 0.955224 -10.633 deleted 0.587571 -10.6334 propbank 0.6 -10.6335 induce 0.761905 -10.6335 thousands of 0.917808 -10.6347 reorderings 0.681818 -10.6347 definite 0.955224 -10.6353 60% 0.469314 -10.6357 annotated with 0.929577 -10.6357 pose 0.793814 -10.636 wordnet, 0.984127 -10.6361 exhibits 0.730435 -10.6363 questions, 0.648276 -10.6363 logic 0.905405 -10.6365 found, 0.929577 -10.6365 infrequent 0.402116 -10.6368 unknown 0.613497 -10.6369 (8) 0.984127 -10.6371 inferring 0.984127 -10.6371 possibilities 0.822222 -10.6374 42 0.354639 -10.6375 regression 0.477612 -10.6376 to estimate 0.984127 -10.6378 ms 0.60479 -10.6382 propositions 0.793814 -10.6382 much larger 0.984127 -10.6384 delete 0.984127 -10.6385 miss 0.676692 -10.6385 prepositional 0.591954 -10.6386 7. 0.701613 -10.6387 it should be 0.276596 -10.6393 structures 0.696 -10.6393 cells 0.984127 -10.6394 developers 0.522321 -10.6396 considered as 0.864198 -10.64 bootstrap 0.565445 -10.6405 dependency tree 0.864198 -10.6407 someone 0.738739 -10.6414 translated into 0.984127 -10.6416 utilizes 0.75 -10.6417 (where 0.714286 -10.6418 figure 3. 0.873418 -10.6423 beam search 0.894737 -10.6424 mturk 0.692913 -10.643 our goal is to 0.317355 -10.643 parsers 0.954545 -10.643 preliminary experiments 0.983871 -10.6431 pointed out 0.176561 -10.6433 after 0.469091 -10.6434 adapted 0.407104 -10.6436 boundaries 0.657143 -10.6437 expressing 0.717949 -10.6443 1-best 0.470803 -10.6454 theory 0.495935 -10.6454 symbols 0.419075 -10.6457 employ 0.941176 -10.6457 prediction, 0.699187 -10.6457 significant improvement 0.664234 -10.6459 stochastic 0.928571 -10.6462 piece 0.772277 -10.6463 terminology 0.405995 -10.6463 overlap 0.559585 -10.6464 strict 0.983871 -10.647 transduction 0.893333 -10.6471 statistics. 0.415954 -10.6471 again 0.772277 -10.6473 inflected 0.983871 -10.648 broadly 0.520179 -10.6481 learns 0.556122 -10.6481 table 2 shows 0.666667 -10.6488 section 3. 0.853659 -10.6492 et. 0.983871 -10.6494 exclusively 0.621795 -10.6494 curves 0.941176 -10.6503 47 0.845238 -10.6503 compressions 0.941176 -10.6504 bases 0.415473 -10.6506 affect 0.983871 -10.6508 part-of-speech tagging 0.983871 -10.6511 emphasize 0.219649 -10.6516 discourse 0.357294 -10.6517 accurate 0.627451 -10.6518 linguistically 0.797872 -10.6518 sentential 0.845238 -10.6518 plain 0.797872 -10.6519 except that 0.983871 -10.653 noise, 0.345924 -10.6531 design 0.34104 -10.6531 dependencies 0.941176 -10.6531 readings 0.633333 -10.6532 expectation 0.983871 -10.6537 project, 0.983871 -10.6538 0 0 0 . 0.983871 -10.6544 indirectly 0.218427 -10.6549 7 0.445545 -10.655 transliteration 0.813187 -10.6552 non-native 0.777778 -10.6554 scenes 0.547264 -10.6554 clark 0.608696 -10.6555 to translate 0.983871 -10.6557 group, 0.915493 -10.6557 time complexity 0.645833 -10.6558 te 0.248205 -10.6562 category 0.504237 -10.6567 machine translation. 0.725664 -10.6569 ambiguities 0.915493 -10.6569 nothing 0.661765 -10.6569 haitian 0.629139 -10.657 elementary 0.759615 -10.6571 sub 0.721739 -10.6592 concerning 0.789474 -10.6596 referring to 0.6 -10.66 van 0.811111 -10.6611 assuming that 0.383292 -10.6613 (and 0.635135 -10.6614 books 0.702479 -10.6615 order, 0.647887 -10.6623 multimodal 0.835294 -10.6626 reflecting 0.587209 -10.6626 mutual information 0.881579 -10.6631 an incremental 0.541872 -10.6633 axioms 0.968254 -10.6634 factors, 0.610063 -10.6634 i. 0.329091 -10.6634 to determine 0.983607 -10.6634 news commentary 0.927536 -10.664 ever 0.715517 -10.664 co-training 0.927536 -10.6651 psychological 0.613924 -10.6653 (see table 0.983607 -10.6654 discrimination 0.983607 -10.6656 74 0.983607 -10.6657 extensively 0.518018 -10.6657 guidelines 0.940299 -10.6661 timing 0.953846 -10.667 orthogonal 0.745283 -10.6671 feature weights 0.475285 -10.6672 explain 0.579545 -10.6672 examples, 0.818182 -10.6673 suggesting that 0.583815 -10.6674 predicts 0.983607 -10.6674 annotators, 0.881579 -10.6674 omitted 0.953846 -10.6681 previous work, 0.671756 -10.6683 inverse 0.715517 -10.6684 decipherment 0.568306 -10.6688 data set. 0.542289 -10.6692 labeled data 0.705882 -10.6696 initially 0.983607 -10.6696 symbolic 0.557895 -10.6701 each of these 0.851852 -10.6706 connotative 0.818182 -10.6707 tts 0.235294 -10.6708 / 0.983607 -10.671 proportions 0.620915 -10.6711 clean 0.983607 -10.6717 buffer 0.691057 -10.6718 language technology 0.579545 -10.6719 toward 0.679688 -10.6719 changing 0.843373 -10.6719 calls 0.953846 -10.6731 phonological 0.523148 -10.6732 zhang 0.926471 -10.6733 ! 0.542289 -10.6734 variations 0.983607 -10.6736 kn 0.860759 -10.6736 mt systems 0.860759 -10.6746 topic modeling 0.843373 -10.6746 retain 0.983607 -10.6749 66 0.301075 -10.675 predicted 0.983607 -10.6751 pronouns, 0.5311 -10.6752 (koehn et al., 0.186761 -10.6758 automatic 0.914286 -10.6765 shape 0.914286 -10.6768 studying 0.818182 -10.6769 trigrams 0.825581 -10.677 error detection 0.87013 -10.677 movement 0.298485 -10.6777 polarity 0.78125 -10.678 thousand 0.767677 -10.6784 unsupervised learning 0.468165 -10.6791 degree of 0.628378 -10.6791 looking at 0.634483 -10.6794 assignments 0.5 -10.6796 we need to 0.64539 -10.6797 created by 0.5 -10.68 srl 0.773196 -10.6802 source-side 0.694215 -10.6805 κ 0.562162 -10.6806 approaches, 0.677165 -10.6808 grammaticality 0.890411 -10.6809 exp 0.983333 -10.681 dedicated 0.266344 -10.6814 certain 0.685484 -10.6819 (if 0.41994 -10.682 outperform 0.833333 -10.6824 proof 0.233055 -10.6827 tokens 0.694215 -10.6827 to collect 0.767677 -10.6832 to implement 0.585799 -10.6833 across all 0.802198 -10.6834 formally 0.85 -10.6836 (klein and 0.416667 -10.6838 iterations 0.52093 -10.6838 modification 0.983333 -10.6839 gradually 0.397297 -10.6844 al 0.504348 -10.6844 morphology 0.455516 -10.6854 she 0.983333 -10.6856 insufficient 0.54 -10.6859 numerical 0.438944 -10.6863 24 0.773196 -10.6867 really 0.983333 -10.6869 participation 0.630137 -10.6871 phone 0.299383 -10.6874 probabilistic 0.720721 -10.6876 bow 0.368298 -10.6879 interpretation 0.324275 -10.6879 split 0.967742 -10.6881 answered 0.983333 -10.6886 decision trees 0.983333 -10.6896 segmenter 0.473077 -10.6898 preceding 0.559783 -10.6907 sl 0.983333 -10.6908 suffer from 0.88 -10.691 pantel, 0.793478 -10.6912 lexical resources 0.461255 -10.6913 segments 0.773196 -10.6913 pagerank 0.649635 -10.6916 kb 0.938462 -10.692 messages, 0.671875 -10.6924 (p < 0.742857 -10.6926 very high 0.76 -10.6926 ti 0.913043 -10.6932 indicative 1 -10.6932 induction. 1 -10.6932 people. 0.224348 -10.6933 n-gram 0.967213 -10.6933 danish 0.511312 -10.6935 independently 0.356044 -10.6936 longer 0.230699 -10.6937 initial 0.54359 -10.694 challenging 0.195638 -10.6941 gold 0.180236 -10.6944 examples 0.476378 -10.6945 comparative 0.537688 -10.6945 baseline system 0.967213 -10.6945 characterized by 0.586826 -10.695 shorter 0.491597 -10.6955 much more 0.608974 -10.6956 pmi 0.578035 -10.6961 we follow 0.765306 -10.6963 embedding 0.682927 -10.6964 employing 0.485714 -10.6966 p. 0.714286 -10.6966 smaller than 0.868421 -10.6966 professional 0.8 -10.697 literature, 0.595092 -10.6974 bacteria 0.823529 -10.6977 mpqa 0.631944 -10.698 transitive 0.823529 -10.6981 starting with 0.720721 -10.6982 rule, 0.967213 -10.6983 at all. 0.720721 -10.6986 organized as follows. 0.476378 -10.6992 play 0.888889 -10.6994 58 0.8 -10.6996 language pairs. 0.446367 -10.6997 the impact of 0.938462 -10.7 conjunctions 0.700855 -10.7002 most similar 0.306557 -10.7003 identification 0.415663 -10.7008 semantic role 0.967213 -10.7016 wide variety 0.983051 -10.7021 (for example, 0.9 -10.7022 particle 0.301109 -10.7022 distributions 0.236791 -10.7022 mt 0.983051 -10.7024 acquiring 0.983051 -10.7027 rs 0.983051 -10.7031 deemed 0.857143 -10.7036 as well, 0.740385 -10.7036 iterations, 0.983051 -10.7038 liu et al. 0.868421 -10.704 wikipedia articles 0.813953 -10.7042 2006 0.654135 -10.7042 markers 0.348195 -10.7046 review 0.983051 -10.7047 satisfied 0.952381 -10.705 recovery 0.878378 -10.7052 raters 0.839506 -10.7054 updating 0.323583 -10.7056 resource 0.627586 -10.7061 entirely 0.544041 -10.7061 gave 0.322993 -10.7063 testing 0.389333 -10.7066 despite 0.502222 -10.7067 nearly 0.704348 -10.7072 exhaustive 0.328302 -10.7075 character 0.564246 -10.7077 fields 0.888889 -10.7077 semi-supervised learning 0.274769 -10.7078 paraphrase 0.983051 -10.7091 81 0.951613 -10.7097 converted into 0.924242 -10.7102 posterior probability 0.757576 -10.7103 appendix 0.597484 -10.7104 held-out 0.538462 -10.7104 letters 1 -10.7106 themselves. 0.283099 -10.7107 recognition 0.848101 -10.7108 140 0.966667 -10.711 diagonal 0.689076 -10.7114 filled 0.436242 -10.7116 literature 0.623288 -10.7116 discussed in section 0.355556 -10.7117 explicitly 0.59375 -10.7117 semantic relations 0.866667 -10.7118 assumes that 0.300477 -10.712 whereas 0.554348 -10.7121 perceptron 0.554348 -10.7127 phenomenon 0.550802 -10.7127 me 0.757576 -10.7128 65 0.876712 -10.7133 procedures 0.698276 -10.7133 linked to 0.385827 -10.7135 appear in 0.689076 -10.7136 training examples 0.745098 -10.7137 suggestions 0.580838 -10.7138 makes it 0.259833 -10.714 m 0.924242 -10.7142 td 0.876712 -10.7143 se 0.585366 -10.7155 any other 0.238 -10.7158 techniques 0.804598 -10.7158 discovering 0.887324 -10.7159 summarizing 0.361111 -10.716 yields 0.982759 -10.7164 downstream 0.966667 -10.7164 (that 0.982759 -10.7166 lists, 0.194892 -10.7169 thus 0.692308 -10.717 existence of 0.726415 -10.717 segmentation. 0.692308 -10.7176 affects 0.911765 -10.7179 correlations between 0.782609 -10.7179 semantic relatedness 0.733333 -10.7182 is equivalent to 0.982759 -10.7184 rose 0.737864 -10.7187 utilizing 0.9375 -10.719 self 0.768421 -10.7193 anaphoric 0.674797 -10.7194 sort 0.619048 -10.7195 to maximize 0.755102 -10.7195 aware of 0.198054 -10.7195 automatically 0.898551 -10.7199 this. 0.722222 -10.7203 moving 0.637681 -10.7205 source side 0.982759 -10.7208 convenient 0.439863 -10.7208 however 0.726415 -10.7208 thresholds 0.472222 -10.721 methods, 0.637681 -10.7212 target side 0.982759 -10.7214 parametric 0.982759 -10.7216 lookup 0.755102 -10.7225 target-side 0.982759 -10.7228 59 0.982759 -10.7229 mutually 0.75 -10.7231 helpfulness 0.755102 -10.7231 55 0.590062 -10.7234 one, 0.966667 -10.7235 semantically related 0.686441 -10.7235 release 0.737864 -10.724 not yet 0.473896 -10.7246 say 0.485232 -10.725 lines 0.701754 -10.7253 demonstrated that 0.485232 -10.7254 subsequent 0.674797 -10.7258 in table 1, 0.591195 -10.7262 morpheme 0.742574 -10.7266 similarity measure 0.658915 -10.7267 speaking 0.910448 -10.7267 an approximation 0.887324 -10.7269 round 0.760417 -10.7272 quantity 0.898551 -10.7278 1st 0.936508 -10.7281 adjust 0.316071 -10.7282 state-of-the-art 1 -10.7283 segments. 0.982456 -10.7285 authorship attribution 0.695652 -10.7286 keeping 0.936508 -10.7288 described above 0.898551 -10.7289 graphs, 0.142545 -10.7289 rules 0.268041 -10.7291 string 0.755102 -10.7292 bits 0.760417 -10.7293 class labels 0.875 -10.7293 players 0.864865 -10.7295 as can be seen 0.755102 -10.7296 this paper describes 0.719626 -10.7303 as a result 0.32216 -10.7308 we consider 0.855263 -10.7313 longer than 0.689655 -10.7319 subsets of 0.936508 -10.732 functionality 0.591195 -10.7322 we ran 0.248062 -10.7322 verbs 0.982456 -10.7323 noisy channel 0.542553 -10.7325 conversational 0.735294 -10.7325 development, 0.966102 -10.7326 speak 0.578313 -10.7326 p < 0.442105 -10.7327 suitable 0.885714 -10.7328 latent variables 0.885714 -10.7335 compose 0.966102 -10.7339 attributed to 0.835443 -10.7339 system: 0.601307 -10.7339 ir 0.885714 -10.7341 opinionated 0.318099 -10.7348 functions 0.966102 -10.736 bionlp’09 shared 0.419048 -10.7364 simplification 0.563218 -10.7365 classification, 0.982456 -10.7365 ultimately 0.332008 -10.737 considering 0.982456 -10.737 symbols, 0.5875 -10.7374 acquired 0.138002 -10.7375 many 0.786517 -10.7375 variational 0.786517 -10.7375 ei 0.982456 -10.7377 token-level 0.300163 -10.7379 people 0.72381 -10.7379 l2 0.304714 -10.7385 applying 0.699115 -10.7386 stable 0.844156 -10.7387 l, 0.817073 -10.739 a., 0.982456 -10.7391 shows, 0.885714 -10.7392 bulgarian 0.982456 -10.7394 crossing 0.936508 -10.7397 germany 0.864865 -10.7398 bracketing 0.384 -10.74 increasing 0.875 -10.7401 ii) 0.853333 -10.7403 automatically extracted 0.853333 -10.7406 suggested by 0.624113 -10.7412 creation of 0.921875 -10.7414 most popular 0.47541 -10.7415 we take 0.404762 -10.7416 handle 0.608108 -10.7416 notice that 0.72381 -10.742 this kind of 0.875 -10.7426 p-value 0.982456 -10.7428 productions 0.716981 -10.743 mechanical turk 0.982456 -10.7431 hmms 0.583851 -10.7431 research, 0.95082 -10.7431 ensures that 0.875 -10.7434 returned by 0.716981 -10.7435 similarity, 0.809524 -10.7436 rescoring 0.65873 -10.7438 maxent 0.853333 -10.7439 freebase 0.935484 -10.7441 (of 0.809524 -10.7443 obj 0.702703 -10.7446 tries 0.525253 -10.745 terminal 0.272237 -10.7456 support 0.272237 -10.7457 to find 0.384824 -10.7458 yet 0.489083 -10.7458 errors, 0.921875 -10.7458 bin 0.777778 -10.7459 insert 0.453184 -10.7459 table 2. 0.514563 -10.7461 production 1 -10.7463 ontology. 0.853333 -10.7463 far from 0.982143 -10.7463 linguistically motivated 0.613793 -10.7464 articles, 0.897059 -10.7467 ep 0.28614 -10.7468 rank 0.386921 -10.7468 to predict 0.873239 -10.7472 batch 0.965517 -10.7473 1.1 0.921875 -10.7473 features) 0.335391 -10.7474 gain 0.833333 -10.748 significantly outperforms 0.568047 -10.7492 knowledge. 0.702703 -10.7492 usual 0.497738 -10.7493 straightforward 0.853333 -10.7494 stemming 0.863014 -10.7499 feature feedback 0.728155 -10.7499 clustered 0.909091 -10.7503 their respective 0.431507 -10.7506 incremental 0.514563 -10.7509 supporting 0.982143 -10.7513 human-human 0.935484 -10.7515 real-time 0.530928 -10.7518 an initial 0.75 -10.7519 script 0.982143 -10.7519 penalize 0.95 -10.7522 partitions 0.982143 -10.7522 seeks 0.982143 -10.7524 modeling, 0.982143 -10.7524 ranking, 0.784091 -10.7524 namely, 0.884058 -10.7525 2008. 0.572289 -10.7528 described by 0.280172 -10.7529 queries 0.982143 -10.7529 combination, 0.807229 -10.7529 3.0 0.833333 -10.753 an efficient 0.475 -10.7534 a total of 0.467742 -10.7536 acoustic 0.982143 -10.7537 intuitively 0.814815 -10.7538 commercial 0.814815 -10.754 corpus-based 0.240938 -10.7541 decoding 0.287234 -10.7544 disambiguation 0.493274 -10.7544 assuming 0.427609 -10.7545 to achieve 0.14083 -10.7548 methods 0.982143 -10.7549 reconcile 0.227836 -10.7549 observed 0.935484 -10.7551 identification, 0.429054 -10.7556 reliable 0.639098 -10.7557 muc 0.825 -10.7559 coverage. 0.982143 -10.756 invalid 0.884058 -10.7566 though, 0.95 -10.7569 pro 0.557471 -10.7569 mechanism 0.982143 -10.7572 signature modules 0.605442 -10.7576 to annotate 0.41195 -10.7582 plan to 0.459144 -10.7582 effort 0.641221 -10.7583 areas 0.537234 -10.7583 themselves 0.982143 -10.7585 filling 0.769231 -10.7591 binomial 0.568862 -10.7595 faster 0.851351 -10.76 subtypes 0.641221 -10.7601 optimizing 0.737374 -10.7604 implicitly 0.690265 -10.7604 replacement 0.982143 -10.7604 normalised 0.982143 -10.7608 c-lda 0.833333 -10.7608 elements. 0.681034 -10.7608 collecting 0.822785 -10.761 causality 0.907692 -10.7613 (galley 0.982143 -10.7616 theory, 0.982143 -10.7623 92 0.982143 -10.7623 73 0.895522 -10.7627 communities 0.755319 -10.7628 metadata 0.62963 -10.7628 constituency 0.384615 -10.7629 attributes 0.72549 -10.7629 tfidf 0.350797 -10.7635 tuning 0.351724 -10.7638 strategies 0.507177 -10.7638 in-domain 0.601351 -10.7642 team 0.243392 -10.7645 learned 0.605442 -10.7647 (including 0.934426 -10.7647 parse trees. 0.797619 -10.7649 most relevant 0.907692 -10.7649 of) 0.769231 -10.7653 ranker 0.188034 -10.7655 see 0.684211 -10.7658 company 0.907692 -10.766 him 0.176301 -10.7666 statistical 0.964912 -10.7667 (no 0.949153 -10.7668 for this purpose, 0.43617 -10.7668 modules 0.493213 -10.767 simplified 0.861111 -10.768 resolution, 0.284639 -10.7682 counts 0.831169 -10.7685 p r f1 0.25 -10.7687 : 0.57764 -10.7693 x. 0.220217 -10.7694 (e.g. 0.851351 -10.7694 entails 0.703704 -10.7699 marked as 0.981818 -10.7701 helping 0.76087 -10.7702 to our knowledge, 0.981818 -10.7705 filler 0.426621 -10.7707 mainly 0.981818 -10.7707 insight into 0.949153 -10.7709 uturku 0.934426 -10.7709 diseases 0.920635 -10.7712 measurement 0.286371 -10.772 vocabulary 0.366834 -10.7724 exactly 0.964912 -10.7727 wmt11 0.981818 -10.7729 moved 0.781609 -10.7731 sentences containing 0.964912 -10.7732 random sample 0.981818 -10.7734 links, 0.920635 -10.7735 turkish 0.920635 -10.7738 pipelines 0.861111 -10.774 cky 0.981818 -10.774 reaching 0.509804 -10.7741 we chose 0.73 -10.7745 reasonably 0.981818 -10.7745 individually 0.949153 -10.7748 water 0.734694 -10.7749 start with 0.981818 -10.775 capital 0.684211 -10.7752 liu, 0.8125 -10.7758 distributions over 0.981818 -10.7761 parameterized 0.804878 -10.7762 phrase-based smt 0.722772 -10.7763 pre-processing 0.363184 -10.7766 google 0.73 -10.7767 affective 0.578616 -10.7768 pa 0.517588 -10.7769 software 0.781609 -10.7771 three types of 0.193142 -10.7775 under 0.40367 -10.7778 computer science 0.893939 -10.778 cast 0.981818 -10.7781 fn 0.567073 -10.7783 comparison with 0.934426 -10.7784 (lit. 0.308797 -10.7785 real 0.697248 -10.779 to derive 0.919355 -10.7791 57 0.255215 -10.7792 university of 0.8125 -10.7793 10,000 0.949153 -10.7799 inform 0.352941 -10.7799 graphs 0.739583 -10.7801 suffixes 0.707547 -10.7802 rst 0.697248 -10.7803 parallel corpus. 0.981818 -10.7805 ambiguity, 0.84 -10.7809 compressed 0.882353 -10.7809 preprocessed 0.766667 -10.7811 collapsed 0.981818 -10.7812 89 0.919355 -10.7814 hand-crafted 0.266399 -10.7823 help 0.177804 -10.7829 domain 0.681416 -10.783 ignore 0.766667 -10.783 participated in 0.507317 -10.783 preferences 0.711538 -10.7831 reveal 0.933333 -10.7833 portions of 1 -10.7833 practice. 0.551724 -10.7835 ace 0.387464 -10.7835 noise 0.608392 -10.7837 categorization 0.849315 -10.7838 mira 0.752688 -10.7839 reaction 0.562874 -10.7841 exploiting 0.567073 -10.7842 greedy 0.559524 -10.7842 locations 0.828947 -10.7842 creole 0.84 -10.7845 strengths 0.681416 -10.7846 speech, 0.869565 -10.7849 how often 0.933333 -10.7851 another, 0.919355 -10.7852 foreign language 0.90625 -10.7852 label propagation 0.529412 -10.7862 conversion 0.202808 -10.7863 overall 0.77907 -10.7871 obviously 0.398792 -10.7873 because it 0.90625 -10.7873 1999), 0.948276 -10.7875 snippet 0.981481 -10.7878 high-quality 0.325911 -10.7878 q 0.715686 -10.7879 now, 0.859155 -10.788 et al, 0.648 -10.7881 taxonomy 0.648 -10.7886 chiang 0.933333 -10.7886 task-oriented 0.628788 -10.7888 100% 0.572327 -10.79 reward 0.869565 -10.7901 degrees of 0.981481 -10.7903 nevertheless 0.504854 -10.7906 ratings 0.715686 -10.7912 pilot 0.394659 -10.7912 researchers 0.424138 -10.7913 increased 0.981481 -10.7914 centrality 0.772727 -10.7914 synthetic 0.437956 -10.7915 place 0.538889 -10.7915 annotation, 0.675439 -10.7918 searching 0.47807 -10.7919 reverse 0.609929 -10.7923 80% 0.981481 -10.7925 cascaded 0.529412 -10.7926 modeled 0.981481 -10.7926 outperforming 0.727273 -10.793 characteristic 0.103406 -10.793 2 0.981481 -10.7932 hall 0.49763 -10.7935 we did not 0.77907 -10.7941 dominant 0.744681 -10.7943 wizard 0.981481 -10.7945 accounting for 0.62406 -10.7946 summarize 0.785714 -10.7949 axiom 0.981481 -10.7949 absent 0.981481 -10.795 converges 0.948276 -10.7952 ldc 0.628788 -10.7955 n. 0.880597 -10.7957 1.3 0.981481 -10.7959 cross-entropy 0.635659 -10.7959 computationally 0.948276 -10.7959 maintains 0.981481 -10.7962 clauses, 0.964286 -10.7965 distributional semantics 0.981481 -10.7965 formation 0.918033 -10.7967 condensed 0.981481 -10.7968 highlighting 0.493023 -10.7969 producing 0.981481 -10.797 negatives 0.981481 -10.7971 tc 0.857143 -10.7972 nlp tasks 0.981481 -10.7974 pdtb 0.981481 -10.7976 distinguished 0.892308 -10.7979 web-based 0.981481 -10.798 62 0.785714 -10.7982 averaging 0.918033 -10.7982 previous studies 0.269553 -10.7984 indicate 0.948276 -10.7985 evaluations, 0.770115 -10.7986 paragraphs 0.981481 -10.7987 61 0.731959 -10.7987 fluent 0.572327 -10.7989 extensive 0.981481 -10.799 happen 0.785714 -10.7991 receives 0.652893 -10.7994 inserted 0.675439 -10.7997 not, 0.904762 -10.8 asks 0.468354 -10.8001 perplexity 0.40836 -10.8002 mixture 0.669565 -10.8002 encoded in 0.857143 -10.8005 two or more 0.981481 -10.8006 two, 0.981481 -10.801 press 0.810127 -10.8012 minutes 0.253713 -10.8018 definition 0.688073 -10.8021 equivalence 0.51269 -10.8022 need for 0.818182 -10.8024 pt 0.50495 -10.8025 specify 0.515464 -10.8026 holds 0.792683 -10.8027 connections 0.678571 -10.8032 extends 0.502439 -10.8035 tutor 0.141696 -10.8037 i 0.367188 -10.8039 currently 0.904762 -10.8041 low frequency 0.470085 -10.8041 in general, 0.776471 -10.8051 tendency 0.8 -10.8052 extractions 0.376374 -10.8059 biomedical 0.963636 -10.806 contrary to 0.507463 -10.8061 nouns, 0.947368 -10.8063 window size 0.601399 -10.8063 or not. 0.446154 -10.8067 performance, 0.878788 -10.8068 attached to 0.712871 -10.807 3.5 0.755556 -10.807 et al 0.632812 -10.8076 prove 0.981132 -10.8077 chunker 0.219127 -10.8078 news 0.355392 -10.8079 results, 0.691589 -10.8081 parameters, 0.379888 -10.8082 domains 0.614815 -10.8094 absence of 0.741935 -10.8096 employs 0.981132 -10.8097 relied 0.652893 -10.8099 applicable to 0.154273 -10.81 like 0.471861 -10.8101 difficulty 0.632812 -10.8108 conceptual 0.56962 -10.8109 exhibit 0.947368 -10.811 ami 0.981132 -10.8112 fits 0.981132 -10.8114 posts, 0.981132 -10.8116 highlights 0.981132 -10.812 norms 0.325103 -10.8121 optimization 0.8 -10.8121 insights 0.947368 -10.8123 rapidly 0.963636 -10.8124 ni 0.918033 -10.8126 tm 0.867647 -10.8127 square 0.783133 -10.8128 esl 0.947368 -10.813 negatively 0.981132 -10.813 knowledge-based 0.614815 -10.8131 cue 0.981132 -10.8133 behaviors 0.396923 -10.8134 we report 0.826667 -10.8135 only a few 0.981132 -10.8137 prepared 0.932203 -10.814 sessions 0.981132 -10.814 tesla-f 0.815789 -10.8141 cohesion 0.890625 -10.8142 multi-class 0.947368 -10.8144 brevity 0.867647 -10.8144 ablation 0.890625 -10.8144 a* 0.981132 -10.8144 manage 0.826667 -10.8145 9. 0.695238 -10.8145 corrected 0.932203 -10.8146 69 0.807692 -10.8147 resolving 0.578947 -10.815 unlikely 0.675676 -10.8151 … 0.84507 -10.8154 isolated 0.826667 -10.8157 0.84 0.963636 -10.8157 note that, 0.587838 -10.8158 automatic evaluation 0.486111 -10.8159 competitive 0.947368 -10.816 non 0.855072 -10.816 segmenting 0.981132 -10.816 true, 0.747253 -10.8162 symmetric 0.947368 -10.8165 dates 0.835616 -10.8166 globally 0.963636 -10.8166 description. 0.54023 -10.817 broad 0.981132 -10.8171 proposals 0.981132 -10.8174 adverbial 0.65812 -10.8176 wang 0.507614 -10.8179 choices 0.931034 -10.818 next section. 0.65812 -10.8184 las 0.614815 -10.8185 by combining 0.981132 -10.8189 separating 0.981132 -10.8194 ignores 0.717172 -10.8195 model: 0.185435 -10.8197 through 0.790123 -10.8207 more detailed 0.614815 -10.8207 ne 0.642276 -10.8208 has shown 0.695238 -10.8215 completion 0.197097 -10.8216 unsupervised 0.747253 -10.8217 instruction 1 -10.8218 true. 0.980769 -10.8218 cost. 0.783133 -10.8219 ip 0.507614 -10.8221 impact on 0.980769 -10.8227 artificial intelligence 0.598592 -10.8228 near 0.797468 -10.8229 almost all 0.32839 -10.8229 response 0.20462 -10.823 question 0.695238 -10.8232 collected from 0.685185 -10.8233 requirements 0.903226 -10.8234 grammatically 0.315068 -10.8237 g 0.931034 -10.8238 argument, 0.962963 -10.8239 speculation 0.699029 -10.824 white 0.962963 -10.8246 idiomatic 0.903226 -10.8249 generalizes 0.322382 -10.8253 core 0.54386 -10.8254 learners 0.962963 -10.8254 convex 0.946429 -10.8257 pairing 0.699029 -10.8258 0.01 0.699029 -10.8258 mode 0.790123 -10.8259 unknown words 0.353086 -10.826 noisy 0.554878 -10.8267 onto 0.58 -10.8267 costs 0.865672 -10.8273 listener 0.65 -10.8274 obtains 0.758621 -10.8275 false positive 0.449799 -10.8275 page 0.962963 -10.8277 3the 0.31094 -10.8282 works 0.44664 -10.8288 relevance 0.903226 -10.8294 as, 0.623077 -10.8298 authority 0.721649 -10.8302 simulation 0.980769 -10.8304 communicate 0.980769 -10.8304 isolation 0.980769 -10.8304 blind 0.876923 -10.8304 naming 0.205882 -10.8306 able to 0.980769 -10.8306 better, 0.980769 -10.8306 properties, 0.278034 -10.8308 limited 0.35589 -10.8308 clearly 0.758621 -10.831 phoneme 0.66087 -10.8312 tense 0.916667 -10.8313 clue 0.946429 -10.8317 fe 0.946429 -10.8317 coreference resolution. 0.876923 -10.8319 met 0.931034 -10.8321 edited 0.980769 -10.8322 grasp 0.980769 -10.8323 optimizer 0.347722 -10.8325 database 0.876923 -10.8326 augmented with 0.980769 -10.8329 capacity 0.403909 -10.833 index 0.337104 -10.8332 to produce 0.780488 -10.8333 removed from 0.726316 -10.8334 2009. 0.744444 -10.8335 table 3, 0.797468 -10.8341 api 0.805195 -10.8345 arabic, 0.980769 -10.8345 negated 0.244922 -10.8349 just 0.805195 -10.8351 controls 0.607407 -10.8352 offers 0.946429 -10.8352 ranked list 0.805195 -10.8352 nil 0.797468 -10.8356 adjoining 0.797468 -10.8358 imt 0.205387 -10.8358 bilingual 0.534091 -10.8363 sizes 0.833333 -10.8364 knowledge sources 0.915254 -10.8375 84 0.591549 -10.8375 resolve 0.980769 -10.8377 objectives 0.980769 -10.8378 88 0.980769 -10.8379 multi-document summarization 0.23388 -10.8382 users 0.980769 -10.8385 german-english 0.431227 -10.8388 never 0.980769 -10.839 went 0.497512 -10.8397 nps 0.75 -10.8398 computational linguistics volume 0.672727 -10.8399 critical 0.626984 -10.8399 10-fold 0.515957 -10.8401 movie 0.794872 -10.8403 gibbs sampling 0.945455 -10.8403 driven by 0.962264 -10.8409 talking 0.980769 -10.8409 heart 1 -10.8416 variants. 1 -10.8416 usage. 0.280315 -10.8417 finding 0.272727 -10.8419 partial 0.901639 -10.8421 (step 0.297683 -10.8424 product 0.696078 -10.8426 semantic representation 0.682243 -10.8428 treebanks 0.586207 -10.8429 all, 0.736264 -10.8431 tracking 0.929825 -10.8433 non-local 0.842857 -10.8434 different levels 0.852941 -10.8438 headline 0.915254 -10.8439 unstructured 0.564103 -10.8439 to calculate 0.71875 -10.8439 voting 0.901639 -10.8441 (van 0.901639 -10.8443 augmenting 0.842857 -10.8444 authors, 0.465217 -10.8449 europarl 0.406667 -10.845 in contrast, 0.888889 -10.8451 generative models 0.490291 -10.8452 italian 0.7875 -10.8452 performances 0.888889 -10.8453 0.70 0.821918 -10.8458 concern 0.915254 -10.8459 confusion network 0.25876 -10.8459 manual 0.401961 -10.8466 fewer 0.945455 -10.8468 surface form 0.980392 -10.847 incorporation 0.980392 -10.8471 two-stage 0.945455 -10.8478 nuggets 0.398714 -10.8479 none 0.31893 -10.8479 adaptation 0.962264 -10.8485 q, 0.794872 -10.8486 degradation 0.275651 -10.8489 performs 0.521978 -10.8492 three different 0.545455 -10.8494 kept 0.777778 -10.8497 ρ 0.592857 -10.8503 except for 0.980392 -10.8504 assessments 0.755814 -10.8507 characters. 0.652174 -10.8508 4) 0.538012 -10.8509 in this work 0.163597 -10.851 pos 0.755814 -10.8515 precisely 0.636364 -10.8516 essential 0.945455 -10.8516 sentence compression 0.679245 -10.8518 got 0.777778 -10.8518 in future work, 0.154143 -10.8521 research 0.929825 -10.8528 construction, 0.327511 -10.8531 world 0.28619 -10.8531 found in 0.875 -10.8531 vowel 0.644068 -10.8532 are extracted from 0.174154 -10.8532 further 0.629032 -10.8532 a) 0.497487 -10.8534 decrease 0.962264 -10.8536 dual decomposition 0.980392 -10.8541 interpretation, 0.945455 -10.8542 in principle, 0.980392 -10.8542 surprisingly 0.980392 -10.8542 4), 0.980392 -10.8542 experimenting 0.497487 -10.8545 primarily 0.464912 -10.8545 inside 0.296429 -10.8545 computing 0.552795 -10.8546 judge 0.9 -10.855 curran 0.289608 -10.855 separate 0.423358 -10.855 earlier 0.321429 -10.8556 # 0.431818 -10.8557 synsets 0.850746 -10.8557 studies, 0.318841 -10.8561 special 0.980392 -10.8564 dice 0.483254 -10.8564 occurrences of 0.980392 -10.8565 spam 0.830986 -10.8567 presenting 0.945455 -10.8569 hundred 0.810811 -10.8569 37, number 0.313883 -10.8571 derivation 0.962264 -10.8579 table 8: 0.403333 -10.8581 integrated 0.326797 -10.8587 around 0.682692 -10.859 significant improvements 0.980392 -10.8592 urls 0.693069 -10.8593 asking 0.980392 -10.8594 sk 0.980392 -10.8595 calculates 0.913793 -10.8599 sim 0.237875 -10.86 last 0.887097 -10.8601 special case 0.693069 -10.8602 plausible 0.280645 -10.8606 called 0.121422 -10.8608 english 0.768293 -10.8608 worked 0.682692 -10.8609 nivre 0.9 -10.8611 bar 0.624 -10.8612 lengths 0.443548 -10.8617 her 1 -10.8618 french. 0.850746 -10.8621 compare against 0.532164 -10.8627 coherent 0.98 -10.8628 (based on 0.594203 -10.8631 c1 0.9 -10.8632 pantel 0.17095 -10.8634 there are 0.663636 -10.8634 statistical significance 0.339667 -10.8638 to compute 0.768293 -10.8645 region 0.861538 -10.8645 ui 0.432432 -10.8648 to get 0.78481 -10.865 (1), 0.693069 -10.8652 enhanced 0.913793 -10.8655 greek 0.9 -10.8655 country 0.98 -10.8655 implementation. 0.654867 -10.8657 partly 0.693069 -10.8657 knowledge about 0.792208 -10.866 morphological analysis 0.446721 -10.8667 letter 0.646552 -10.8668 across different 0.619048 -10.8669 n, 0.861538 -10.8669 reordering, 0.313131 -10.8672 5. 0.210191 -10.8674 local 0.483092 -10.8674 lowest 0.340476 -10.8676 follow 0.682692 -10.8676 similarity measures 0.98 -10.8678 grounded 0.686275 -10.8679 dependency structure 0.309665 -10.8681 predicates 0.733333 -10.8689 played 0.392971 -10.869 answers 0.928571 -10.8693 california 0.344743 -10.8695 hidden 0.928571 -10.8698 crisis 0.98 -10.8701 features; 0.98 -10.8702 sensitivity 0.98 -10.8704 co-occurring 0.185028 -10.8707 query 0.641026 -10.8712 determiner 0.72043 -10.8712 language pair 0.55414 -10.8713 to understand 0.69697 -10.8715 favor 0.98 -10.8716 strategies, 0.98 -10.8717 posteriors 0.98 -10.8719 attachments 0.272171 -10.872 strategy 0.898305 -10.8731 discontinuous 0.614173 -10.8733 closed 0.98 -10.8733 phonemes 0.408304 -10.8734 tweet 0.67619 -10.8735 (10) 0.961538 -10.8737 ptb 0.402685 -10.8741 compression 0.98 -10.8743 scalable 0.98 -10.8743 excellent 0.98 -10.8743 interest, 0.738636 -10.8743 fails to 0.961538 -10.8744 generative process 0.828571 -10.8745 contextual features 0.98 -10.8746 technique, 0.782051 -10.8748 table 4, 0.98 -10.8749 retrieving 0.164607 -10.8751 n 0.485294 -10.8761 li 0.873016 -10.8765 very few 0.562914 -10.8766 partition 0.98 -10.8767 1.6 0.98 -10.8769 smooth 0.712766 -10.877 2010, 0.838235 -10.8773 (collins, 0.151645 -10.8773 relations 0.554839 -10.8774 considerably 0.738636 -10.8775 star 0.508108 -10.8776 generated from 0.482927 -10.8777 splitting 0.782051 -10.8783 transliterations 0.759036 -10.8785 an utterance 0.98 -10.8787 classes: 0.176738 -10.8788 quality 0.912281 -10.8789 synonymous 0.898305 -10.8791 co-occur with 0.686275 -10.8793 evaluation metric 0.98 -10.8794 described above. 0.98 -10.8798 still, 0.643478 -10.8799 problems, 0.550633 -10.8804 2010. 0.321121 -10.8804 person 0.207921 -10.8807 agreement 0.5 -10.8809 genia 0.898305 -10.8811 families 0.669811 -10.8812 table 5. 0.572414 -10.8818 si 0.927273 -10.8818 presented here 0.725275 -10.8823 years. 0.505376 -10.8824 (7) 1 -10.8824 known. 1 -10.8824 number. 0.474178 -10.8825 integration 0.730337 -10.8825 recorded 0.258741 -10.8832 candidates 0.979592 -10.8833 conversations. 0.456522 -10.8834 outside 0.859375 -10.8834 module, 0.979592 -10.8837 named entities, 0.201531 -10.8838 final 0.554839 -10.8838 topic models 0.960784 -10.8839 results: 0.37574 -10.8841 formal 0.622951 -10.8846 generation, 0.838235 -10.8854 distinction between 0.635593 -10.8855 aggregate 0.712766 -10.8856 can be computed 0.838235 -10.8858 poses 0.838235 -10.8859 we intend to 0.420074 -10.8861 book 0.306931 -10.8863 span 0.572414 -10.8863 readers 0.567568 -10.8865 iteratively 0.912281 -10.8865 nl 0.744186 -10.8868 mentions, 0.960784 -10.8871 do so, 0.31524 -10.8875 defined as 0.960784 -10.8879 minimizes 0.563758 -10.8881 selectional 0.943396 -10.8882 pointing 0.58156 -10.8885 risk 0.730337 -10.8886 wil 0.151007 -10.8888 classification 0.979592 -10.8889 supertagging 0.183099 -10.8889 r 0.826087 -10.889 85 0.194444 -10.8898 selection 0.816901 -10.8899 figure 4. 0.943396 -10.8902 evenly 0.927273 -10.8904 arbitrarily 0.765432 -10.8905 types: 0.835821 -10.8906 synthesis 0.54717 -10.8907 prepositions 0.943396 -10.8911 people, 0.912281 -10.8912 seem to be 0.979592 -10.8912 quantification 0.654545 -10.8916 constitute 0.494845 -10.8922 to handle 0.859375 -10.8926 attributes. 0.979592 -10.8927 regularities 0.896552 -10.8928 pos tags. 0.756098 -10.893 m., 0.979592 -10.8931 aggregated 0.979592 -10.8932 tool, 0.870968 -10.8933 mismatch 0.979592 -10.8936 prominence 0.979592 -10.8938 grow 0.489796 -10.8939 adopted 0.29326 -10.8939 a particular 0.979592 -10.8941 high-dimensional 0.835821 -10.8944 buy 0.979592 -10.8946 bear 0.663551 -10.8947 powerful 0.979592 -10.895 scalability 0.979592 -10.8951 since, 0.979592 -10.8951 successive 0.979592 -10.8951 classifications 0.640351 -10.8954 presentation 0.979592 -10.8958 platform 0.979592 -10.8958 amongst 0.454148 -10.8959 phrases, 0.405594 -10.8963 measuring 0.979592 -10.8965 favors 0.797297 -10.8966 india 0.741176 -10.897 turn-taking 0.457778 -10.8971 hand 0.625 -10.8971 cm 0.582734 -10.8971 treatment 0.979592 -10.8972 yago 0.583942 -10.8973 predicate-argument 0.979592 -10.8973 text categorization 0.741176 -10.8974 genres 0.883333 -10.8979 que 0.826087 -10.8987 0.55 0.835821 -10.8989 notable 0.203524 -10.9004 reordering 0.870968 -10.9005 2.0 0.548387 -10.9007 a series 0.979592 -10.9007 irony 0.883333 -10.9009 units, 0.979592 -10.9009 109 0.846154 -10.901 character-based 0.301741 -10.901 subset of 0.109526 -10.9012 3 0.942308 -10.9013 interact with 0.979592 -10.902 search engines 0.657407 -10.9021 expresses 0.727273 -10.9024 epsilon1 0.756098 -10.9024 counts. 0.544304 -10.9025 manner 0.857143 -10.9031 a: 0.96 -10.9032 checks 0.427451 -10.9033 synchronous 0.942308 -10.9033 m1 0.814286 -10.9033 svm classifier 0.455752 -10.9038 collocation 0.432 -10.9041 similarly 0.564626 -10.9046 we choose 0.96 -10.9053 finite state 0.96 -10.9053 lexical substitution 0.233179 -10.9055 before 0.619835 -10.9055 @ 0.219814 -10.9055 evaluated 0.896552 -10.9061 deviant 0.37963 -10.9062 express 0.632479 -10.9062 pronunciation 0.372024 -10.9067 asr 0.68 -10.9068 summarization, 0.686869 -10.9068 7 conclusion 0.632479 -10.907 maximal 0.640351 -10.907 changed 0.226623 -10.9072 against 0.120459 -10.9077 4 0.769231 -10.9078 clues 0.477833 -10.9078 characteristics of 0.68 -10.908 overall performance 0.465116 -10.908 wi 0.925926 -10.9081 na 0.284722 -10.9083 tagging 0.786667 -10.9084 very small 0.395973 -10.9084 influence 0.857143 -10.9085 at least two 0.410909 -10.9086 are often 0.436735 -10.9086 involve 0.3769 -10.909 challenge 0.242424 -10.9091 hierarchical 0.896552 -10.9092 bb 0.910714 -10.9092 severe 0.601562 -10.9093 phrase structure 0.399317 -10.9094 involves 0.96 -10.9098 effect, 0.979167 -10.9099 data) 0.96 -10.91 rouge-1 0.881356 -10.9103 v1 0.732558 -10.9103 cognate 0.942308 -10.9103 task success 0.533742 -10.9105 demonstrate that 0.302554 -10.9106 interaction 0.7625 -10.9107 cohen 0.794521 -10.9107 agents 0.634783 -10.9109 measured by 0.425781 -10.9112 max 0.418868 -10.9114 runs 0.319383 -10.9115 systems, 0.979167 -10.9118 truly 0.910714 -10.9118 signs 0.436214 -10.9119 consistently 0.979167 -10.912 matching, 0.833333 -10.9122 tailored 0.690722 -10.9123 unrelated 0.979167 -10.9127 uncertain 0.979167 -10.9129 passes 0.979167 -10.913 1996) 0.706522 -10.9131 wall street 0.979167 -10.9132 sun 0.846154 -10.9133 lexically 0.690722 -10.9134 manner. 0.979167 -10.9135 experimentally 0.642857 -10.9135 tree-to-string 0.634783 -10.9137 st 0.868852 -10.9137 dropped 0.719101 -10.914 profile 0.776316 -10.914 re-ranking 0.786667 -10.9143 spearman’s 0.942308 -10.9144 ans 0.979167 -10.9145 activities 0.942308 -10.9146 dot 0.614754 -10.9146 was done 0.428571 -10.9147 used, 0.746988 -10.9147 tell 0.823529 -10.9148 bp 0.769231 -10.9148 out. 0.651376 -10.9149 techniques, 0.411765 -10.9153 to reduce 0.979167 -10.9155 top-ranked 0.823529 -10.9155 rapid 0.942308 -10.9158 subtype 0.17743 -10.9159 include 0.802817 -10.916 users’ 0.471154 -10.9161 reducing 0.769231 -10.9162 advantages of 0.979167 -10.9164 displaying 0.979167 -10.9167 travel 0.979167 -10.9167 1.8 0.84375 -10.9167 pos tags, 0.857143 -10.9169 by exploiting 0.979167 -10.917 locations, 0.979167 -10.917 anaphora resolution 0.979167 -10.9173 improvement, 0.738095 -10.9175 correcting 0.449782 -10.9177 criteria 0.881356 -10.9178 seeing 0.979167 -10.9179 offered 0.979167 -10.9181 windows 0.802817 -10.9182 encourage 0.494681 -10.9182 compositionality 0.128674 -10.9184 accuracy 0.769231 -10.9196 prior knowledge 0.979167 -10.9197 fire 0.96 -10.92 social behavior 0.942308 -10.9201 counts, 0.545455 -10.9201 to develop 0.477612 -10.9204 considers 0.52071 -10.9204 pass 0.769231 -10.9204 usability 0.297505 -10.9207 seed 0.457014 -10.9208 pcfg 0.802817 -10.9209 “a 0.894737 -10.921 subsampling 0.833333 -10.9212 parallel corpora. 0.924528 -10.9213 degree to which 0.979167 -10.9214 g, 0.232667 -10.922 produce 0.979167 -10.9221 feel 0.979167 -10.9226 signatures 0.909091 -10.9226 oriented 0.979167 -10.9229 sequence labeling 0.96 -10.9229 set: 0.783784 -10.9229 overlap between 0.811594 -10.923 various types 0.698925 -10.9232 desirable 0.979167 -10.9238 (only 0.924528 -10.9239 joint inference 0.168415 -10.924 recall 0.979167 -10.9246 (more 0.534161 -10.9247 topics, 0.959184 -10.9248 statistics, 0.941176 -10.9248 raising 0.959184 -10.9249 97 0.753086 -10.9249 previous research 0.666667 -10.9251 highlighted 0.959184 -10.9252 anything 0.978723 -10.9253 united states 0.6875 -10.9254 adjectives, 0.959184 -10.9258 101 0.616667 -10.9258 variability 0.694737 -10.9261 initialized 0.959184 -10.9265 previous work. 0.464455 -10.9271 correlations 0.959184 -10.9273 (2003), 0.6875 -10.9275 our approach, 0.978723 -10.9276 approximations 0.350404 -10.9277 computational linguistics, 0.941176 -10.9277 li et al. 0.820896 -10.9281 subtle 0.881356 -10.9282 polarities 0.924528 -10.9285 enter 0.811594 -10.9286 72 0.408759 -10.9287 follows 0.978723 -10.9288 taken into account 0.6875 -10.9289 scales 0.894737 -10.929 distinctive 0.959184 -10.9291 thomas 0.783784 -10.9304 we were able 0.179325 -10.9305 common 0.811594 -10.9308 gap between 0.470874 -10.931 nonterminal 0.87931 -10.931 carefully 0.978723 -10.9312 figure 7: 0.698925 -10.9314 renaming 0.5625 -10.9318 simultaneously 0.866667 -10.9319 robert 0.554054 -10.9322 gets 0.978723 -10.9323 variation, 0.978723 -10.9325 bitexts 0.240202 -10.9326 list of 0.373089 -10.9326 lattice 0.978723 -10.9329 0.58 0.978723 -10.9329 serious 0.567376 -10.9336 benefits 0.924528 -10.9337 confused 0.978723 -10.9337 6) 0.783784 -10.9337 editing 0.242542 -10.9346 mention 0.67 -10.9348 formalism 0.676768 -10.9348 twice 0.58209 -10.9352 qualitative 0.213925 -10.9353 texts 0.811594 -10.9354 man 0.715909 -10.9354 associate 0.978723 -10.9355 bins 0.941176 -10.9355 6 discussion 0.892857 -10.9356 bilingual dictionary 0.460094 -10.9357 belief 0.6875 -10.9359 learnt 0.959184 -10.9363 april 0.978723 -10.9363 ways, 0.59375 -10.9364 ask 0.550336 -10.9366 gradient 0.650943 -10.9366 which, 0.189274 -10.9367 total 0.854839 -10.9371 more detail 0.31441 -10.9372 performing 0.75 -10.9376 dm 0.715909 -10.9378 quantifier 0.978723 -10.9378 minority 0.791667 -10.9378 bring 0.941176 -10.9378 medline 0.404332 -10.938 visual 0.508571 -10.9381 essentially 0.84127 -10.9382 decide whether 0.978723 -10.9384 directly, 0.75 -10.9386 to prevent 0.606557 -10.9387 utilized 0.978723 -10.9387 writers 0.390572 -10.9387 (2010), 0.583333 -10.9392 we get 0.613445 -10.9396 covers 0.691489 -10.9397 b3 0.201432 -10.9398 represent 0.486911 -10.9402 texts, 0.978723 -10.9404 91 0.907407 -10.9406 prone to 0.978723 -10.9406 predominantly 0.978723 -10.9407 pages, 0.341085 -10.9408 author 0.978723 -10.9409 rater 0.773333 -10.9409 session 0.907407 -10.9414 rate, 0.75 -10.9414 passed 0.978723 -10.9414 entries, 0.941176 -10.9418 companies 0.959184 -10.9419 uniquely 0.978723 -10.9421 preserves 0.87931 -10.9421 chinese, 0.941176 -10.9421 news, 0.514793 -10.9422 keyword 0.978723 -10.9423 suffers from 0.978723 -10.9423 compatibility 0.680412 -10.9424 ignoring 0.170763 -10.943 correct 0.521212 -10.9431 rows 0.907407 -10.9432 exercises 0.866667 -10.9434 structured prediction 0.978723 -10.9434 attested 0.978723 -10.9436 in recent years, 0.464115 -10.9438 transitions 0.535032 -10.9441 terms, 0.653846 -10.9449 0.05 0.75 -10.945 operate 0.436441 -10.945 attachment 0.923077 -10.9451 prior polarity 0.660194 -10.9452 ideal 0.892857 -10.9459 top-down 0.538961 -10.9463 continuous 0.892857 -10.9463 auc 1 -10.947 opinions. 1 -10.947 analyses. 1 -10.947 effective. 0.680412 -10.9473 british 0.818182 -10.9477 we wish to 0.978261 -10.9479 wordnet (fellbaum, 0.907407 -10.9483 timebank 0.958333 -10.9486 notions 0.852459 -10.9492 (huang 0.195726 -10.9494 entities 0.907407 -10.9495 linguists 0.923077 -10.9497 ts 0.479381 -10.9499 pi 0.87931 -10.9503 navigation 0.892857 -10.9504 emoticons 0.64486 -10.9505 ? 0.923077 -10.9507 predicate argument 0.94 -10.9509 postprocessing 0.335859 -10.951 map 0.423387 -10.9512 informative 0.94 -10.9512 proper names 0.40824 -10.9512 limit 0.653846 -10.9513 can help 0.852459 -10.9518 singleton 0.712644 -10.9519 are, 0.564286 -10.952 s1 0.684211 -10.9521 web search 0.978261 -10.9524 state-of-theart 0.494505 -10.9525 introduces 0.94 -10.9531 mckeown 0.684211 -10.9532 datasets, 0.673469 -10.9534 president 0.978261 -10.9535 turns out 0.780822 -10.9536 concatenated 0.524691 -10.9541 evaluated on 0.978261 -10.9543 permit 0.7 -10.9547 composing 0.746835 -10.9547 cancer 0.244624 -10.9553 nouns 0.958333 -10.9559 more importantly, 0.978261 -10.9561 separation 0.712644 -10.9562 splits 0.94 -10.9562 basic idea 0.877193 -10.9563 sentiment, 0.94 -10.9564 psycholinguistic 0.808824 -10.9566 when compared to 0.636364 -10.9567 (1 0.978261 -10.9568 simplifications 0.603306 -10.9568 morphologically 0.828125 -10.957 development set, 0.852459 -10.9571 68 0.797101 -10.9572 s.t. 0.978261 -10.9575 ordinal 0.808824 -10.9576 male 0.465686 -10.9577 kernels 0.780822 -10.958 intention 0.746835 -10.9582 expression, 0.46798 -10.9582 paired 0.94 -10.9584 legal 0.746835 -10.9587 manual evaluation 0.617391 -10.9597 reached 0.213622 -10.9597 lexicon 0.978261 -10.9598 yu 0.978261 -10.9598 threads 0.852459 -10.96 sentiment analysis, 0.818182 -10.9601 semantic representations 0.978261 -10.9609 respond 0.753247 -10.961 turned 0.978261 -10.9611 real-valued 0.788732 -10.9618 demonstrating 0.978261 -10.962 materials 0.575758 -10.9621 parallel data 0.717647 -10.9623 treelets 0.890909 -10.9623 mihalcea 0.94 -10.9627 words), 0.717647 -10.9628 into account. 0.978261 -10.9629 pre-defined 0.224629 -10.963 latent 0.978261 -10.9631 spelling correction 0.630631 -10.9634 wn 0.978261 -10.9635 century 0.877193 -10.9635 forests 0.432203 -10.9636 promising 0.753247 -10.9638 merely 0.90566 -10.9638 inferred from 0.518293 -10.9639 extra 0.630631 -10.9639 unannotated 0.476684 -10.9641 successful 0.978261 -10.9642 orderings 0.556338 -10.9643 keyphrase 0.677083 -10.9644 cipher 0.437229 -10.9644 attention 0.677083 -10.9647 uas 0.231235 -10.9648 social 0.828125 -10.9649 underspecified 0.978261 -10.965 parts, 0.90566 -10.965 forming 0.828125 -10.965 included. 0.45933 -10.9652 an unsupervised 0.363363 -10.9652 unit 0.978261 -10.9653 unusual 0.921569 -10.9655 queried 0.777778 -10.9656 dependency relations 0.746835 -10.9657 sentiments 0.511905 -10.9659 contained in 0.641509 -10.9663 word senses 0.306383 -10.9664 logical 0.486486 -10.9664 2011) 0.666667 -10.9665 can easily 0.7 -10.9665 implements 0.731707 -10.9668 to recover 0.731707 -10.9671 meaning, 0.535948 -10.9671 20% 0.958333 -10.9672 sets: 0.547945 -10.9672 tac 0.828125 -10.9672 substitutions 0.978261 -10.9674 smt, 0.515152 -10.9675 w. 0.650485 -10.9675 agreed 0.020696 -10.9676 and 0.978261 -10.9676 ties 0.921569 -10.9677 house 0.330025 -10.9677 provided by 0.80597 -10.9678 82 0.603306 -10.9679 test sets. 0.449541 -10.968 read 0.228844 -10.9681 20 0.921569 -10.9683 martins 0.717647 -10.9684 ml 0.566176 -10.9685 the proposed method 0.306867 -10.9688 such that 0.85 -10.9691 annotation guidelines 0.362275 -10.9693 strings 1 -10.9694 next. 0.977778 -10.9695 (1). 0.890909 -10.9697 ic 0.777778 -10.9698 users, 0.406015 -10.97 significance 0.90566 -10.9701 automatically, 0.566176 -10.9705 strictly 0.463054 -10.9706 align 0.650485 -10.9706 robustness 0.731707 -10.9713 points, 0.617391 -10.9714 chen 0.977778 -10.9714 held out 0.717647 -10.9714 location, 0.921569 -10.9714 clairlib 0.90566 -10.9717 random variables 0.797101 -10.9719 influences 0.7375 -10.9722 (11) 0.402985 -10.9724 power 0.518519 -10.9724 and hence 0.400735 -10.9731 connectives 0.977778 -10.9731 regular expressions 0.957447 -10.9732 systems: 0.85 -10.9738 %) 0.505882 -10.9739 conjunction 0.317865 -10.9739 major 0.293886 -10.9739 ranked 0.627273 -10.974 der 0.90566 -10.9741 aspects, 0.921569 -10.9742 subgradient 0.921569 -10.9745 an indication 0.23375 -10.9748 monolingual 0.692308 -10.9751 pruned 0.90566 -10.9752 net 0.571429 -10.9752 period 0.90566 -10.9753 attack 0.593496 -10.9755 exponential 0.815385 -10.9757 technology, 0.890909 -10.9757 edges, 0.785714 -10.9759 more robust 0.938776 -10.9763 visualization 0.350282 -10.9763 contrast 0.921569 -10.9766 storing 0.728395 -10.9766 gained 0.85 -10.9766 clicks 0.938776 -10.9767 cutoff 0.85 -10.9768 realizations 0.977778 -10.9779 continued 0.875 -10.978 statistically significant. 0.977778 -10.9782 approximates 0.957447 -10.9782 fall into 0.673684 -10.9782 fail to 0.83871 -10.9783 bridge 0.825397 -10.9783 reference translations 0.862069 -10.9783 och 0.371429 -10.9785 f-measure 0.80597 -10.9786 tion 0.363914 -10.9787 definition of 0.505882 -10.9788 answering 0.80597 -10.9789 modal 0.680851 -10.9789 live 0.680851 -10.979 inconsistent 0.977778 -10.9791 nlg 0.670103 -10.9792 things 0.606838 -10.9794 worst 0.888889 -10.9799 figure 1: example 0.90566 -10.9799 music 0.938776 -10.9802 automata 0.619469 -10.9803 (brown et al., 0.977778 -10.9807 retaining 0.977778 -10.9808 motivates 0.214135 -10.9808 features, 0.977778 -10.9815 pictures 0.875 -10.9815 glosses 0.836066 -10.9816 millions of 0.977778 -10.9818 optimum 0.977778 -10.9819 geographic 0.488889 -10.9822 discover 0.977778 -10.9824 hierarchy, 0.977778 -10.9824 so far. 0.977778 -10.9825 precision, recall, 0.75 -10.9827 n-grams. 0.938776 -10.9829 intent 0.606838 -10.983 to optimize 0.977778 -10.9833 projections 0.47644 -10.9834 categories, 0.714286 -10.9835 systematically 0.977778 -10.9835 reason why 0.977778 -10.9836 enabled 0.938776 -10.9836 crosslingual 0.794118 -10.9836 ordinary 0.38 -10.9837 200 0.977778 -10.9839 overly 0.977778 -10.9839 frame, 0.417004 -10.9842 years 0.977778 -10.9843 similarity-based 0.903846 -10.9844 top-level 0.448598 -10.9845 coherence 0.80303 -10.9847 planning 0.977778 -10.9848 applied, 0.80303 -10.985 joshua 0.977778 -10.9851 pattern-based 0.774648 -10.9852 outline 0.815385 -10.9857 lexical resource 0.656566 -10.9858 sketch 0.977778 -10.9859 decoders 0.903846 -10.9859 (e) 0.903846 -10.986   0. 0.977778 -10.9861 briefly describe 0.6 -10.9868 hours 0.938776 -10.987 there exist 0.656566 -10.9873 streams 0.225768 -10.9874 dictionary 0.977778 -10.9876 information retrieval, 0.977778 -10.9882 theoretically 0.767123 -10.9886 units. 0.847458 -10.9891 conservative 0.608696 -10.9894 pool 0.977778 -10.9897 behavior, 0.977778 -10.9903 imagine 0.525974 -10.9904 the majority of 0.250733 -10.9904 concepts 0.957447 -10.9905 2005. 0.888889 -10.9906 coreferent 0.875 -10.9906 punctuation. 0.92 -10.991 society 0.446512 -10.9912 phase 0.977778 -10.9913 5.5 0.977778 -10.9913 mentioning 0.688889 -10.9917 knowing 0.888889 -10.992 transliteration pairs 0.956522 -10.9921 money 0.956522 -10.9922 2008, 1 -10.9924 first. 1 -10.9924 experts. 0.836066 -10.9925 membership 0.677419 -10.9929 uk 0.693182 -10.9931 to minimize 0.464646 -10.9931 ccg 0.977273 -10.9933 chinese. 0.405405 -10.9934 history 0.859649 -10.9934 gappy 0.977273 -10.9937 randomly sampled 0.608696 -10.9937 principle 0.411067 -10.9939 derivations 0.693182 -10.9941 1), 0.578125 -10.9941 configurations 0.80303 -10.9942 ceaf 0.888889 -10.9944 domain knowledge 0.903846 -10.9947 conjoined 0.977273 -10.9951 non-trivial 0.9375 -10.9953 w0 0.9375 -10.9954 pertaining 0.541096 -10.9956 90 0.705882 -10.9956 (1998) 0.9375 -10.9957 motivating 0.13004 -10.9962 relation 0.269036 -10.9962 sequences 0.977273 -10.9964 biases 0.486188 -10.9965 0.1 0.416327 -10.9967 mark 0.21802 -10.9967 f 0.75 -10.9968 49 0.9375 -10.9968 q. 0.8125 -10.9968 device 0.977273 -10.9971 ensuring 0.118399 -10.9972 we use 0.368254 -10.9974 for example 0.310658 -10.9978 spanish 0.859649 -10.998 begin with 0.610619 -10.9985 we know 0.28463 -10.9986 presented in 0.8125 -10.9986 an explicit 0.8125 -10.9986 stylistic 0.956522 -10.9989 furthermore 0.977273 -10.9989 reg 0.656566 -10.9989 out-of-domain 0.80303 -10.999 87 0.956522 -10.9993 meta 0.719512 -10.9993 1998), 0.497076 -10.9996 on the basis 0.847458 -10.9998 distributions, 0.977273 -11 passages 0.659794 -11 distant 0.977273 -11.0007 task-specific 0.847458 -11.0007 n-grams, 0.977273 -11.001 independently, 0.155017 -11.001 verb 0.956522 -11.0011 speakers, 0.430435 -11.0013 (6) 0.65 -11.0013 discard 0.725 -11.0015 0.83 0.515723 -11.0015 solving 0.977273 -11.0022 during training, 0.569231 -11.0025 decoding, 0.40625 -11.0026 investigated 0.977273 -11.0027 situated 0.977273 -11.0027 extracted, 0.473404 -11.0027 meteor 0.92 -11.003 ve 0.977273 -11.0033 keyphrase extraction 0.659794 -11.0033 stages 0.284895 -11.0035 states 0.903846 -11.0036 hi 0.558824 -11.0036 translations, 0.242382 -11.004 give 0.301724 -11.0041 root 0.977273 -11.0041 cycle 0.9375 -11.0042 everything 0.618182 -11.0046 ii 0.244039 -11.0047 procedure 0.886792 -11.0048 monosemous 0.977273 -11.005 dissimilar 0.977273 -11.0051 precision/recall 0.977273 -11.0052 stopwords 0.92 -11.0056 lemmatization 0.585366 -11.0058 morphemes 0.977273 -11.006 resort to 0.977273 -11.006 description, 0.329049 -11.0061 pairs, 0.433036 -11.0064 learning, 0.977273 -11.0069 noisy, 0.550725 -11.0069 notation 0.65 -11.0071 among all 0.486034 -11.0072 50% 0.554745 -11.0072 heads 0.977273 -11.0075 his/her 0.8 -11.0077 by introducing 0.20374 -11.0078 existing 0.977273 -11.0079 proposing 0.164427 -11.008 additional 0.8125 -11.0081 les 0.872727 -11.0082 questionnaire 0.977273 -11.0086 cmu 0.763889 -11.0086 avg 0.833333 -11.0087 would require 0.681319 -11.0095 independence 0.400763 -11.0097 graphical 0.977273 -11.0103 change, 0.519231 -11.0109 complementary 0.977273 -11.011 greatest 0.763889 -11.011 zhou 0.74026 -11.0111 inferences 0.449275 -11.0112 stream 0.716049 -11.0114 syntactic structure 0.383803 -11.0114 determining 0.564885 -11.0115 flexible 0.569231 -11.012 ≥ 0.977273 -11.0123 error rates 0.872727 -11.0123 lab 0.337875 -11.0123 actions 0.685393 -11.0124 40% 0.844828 -11.0124 mathematical 0.901961 -11.0127 yang, 0.977273 -11.0129 illustrating 0.612613 -11.0129 disambiguate 0.537931 -11.0129 for instance 0.488636 -11.013 kappa 0.977273 -11.0132 today 0.753425 -11.0133 cumulative 0.542254 -11.0135 ensure that 0.247093 -11.0137 improved 0.956522 -11.0139 constructions. 0.274686 -11.0139 building 0.135312 -11.0143 type 0.753425 -11.0145 seconds 0.763889 -11.0145 2005 0.977273 -11.0148 outputs, 0.809524 -11.0148 explores 0.977273 -11.015 alone, 0.763889 -11.0154 to simplify 0.955556 -11.0156 interpreting 0.434389 -11.0157 operation 1 -11.016 do. 1 -11.016 architecture. 1 -11.016 difference. 0.955556 -11.0163 2the 0.791045 -11.0166 adequate 0.697674 -11.0166 number, 0.872727 -11.0167 mvm 0.108078 -11.0168 learning 0.449275 -11.017 user’s 0.955556 -11.0171 lvcs 0.308924 -11.0171 underlying 0.710843 -11.0171 constitutes 0.457286 -11.0171 drop 0.519481 -11.0174 environment 0.955556 -11.0176 99 0.746667 -11.0176 lin, 0.59322 -11.018 compact 0.443396 -11.018 regularization 0.65625 -11.018 localization 0.771429 -11.0181 each iteration 0.210359 -11.0182 wikipedia 0.465969 -11.0182 citation 0.360248 -11.0183 eq. 0.901961 -11.0185 determination 0.716049 -11.0187 getting 0.771429 -11.0187 has become 0.612613 -11.0189 to discover 0.331579 -11.019 decoder 0.736842 -11.019 optional 0.771429 -11.0192 affinity 0.403922 -11.0196 foreign 0.93617 -11.0197 parameter estimation 0.315789 -11.0197 compound 0.560606 -11.0198 1998) 0.523179 -11.0199 in fact 0.833333 -11.0202 in our work, 0.779412 -11.0203 feasible 0.976744 -11.0205 based upon 0.258842 -11.0209 smt 0.753425 -11.021 data: 0.8 -11.021 ungrammatical 0.209474 -11.0211 classes 0.736842 -11.0215 web, 0.901961 -11.022 significantly. 0.637255 -11.0224 anchor 0.976744 -11.0225 stevenson 0.886792 -11.0227 identifiers 0.833333 -11.0228 ji 0.622642 -11.0229 goes 0.512658 -11.023 context-free 0.249626 -11.0231 alignments 0.809524 -11.0235 illustration 0.976744 -11.0236 analyse 0.463542 -11.0237 da 0.402344 -11.0241 helps 0.976744 -11.0242 useful, 0.93617 -11.0244 empty categories 0.607143 -11.0248 journal of 0.497006 -11.0249 in our experiments, 0.340782 -11.0254 for every 0.93617 -11.0259 running time 0.506173 -11.026 problems. 0.976744 -11.026 diagnostic 0.320896 -11.0261 preposition 0.976744 -11.0261 discriminating 0.976744 -11.0264 lengthening 0.488506 -11.0267 volume 0.976744 -11.0269 unexpected 0.819672 -11.0271 5 discussion 0.93617 -11.0274 h1 0.190061 -11.0274 parameters 0.844828 -11.0274 paid 0.976744 -11.0275 asymmetric 0.491228 -11.0276 acceptable 0.976744 -11.0279 tempeval 0.628571 -11.028 segmentations 0.33515 -11.0281 emotion 0.976744 -11.0285 rationale 0.857143 -11.0287 96 0.649485 -11.0287 progress 0.976744 -11.0287 gazetteer 0.689655 -11.0289 0.75 0.622642 -11.0289 modelling 0.976744 -11.029 time series 0.918367 -11.0291 general framework 0.210753 -11.0291 3. 0.976744 -11.0292 cca 0.428571 -11.0301 analyzing 0.278719 -11.0302 choose 0.976744 -11.0302 1996), 0.976744 -11.0302 considered, 0.126401 -11.0305 for example, 0.328125 -11.031 filter 0.87037 -11.0311 2% 0.2103 -11.0316 binary 0.93617 -11.0317 branch 0.976744 -11.032 900 0.707317 -11.0327 eisner 0.976744 -11.0327 appealing 0.976744 -11.0327 1.7 0.976744 -11.0328 behave 0.976744 -11.0328 model) 0.768116 -11.0338 it is necessary 0.787879 -11.0341 10k 0.916667 -11.0341 phones 0.721519 -11.0342 together, 0.631068 -11.0345 pseudo 0.609091 -11.0347 and klein, 0.557252 -11.035 coefficients 0.760563 -11.0351 1995; 0.250382 -11.0353 introduced 0.976744 -11.0354 organizations 0.9 -11.0355 g1 0.721519 -11.0359 text classification 0.707317 -11.036 similarly to 0.262185 -11.0361 e.g. 0.796875 -11.0362 formed by 0.976744 -11.0364 conflicting 0.339888 -11.037 al., 2009) 0.698795 -11.037 aspects. 0.246657 -11.0371 mapping 0.440758 -11.0375 2009 0.884615 -11.0375 months 0.976744 -11.0375 randomized 0.976744 -11.0376 inconsistency 0.362179 -11.0379 the rest of 0.934783 -11.038 clarke 0.760563 -11.0381 lexemes 0.806452 -11.0381 reductions 0.819672 -11.0382 lexical categories 0.976744 -11.0383 abbreviation 0.87037 -11.0385 loop 0.25273 -11.0385 shows that 0.649485 -11.0386 summary, 0.976744 -11.0388 section 3.1. 0.633663 -11.0388 τ 0.954545 -11.039 recommended 0.75 -11.0393 4-gram 0.976744 -11.0393 hot 0.631068 -11.0394 very different 0.727273 -11.0395 components, 0.916667 -11.0395 next step 0.64 -11.0401 this suggests that 1 -11.0401 field. 1 -11.0401 shown. 0.954545 -11.0402 anchors 0.652632 -11.0402 these, 0.97619 -11.0404 more precisely, 0.356037 -11.0406 pipeline 0.916667 -11.0409 mental 0.87037 -11.041 aim at 0.548148 -11.0411 nodes, 0.884615 -11.0412 folds 0.97619 -11.0413 domain-independent 0.438679 -11.0414 resources. 0.768116 -11.0417 oovs 0.404 -11.0417 developing 0.578512 -11.0422 le 0.796875 -11.0424 confidence score 0.954545 -11.0428 lesser 0.584746 -11.0431 mentions. 0.472527 -11.0435 theoretical 0.477528 -11.0435 input sentence 0.7125 -11.0438 breaks 0.842105 -11.044 increases, 0.5 -11.044 there are two 0.389513 -11.0442 judgments 0.854545 -11.0442 refine 0.854545 -11.0442 2004 0.768116 -11.0442 sense induction 0.548148 -11.0443 genre 0.884615 -11.0444 amani 0.954545 -11.0444 bless 0.9 -11.0448 argument identification 0.415254 -11.0451 application of 0.5625 -11.0453 ng 0.572581 -11.0453 robot 0.619048 -11.0454 taggers 0.97619 -11.0454 4% 0.97619 -11.0458 biased towards 0.816667 -11.0462 86 0.806452 -11.0463 definition, 0.313869 -11.0465 percentage of 0.272894 -11.0469 too 0.784615 -11.047 pos, 0.97619 -11.047 skewed 0.9 -11.0471 grouped into 0.509677 -11.0471 status 0.867925 -11.0472 1.000 0.663043 -11.0472 semantics, 0.884615 -11.0473 more recently, 0.674157 -11.0475 items, 0.698795 -11.0475 rl 0.97619 -11.0476 cues, 0.472527 -11.0477 giving 0.784615 -11.0477 presumably 0.703704 -11.0479 notes 0.806452 -11.0483 different ways 0.179508 -11.0483 during 0.301124 -11.0487 matches 0.97619 -11.0488 phrase-level 0.796875 -11.0489 avg. 0.337079 -11.0491 fragments 0.97619 -11.0495 listeners 0.97619 -11.0495 satisfactory 0.806452 -11.0498 lost 0.954545 -11.05 (rather 0.655914 -11.05 adverb 0.784615 -11.0503 sequences, 0.97619 -11.0503 inappropriate 0.816667 -11.0504 ce 0.389513 -11.0506 an example of 0.954545 -11.0507 association measures 0.125534 -11.0508 • 0.458333 -11.0509 and/or 0.686047 -11.0509 negationslash= 0.867925 -11.051 too many 0.678161 -11.051 equations 0.854545 -11.0513 source, 0.427273 -11.0513 80 0.97619 -11.0514 0.88 0.97619 -11.0516 partner 0.816667 -11.0516 prevents 0.422222 -11.0517 they were 0.591304 -11.0517 detection, 0.867925 -11.052 virtually 0.544118 -11.052 recover 0.733333 -11.0523 lp 0.97619 -11.0524 custom 0.539568 -11.0525 decreases 0.954545 -11.0529 1a 0.784615 -11.0532 correspondence between 0.793651 -11.0532 77 0.733333 -11.0534 correlate with 0.464865 -11.0538 biased 0.97619 -11.0539 hmm-based 0.274254 -11.0539 components 0.613208 -11.0542 dialogue. 0.916667 -11.0542 discussing 0.916667 -11.0545 moderate 0.934783 -11.0545 lm, 0.954545 -11.0549 microblog 0.739726 -11.055 pubmed 0.854545 -11.0552 pay 0.97619 -11.0553 mesh 0.97619 -11.0554 elements, 0.48538 -11.0555 modify 0.97619 -11.0557 comprise 0.776119 -11.0557 appeared in 0.101298 -11.0564 text 0.97619 -11.0565 exceeds 0.636364 -11.0567 shared task. 0.535714 -11.0568 to overcome 0.703704 -11.0568 submodular 0.613208 -11.0569 there are no 0.897959 -11.057 imbalanced 0.97619 -11.0572 wel 0.645833 -11.0575 e, 0.784615 -11.0576 pointwise mutual 0.451777 -11.0577 2002), 0.613208 -11.058 static 0.97619 -11.0582 candidate, 0.723684 -11.0582 high, 0.954545 -11.0582 advance 0.97619 -11.0584 (equation 0.121334 -11.0586 0 0.97619 -11.0588 71 0.97619 -11.0592 setup, 0.230159 -11.0592 summarization 0.816667 -11.0593 language-independent 0.97619 -11.0594 weeks 0.897959 -11.0595 subordinate 0.436019 -11.0599 segmented 0.135184 -11.0602 topic 0.228275 -11.0603 recent 0.690476 -11.0608 weight vector 0.559055 -11.0611 noun, 0.739726 -11.0613 in practice 0.97619 -11.0614 cij 0.273921 -11.0617 once 0.380435 -11.0619 conditions 0.703704 -11.062 gaussian 0.659341 -11.062 dependency parsers 0.827586 -11.062 chen, 0.933333 -11.0627 integer linear programming 0.97619 -11.0627 substituting 0.52027 -11.0629 granularity 0.97619 -11.0635 during training. 0.97619 -11.0636 solution, 0.97619 -11.0638 criteria, 0.746479 -11.064 manual annotation 0.953488 -11.064 avatar 0.784615 -11.0644 present. 1 -11.0647 version. 1 -11.0647 matrices. 1 -11.0647 situation. 0.621359 -11.0647 finite-state 0.431925 -11.0648 reliability 0.933333 -11.0649 3we 0.682353 -11.0651 ≈ 0.97561 -11.0652 selectional preferences 0.723684 -11.0653 smt systems 0.524138 -11.0653 density 0.559055 -11.0653 adequacy 0.459893 -11.0656 try to 0.418502 -11.0656 comments 0.24961 -11.0656 distributional 0.867925 -11.0657 (on 0.953488 -11.0659 clc 0.897959 -11.066 pieces of 0.933333 -11.0661 ccl 0.953488 -11.0663 2002 0.867925 -11.0665 english: 0.587719 -11.0665 usefulness 0.953488 -11.0669 mixing 0.851852 -11.0671 conditioning 0.195215 -11.0675 manually 0.308612 -11.0679 unlike 0.496894 -11.0681 classes, 0.682353 -11.0684 to facilitate 0.276392 -11.0685 answer 0.7 -11.0688 making it 0.474286 -11.0689 calculating 0.520548 -11.069 double 0.97561 -11.069 themes 0.97561 -11.0692 sensible 0.395257 -11.0694 trigram 0.414847 -11.0696 course 0.359477 -11.0703 detecting 0.803279 -11.0705 arises 0.97561 -11.0708 adjusting 0.914894 -11.071 newspapers 0.2225 -11.071 mentions 0.97561 -11.0711 investigations 0.528169 -11.0711 a, 0.97561 -11.0716 japanese, 0.953488 -11.0716 coordinating 0.803279 -11.072 0.82 0.851852 -11.0721 polar 0.772727 -11.0723 we argue that 0.953488 -11.0724 (stolcke, 2002) 0.914894 -11.0727 synonyms, 0.72973 -11.073 obama 0.97561 -11.0731 (wu 0.933333 -11.0735 new, 0.953488 -11.0737 website 0.453125 -11.0739 k = 0.851852 -11.0739 push 0.839286 -11.0742 discourse structure 0.97561 -11.0744 straightforwardly 0.97561 -11.0751 weaker 0.375 -11.0753 demonstrated 0.97561 -11.0755 generalizing 0.349845 -11.0756 settings 0.376812 -11.0758 when using 0.524476 -11.076 big 0.97561 -11.076 subsequently, 0.97561 -11.076 hamming 0.97561 -11.0761 dynamically 0.117709 -11.077 systems 0.736111 -11.0771 e2 0.545455 -11.0772 min 0.7 -11.0776 pitch 0.933333 -11.0781 programs 0.56 -11.0781 library 0.933333 -11.0782 market 0.772727 -11.0782 automaton 0.933333 -11.0784 n2 0.97561 -11.0786 constraint, 0.309756 -11.0786 responses 0.623762 -11.079 looks 0.97561 -11.079 speaker, 0.97561 -11.0792 publications 0.933333 -11.0792 bracket 0.97561 -11.0793 predictions, 0.545455 -11.0794 bottom 0.555556 -11.0796 reaches 0.97561 -11.0797 lowercase 0.97561 -11.0798 works, 0.642105 -11.0798 fillers 0.953488 -11.0799 johns hopkins university 0.5 -11.08 nevertheless, 0.953488 -11.0801 comparable corpora, 0.97561 -11.0802 told 0.132353 -11.0804 extraction 0.72 -11.0806 coming 0.97561 -11.0807 considerations 0.97561 -11.0816 6 related work 0.895833 -11.0816 evaluated against 0.429245 -11.0821 selects 0.32021 -11.0822 kernel 0.742857 -11.0823 real world 0.623762 -11.0824 specifying 0.394422 -11.0824 bias 0.97561 -11.0831 aggressive 0.836364 -11.0834 concatenation of 0.510067 -11.0834 intuitively, 0.97561 -11.0835 punctuation, 0.550388 -11.0836 yielding 0.546154 -11.0837 solved 0.16229 -11.0838 corpora 0.97561 -11.0838 broadcast news 0.510067 -11.0844 model’s 0.736111 -11.0847 gloss 0.262324 -11.0847 complexity 0.429245 -11.0847 10% 0.652174 -11.0847 influenced by 0.97561 -11.085 finegrained 0.753623 -11.0851 we observed that 0.476744 -11.0852 distortion 0.28692 -11.0852 sets of 0.813559 -11.0852 grade 0.97561 -11.0854 t3 0.97561 -11.0854 uc 0.123346 -11.0855 annotation 0.642105 -11.0857 smt system 0.380074 -11.086 down 0.167988 -11.0861 instances 0.769231 -11.0862 enron 0.913043 -11.0864 comparability 0.389105 -11.0867 system’s 0.865385 -11.0876 html 0.836364 -11.0876 psychology 0.761194 -11.0877 pronominal 0.97561 -11.0877 asian 0.865385 -11.088 harvesting 0.865385 -11.088 work well 0.97561 -11.088 ptk 0.97561 -11.0882 rough 0.97561 -11.0883 (up 0.59633 -11.0884 classifiers. 0.23639 -11.0885 sentences, 0.710526 -11.0887 large, 0.836364 -11.089 93 0.742857 -11.0891 becoming 0.172811 -11.0892 reference 0.849057 -11.0892 multiword expressions 0.347826 -11.0893 corpora, 0.710526 -11.0893 a2 0.88 -11.0895 high frequency 0.655556 -11.0897 we tried 0.952381 -11.0897 dmv 0.27668 -11.09 network 1 -11.0901 ordering. 1 -11.0901 added. 1 -11.0901 instead. 1 -11.0901 daum´e iii 1 -11.0901 comparisons. 1 -11.0901 clauses. 0.710526 -11.0908 and roth, 0.537313 -11.0911 updates 0.895833 -11.0912 back-off 0.313924 -11.0915 dynamic 0.913043 -11.0918 100, 0.952381 -11.0918 partitioned 0.405063 -11.0925 patterns. 0.836364 -11.0926 frameworks 0.75 -11.0927 (13) 0.975 -11.0933 gibbs sampler 0.975 -11.0935 heuristically 0.517241 -11.0935 ∗ 0.528986 -11.0936 5-gram 0.810345 -11.0937 links between 0.913043 -11.0941 successfully applied 0.88 -11.0941 humans. 0.8 -11.0943 ast 0.57265 -11.0949 experience 0.732394 -11.0952 section 5, 0.691358 -11.0957 devices 0.895833 -11.0957 lacks 0.381132 -11.0957 recently 0.862745 -11.0958 hardly 0.682927 -11.096 scanning 0.659091 -11.0966 on, 0.913043 -11.0968 profiles 0.513699 -11.0969 ontonotes 0.952381 -11.097 generative story 0.952381 -11.0971 f1-measure 0.461111 -11.0971 begin 0.952381 -11.0972 110 0.541985 -11.0974 90% 0.75 -11.0976 figure 3, 0.931818 -11.0977 8 conclusion 0.849057 -11.0978 (2). 0.975 -11.0979 segmental 0.605769 -11.098 tagging, 0.975 -11.0982 phrase-structure 0.975 -11.0986 decisions, 0.862745 -11.0986 entailed 0.952381 -11.0989 social network 0.849057 -11.0989 subtask 0.393574 -11.099 fragment 0.975 -11.0993 generality 0.659091 -11.0994 encountered 0.931818 -11.0996 question: 0.201477 -11.0997 cluster 0.810345 -11.0998 c) 0.777778 -11.0998 converting 0.810345 -11.0999 tl 0.975 -11.1003 models’ 0.975 -11.1008 differentiate between 0.975 -11.1008 matches, 0.541985 -11.1011 extending 0.8 -11.1011 submissions 0.975 -11.1014 private 0.975 -11.1014 plotted 0.836364 -11.1015 textrunner 0.394309 -11.1015 conll 0.849057 -11.1016 quantified 0.169277 -11.1016 including 0.757576 -11.1017 merges 0.975 -11.102 role, 0.682927 -11.1021 standards 0.975 -11.1021 annotated, 0.701299 -11.1021 most frequently 0.975 -11.1023 modules, 0.952381 -11.1028 word maturity 0.975 -11.1028 xu 0.836364 -11.1029 before. 0.975 -11.103 entailments 0.339339 -11.1031 characteristics 0.552 -11.1031 in our experiments 0.975 -11.1031 referring expressions 0.952381 -11.1032 reference translations. 0.363322 -11.1036 media 0.877551 -11.1038 prevalent 0.244548 -11.1038 discussion 0.862745 -11.1039 recording 0.975 -11.1041 disfluency 0.975 -11.1042 adopts 0.931818 -11.1044 generalizations 0.414414 -11.1045 returns 0.310606 -11.1045 produces 0.975 -11.1045 students’ 0.975 -11.1045 determiners, 0.913043 -11.1046 multiparty 0.696203 -11.1046 converted to 0.638298 -11.1047 state, 0.849057 -11.1047 er 0.429952 -11.1048 multiword 0.913043 -11.1049 people who 0.849057 -11.1052 b: 0.23988 -11.1054 showed 0.651685 -11.1054 arg1 0.8 -11.1054 inherently 0.975 -11.1056 environments 0.862745 -11.1058 preserved 0.975 -11.106 variants, 0.542636 -11.1061 lemmas 0.931818 -11.1066 path. 0.484663 -11.1067 something 0.975 -11.1072 generated, 0.975 -11.1072 ncs 0.849057 -11.1073 lscript 0.833333 -11.1074 san 0.567797 -11.1075 highlight 0.484663 -11.1075 expand 0.975 -11.1075 general-domain 0.192271 -11.1079 topics 0.341463 -11.108 condition 0.422535 -11.1082 al., 2006) 0.877551 -11.1084 nonstandard 0.975 -11.1085 picks 0.975 -11.1086 parameterization 0.151625 -11.1086 general 0.266791 -11.1087 numbers 0.893617 -11.1087 parse tree. 0.757576 -11.1087 association between 0.411504 -11.1087 motivation 0.952381 -11.1088 nlp tasks, 0.765625 -11.1091 mt output 0.975 -11.1093 sarcastic 0.513889 -11.1093 huang 0.931818 -11.1094 penalty. 0.877551 -11.1096 make up 0.911111 -11.1097 transducers 0.468208 -11.1097 salient 0.975 -11.1099 tracks 0.862745 -11.1099 63 0.975 -11.1101 manually created 0.975 -11.1103 numbers, 0.975 -11.1103 α, 0.862745 -11.1104 raw text 0.911111 -11.1106 to: 0.975 -11.1106 picking 0.862745 -11.1109 stop words 0.25784 -11.111 quite 0.73913 -11.111 match. 0.911111 -11.111 arabic-english 0.732394 -11.111 inverted 0.862745 -11.1111 prp 0.616162 -11.1113 r, 0.166545 -11.1116 state 0.911111 -11.1116 opportunities 0.975 -11.1117 130 0.283898 -11.1117 characters 0.651685 -11.1118 alone. 0.975 -11.112 precede 0.877551 -11.1122 exceed 0.952381 -11.1125 fully supervised 0.382239 -11.1125 matched 0.821429 -11.1125 roark 0.580357 -11.1127 function words 0.975 -11.1135 crawled 0.247191 -11.114 takes 0.301435 -11.1141 proper 0.975 -11.1141 readability assessment 0.862745 -11.1143 nns 0.79661 -11.1144 non-terminals 0.975 -11.1146 concentrate on 0.634409 -11.1147 l1 0.975 -11.1147 timeline 0.975 -11.1147 reactions 0.95122 -11.1148 friends 0.692308 -11.1148 fl 0.651685 -11.1149 college 0.706667 -11.1149 semantic space 0.893617 -11.1151 discounts 0.786885 -11.1156 state-ofthe-art 0.679012 -11.1156 makes use 0.250831 -11.1159 a number of 1 -11.116 occurs. 1 -11.116 more. 1 -11.116 computed. 1 -11.116 references. 1 -11.116 changes. 1 -11.116 modifiers. 0.974359 -11.1162 configurations. 0.679012 -11.1169 sds 0.757576 -11.1176 graphical model 0.465517 -11.1178 multinomial 0.765625 -11.1182 facilitates 0.601942 -11.1184 to recognize 0.625 -11.1189 clustering. 0.821429 -11.1189 nb 0.95122 -11.1193 6.4 0.73913 -11.1194 marcu 0.644444 -11.1196 basis for 0.429268 -11.12 referred to 0.666667 -11.1201 informed 0.95122 -11.1203 russian 0.821429 -11.1204 (mt) 0.601942 -11.1208 realistic 0.893617 -11.1212 vv 0.59434 -11.1216 charniak 0.634409 -11.1217 counted 0.481481 -11.1218 for identifying 0.662791 -11.1218 v, 0.215517 -11.1221 mean 0.23696 -11.1221 semantics 0.552846 -11.1222 annotated data 0.493506 -11.1222 further, 0.893617 -11.1223 bc 0.666667 -11.1226 enriched 0.232759 -11.1226 phrase-based 0.455556 -11.1226 file 0.525926 -11.1227 recognizer 0.644444 -11.123 enforce 0.24037 -11.1231 weighted 0.174475 -11.1232 top 0.846154 -11.1234 clasification 0.95122 -11.1235 surface forms 0.203297 -11.1236 combined 0.974359 -11.1237 (shen 0.833333 -11.1239 dr 0.568966 -11.1242 metrics, 0.644444 -11.1243 gesture 0.974359 -11.1243 basis, 0.893617 -11.1243 case study 0.706667 -11.1243 explained by 0.568966 -11.1244 researchers have 0.722222 -11.1245 there have been 0.974359 -11.1247 urdu-english 0.95122 -11.1249 traffic 0.475904 -11.125 alone 0.534351 -11.1253 causal 0.3125 -11.1254 become 0.697368 -11.1254 load 0.191969 -11.1255 coreference 0.875 -11.1261 letters, 0.217997 -11.1262 senses 0.95122 -11.1264 on average. 0.347267 -11.1265 analyses 0.462857 -11.1268 identifies 0.974359 -11.127 satisfying 0.140241 -11.127 average 0.974359 -11.1272 o. 0.634409 -11.1273 pr 0.930233 -11.1273 infants 0.930233 -11.1275 parents 0.974359 -11.1278 aps 0.930233 -11.1281 in isolation. 0.974359 -11.1282 parses, 0.974359 -11.1282 subjects, 0.243671 -11.1284 alternative 0.506849 -11.1284 variant of 0.783333 -11.1286 galley 0.683544 -11.1287 improvements. 0.930233 -11.1287 explanations 0.0635091 -11.1289 = 0.974359 -11.1295 whereby 0.679012 -11.1295 positive or negative 0.553719 -11.1296 the most important 0.974359 -11.1297 clarification 0.655172 -11.1298 in table 2, 0.251265 -11.1299 article 0.227462 -11.1301 decision 0.746269 -11.1302 valence 0.544 -11.1302 incorrectly 0.390947 -11.1303 sms 0.634409 -11.1304 ones, 0.79661 -11.1306 interaction between 0.974359 -11.1307 extent, 0.712329 -11.1307 5000 0.86 -11.1307 0.20 0.807018 -11.1308 confinement 0.305764 -11.1309 equivalent 0.974359 -11.131 sentence; 0.974359 -11.1311 searched 0.534351 -11.1311 scaling 0.753846 -11.1312 purposes, 0.974359 -11.1312 knows 0.911111 -11.1316 source. 0.974359 -11.1317 lacking 0.974359 -11.132 intractable 0.596154 -11.1321 this task. 0.207898 -11.1323 indicates 0.930233 -11.1323 79 0.670732 -11.1326 dependency graph 0.522059 -11.1328 collins 0.95122 -11.1329 primitive 0.974359 -11.133 differences, 0.974359 -11.133 aggregation 0.61 -11.1331 hashtags 0.237879 -11.1333 e.g., 0.25 -11.1338 technique 0.974359 -11.134 harvested 0.95122 -11.134 noun phrases, 0.930233 -11.1341 consequences 0.95122 -11.1342 han 0.974359 -11.1344 prepositions, 0.974359 -11.1344 tests, 0.132084 -11.1345 time 0.290909 -11.1346 false 0.783333 -11.1346 aproach 0.314667 -11.1348 to capture 0.974359 -11.135 penalties 0.522059 -11.1354 statements 0.974359 -11.1354 contemporary 0.397436 -11.1354 describing 0.974359 -11.1355 2003 0.279412 -11.1356 selecting 0.974359 -11.136 gm 0.165803 -11.136 level 0.807018 -11.1363 unconstrained 0.564103 -11.1363 event, 0.761905 -11.1363 normalizing 0.95122 -11.1365 parse correction 0.735294 -11.1366 acronyms 0.40625 -11.1366 configuration 0.974359 -11.1368 left, 0.71831 -11.1376 classifies 0.670732 -11.1377 negative, 0.457627 -11.1379 it has been 0.0643399 -11.1384 these 0.341772 -11.1385 reasonable 0.280851 -11.1387 u 0.830189 -11.139 jiang 0.974359 -11.1394 abbreviated 0.742424 -11.1395 normalisation 0.974359 -11.1396 long distance 0.670732 -11.14 host 0.46988 -11.1403 language model. 0.783333 -11.1403 η 0.974359 -11.1404 pos-tagged 0.909091 -11.1404 human-generated 0.177626 -11.1406 contain 0.974359 -11.1409 98 0.496644 -11.1413 away 0.95 -11.1416 methods: 0.783333 -11.1417 designing 0.571429 -11.1418 e1 0.875 -11.1418 hypernyms 0.974359 -11.1419 product, 0.928571 -11.142 hierarchical structure 0.457143 -11.1421 finds 0.95 -11.1422 formalized 1 -11.1427 convergence. 1 -11.1427 vol. 1 -11.1427 limitations. 1 -11.1427 seeds. 0.95 -11.1428 partners 0.187441 -11.1428 individual 0.95 -11.1434 accounted for 0.973684 -11.1435 packed forest 0.830189 -11.1436 yahoo! 0.761905 -11.1436 exactly the same 0.928571 -11.1439 author, 0.973684 -11.144 markov logic 0.891304 -11.1442 (without 0.461988 -11.1444 φ 0.683544 -11.1445 an online 0.875 -11.1446 non-negative 0.818182 -11.1447 m2 0.928571 -11.1451 zhang et al. 0.95 -11.1452 binarized 0.973684 -11.1456 empty elements 0.518248 -11.1456 study. 0.352542 -11.1458 suggests that 0.702703 -11.1462 education 0.226257 -11.1463 type of 0.843137 -11.1464 contextual information 0.761905 -11.1464 skip 0.514493 -11.147 originally 0.783333 -11.147 usages 0.95 -11.1474 (baroni 0.295238 -11.1474 levels 0.793103 -11.1476 perception 0.216945 -11.1477 group 0.322857 -11.148 students 0.95 -11.1484 objective, 0.909091 -11.1489 plsa 0.928571 -11.1492 linguistic phenomena 0.325581 -11.1493 ontology 0.830189 -11.1494 94 0.909091 -11.1494 morphology, 0.875 -11.1494 symbol. 0.708333 -11.1496 automatically. 0.708333 -11.1496 3), 0.312834 -11.1496 induction 0.95 -11.1499 war 0.702703 -11.1499 translates 0.857143 -11.1502 basque 0.630435 -11.1505 stems 0.973684 -11.1506 (manning 0.675 -11.1506 reranker 0.973684 -11.1508 quick 0.973684 -11.1508 templates, 0.973684 -11.1509 equals 0.467066 -11.1509 reach 0.510791 -11.1509 c, 0.662651 -11.1511 ri 0.403587 -11.1515 mapped 0.973684 -11.1515 cost, 0.651163 -11.1515 plaintext 0.830189 -11.1515 drug 0.554622 -11.1516 remain 0.973684 -11.1519 engine, 0.590476 -11.1519 formulated 0.675 -11.152 additional information 0.928571 -11.1521 privacy 0.973684 -11.1522 occasionally 0.909091 -11.1522 playing 0.621053 -11.1522 95% 0.483871 -11.1523 discussions 0.973684 -11.1527 unbounded 0.891304 -11.1528 dependency path 0.416268 -11.1531 remains 0.793103 -11.1532 writer 0.493333 -11.1534 to solve 0.803571 -11.1534 tractable 0.928571 -11.1535 strings, 0.973684 -11.1536 bounds 0.662651 -11.1536 c&c 0.6 -11.1538 parallel sentences 0.572727 -11.1538 interpreted as 0.928571 -11.1541 interacting with 0.909091 -11.1542 conveyed by 0.857143 -11.1543 feature-based 0.928571 -11.1544 march 0.973684 -11.1546 tunable 0.5 -11.1548 2008 0.555556 -11.1549 literal 0.814815 -11.1553 fscore 0.973684 -11.1554 wikipedia’s 0.793103 -11.1554 deciding 0.793103 -11.1554 basically 0.403587 -11.1556 differences in 0.973684 -11.1558 themselves, 0.973684 -11.1561 frequency-based 0.623656 -11.1562 sites 0.973684 -11.1562 unnecessary 0.973684 -11.1568 non-parallel 0.857143 -11.157 bergsma 0.973684 -11.1574 operation, 0.973684 -11.1577 inserting 0.95 -11.1577 issues, 0.888889 -11.1582 stand for 0.928571 -11.1584 sarcasm 0.973684 -11.1586 nearby 0.973684 -11.1587 variable, 0.714286 -11.159 alleviate 0.973684 -11.1591 fixed, 0.459302 -11.1591 tuple 0.614583 -11.1593 site 0.572727 -11.1593 xi 0.973684 -11.1594 organization, 0.973684 -11.1594 preservation 0.770492 -11.1594 energy 0.714286 -11.1595 keys 0.427136 -11.1596 source sentence 0.459302 -11.1603 accurately 0.973684 -11.1605 acts, 0.770492 -11.1606 do you 0.459302 -11.1606 are: 0.679487 -11.1612 histogram 0.514706 -11.1617 instructions 0.486842 -11.162 nonterminals 0.973684 -11.1622 previously, 0.333333 -11.1622 confusion 0.906977 -11.1625 0.71 0.973684 -11.163 nes 0.888889 -11.163 wiebe 0.654762 -11.163 user. 0.666667 -11.1631 disease 0.906977 -11.1631 atom 0.567568 -11.1634 semantic class 0.69863 -11.1634 120 0.95 -11.1638 5 related work 0.407407 -11.1639 lemma 0.608247 -11.1642 compiled 0.906977 -11.1642 affixes 0.271984 -11.1643 construction 0.973684 -11.1644 tight 0.658537 -11.1645 we remove 0.383673 -11.1646 great 0.285068 -11.1646 classified 0.633333 -11.1647 uniformly 0.163462 -11.1648 content 0.731343 -11.1649 classic 0.906977 -11.165 washington 0.973684 -11.1651 combinations, 0.69863 -11.1651 word forms 0.973684 -11.1655 oral 0.888889 -11.1656 language pairs, 0.803571 -11.1656 heavy 0.906977 -11.1658 binomials 0.87234 -11.1661 noun compound 0.19375 -11.1661 low 0.417476 -11.1663 features: 0.693333 -11.1663 tagging. 0.973684 -11.1665 patients 0.256881 -11.1665 extracting 0.814815 -11.1677 cat 0.332308 -11.1679 restricted 0.522727 -11.1682 i, 0.626374 -11.1682 comprehension 0.523077 -11.1682 leverage 0.906977 -11.1682 cascade 0.731343 -11.1683 decreased 0.738462 -11.1686 bigger 0.789474 -11.1689 interests 0.973684 -11.169 species 0.779661 -11.1692 0.15 0.617021 -11.1693 billion 0.75 -11.1694 much smaller 0.179372 -11.1695 still 0.507246 -11.1698 mst 0.906977 -11.1698 prototypical 0.658537 -11.17 positively 1 -11.1701 bias. 1 -11.1701 broken down 0.972973 -11.1702 slots. 0.826923 -11.1703 supervision. 0.948718 -11.1704 literary 0.412322 -11.1706 computational linguistics, pages 0.518797 -11.1708 metric, 0.948718 -11.1709 notably, 0.670886 -11.171 became 0.738462 -11.171 capabilities 0.474684 -11.1711 don’t 0.758065 -11.1716 vn 0.766667 -11.1717 primitives 0.608247 -11.1718 utterances, 0.972973 -11.1726 direct transfer 0.658537 -11.1728 stress 0.948718 -11.1728 substitutes 0.13806 -11.173 documents 0.187255 -11.1731 made 0.461078 -11.1732 emotional 0.87234 -11.1737 verification 0.948718 -11.1742 fmeasure 0.208232 -11.1743 metrics 0.87234 -11.1743 do so 0.40553 -11.1746 filtered 0.87234 -11.1746 categorical 0.720588 -11.1751 cross-language 0.972973 -11.1751 albeit 0.826923 -11.1751 agree with 0.461078 -11.1752 layer 0.567568 -11.1754 an n-gram 0.344482 -11.1754 sufficient 0.84 -11.1755 tools. 0.689189 -11.1755 lexical similarity 0.926829 -11.176 dialogues, 0.888889 -11.1762 part-whole 0.30102 -11.1763 minimum 0.926829 -11.1765 relatively low 0.388186 -11.1767 levels of 0.720588 -11.1767 distance. 0.972973 -11.1767 appropriately 0.112248 -11.1768 syntactic 0.766667 -11.1769 influenza 0.689189 -11.177 lexeme 0.474684 -11.1771 considerable 0.458333 -11.1772 contexts. 0.926829 -11.1772 often, 0.948718 -11.1778 termed 0.507246 -11.178 we employ 0.906977 -11.1781 birth 0.789474 -11.1782 pages. 0.888889 -11.1783 inflections 0.84 -11.1784 appropriateness 0.948718 -11.1785 ideally 0.372549 -11.1787 implementation of 0.972973 -11.1788 categorize 0.298246 -11.1788 evaluating 0.789474 -11.1789 83 0.854167 -11.179 adopting 0.149095 -11.1793 negative 0.557522 -11.1796 normally 0.972973 -11.1799 vandalism 0.416667 -11.18 directions 0.580952 -11.1802 plural 0.527559 -11.1803 argue that 0.972973 -11.1803 word-to-word 0.972973 -11.1803 unobserved 0.435484 -11.1804 advanced 0.496503 -11.181 wikipedia. 0.84 -11.1813 responsiveness 0.720588 -11.1818 starts with 0.223001 -11.1822 for instance, 0.727273 -11.1823 similarity between two 0.321637 -11.1823 n-best 0.869565 -11.1824 overlap, 0.689189 -11.1824 an effective 0.8 -11.1824 biology 0.789474 -11.1826 majority class 0.926829 -11.1827 stsg 0.926829 -11.1829 hdp 0.906977 -11.1829 0.30 0.8 -11.183 euclidean 0.228951 -11.1832 built 0.972973 -11.1835 imperative 0.84 -11.1837 discarding 0.224928 -11.1841 added 0.972973 -11.1841 conducting 0.972973 -11.1841 unreliable 0.972973 -11.1845 passing 0.541667 -11.1845 percent 0.972973 -11.1847 up, 0.972973 -11.1847 s2net 0.972973 -11.1847 counter 0.480263 -11.1847 singular 0.972973 -11.185 each, 0.972973 -11.1852 transcript 0.972973 -11.1854 information: 0.710145 -11.1861 meet 0.746032 -11.1862 to enhance 0.972973 -11.1862 oc 0.972973 -11.1865 news. 0.246552 -11.1865 contexts 0.694444 -11.1865 helped 0.84 -11.1866 descent 0.746032 -11.1867 student’s 0.926829 -11.187 fixing 0.906977 -11.187 incorrect. 0.972973 -11.1872 multiclass 0.926829 -11.1873 (2002), 0.926829 -11.1874 mte 0.972973 -11.1875 monotonic 0.231355 -11.1877 items 0.766667 -11.1881 home 0.869565 -11.1882 button 0.869565 -11.1883 choi 0.926829 -11.1885 dcs 0.972973 -11.1885 best-performing 0.972973 -11.1887 page, 0.823529 -11.189 sampling, 0.727273 -11.1891 end-to-end 0.972973 -11.1891 regularizer 0.972973 -11.1892 restricts 0.642857 -11.1892 ned 0.563636 -11.1892 even more 0.68 -11.1893 necessary. 0.972973 -11.1897 1994), 0.232308 -11.1897 though 0.972973 -11.1898 attacks 0.68 -11.1901 daily 0.972973 -11.1901 preceded by 0.582524 -11.1905 to infer 0.904762 -11.1906 massive 0.563636 -11.1907 can improve 0.972973 -11.1908 analysis: 0.239804 -11.1908 appropriate 0.430851 -11.1909 supervision 0.650602 -11.1913 to illustrate 0.972973 -11.1913 firstly 0.29602 -11.1914 update 0.972973 -11.1915 character, 0.904762 -11.1917 (possibly 0.582524 -11.1918 (zhang 0.926829 -11.192 preferences. 0.811321 -11.192 ranges from 0.387931 -11.1921 (table 0.511278 -11.1921 learning algorithm 0.972973 -11.1922 6.5 0.972973 -11.1924 f-score, 0.904762 -11.1926 better understand 0.710145 -11.1929 attributes, 0.511278 -11.193 (2011) 0.823529 -11.1931 weight. 0.972973 -11.1931 iv 0.184825 -11.1931 comparison 0.293827 -11.1933 combinations 0.972973 -11.1937 well suited 0.675325 -11.1939 complexity. 0.972973 -11.194 license 0.886364 -11.1951 narrow 0.972973 -11.1952 parameter settings 0.948718 -11.1953 chains. 0.775862 -11.1957 there may be 0.191983 -11.1958 our approach 0.972973 -11.1959 set; 0.523438 -11.1959 have no 0.886364 -11.1961 non-expert 0.886364 -11.1963 in some cases, 0.785714 -11.1966 0.90 0.972973 -11.1966 vignette 0.445714 -11.1967 may have 0.972973 -11.1967 make sure 0.972973 -11.1968 wait 0.32622 -11.1971 useful for 0.654321 -11.1972 genes 0.811321 -11.1972 anotation 0.775862 -11.1977 combinatorial 0.410628 -11.1978 acl 0.775862 -11.1982 overfitting 0.886364 -11.1982 wikulu 1 -11.1983 hindi. 1 -11.1983 cross validation. 1 -11.1983 occur. 1 -11.1983 created. 0.785714 -11.1984 deployed 0.727273 -11.1989 reference, 0.886364 -11.1989 dev test 0.532787 -11.1992 systematic 0.904762 -11.1994 ab 0.716418 -11.1999 games 0.716418 -11.2001 feature representation 0.625 -11.2001 perceived 0.947368 -11.2003 if, 0.356364 -11.2003 (figure 0.278396 -11.2005 want to 0.836735 -11.2008 nsummationdisplay i=1 0.947368 -11.2012 (especially 0.886364 -11.2013 conceptresolver 0.796296 -11.2019 formality 0.426316 -11.2021 combinations of 0.972222 -11.2025 ppi 0.658228 -11.2026 cultural 0.248227 -11.2028 making 0.851064 -11.2037 feasibility of 0.443182 -11.2039 is usually 0.642857 -11.2039 1993). 0.972222 -11.2043 enriching 0.972222 -11.2046 pos-based 0.50365 -11.2051 6, 0.666667 -11.2053 iff 0.972222 -11.2054 cohesive 0.741935 -11.2056 when comparing 0.886364 -11.2056 dialogs 0.972222 -11.2058 nearest neighbor 0.972222 -11.2059 dealt with 0.851064 -11.2062 section describes 0.140893 -11.2064 sentiment 0.851064 -11.2068 enhancing 0.741935 -11.2075 each. 0.851064 -11.2079 raised 0.972222 -11.2079 information retrieval. 0.18866 -11.2081 provides 0.392857 -11.2081 system combination 0.576923 -11.2084 target language. 0.972222 -11.2086 scoping 0.836735 -11.2092 resnik 0.972222 -11.2093 skeleton 0.947368 -11.2094 koppel 0.925 -11.2095 100,000 0.467949 -11.2096 there were 0.836735 -11.2096 complements 0.866667 -11.2099 0.67 0.866667 -11.2099 latin 0.972222 -11.21 everyday 0.465409 -11.21 leave 0.45509 -11.2101 artificial 0.807692 -11.2101 hashtag 0.823529 -11.2102 h, 0.35 -11.2102 also, 0.723077 -11.2102 to acquire 0.35 -11.2103 check 0.904762 -11.2105 inconsistencies 0.972222 -11.2109 outcomes 0.866667 -11.2109 credit 0.972222 -11.2111 deductive 0.866667 -11.2116 translation: 0.925 -11.2117 hyperedges 0.925 -11.2117 periods 0.7 -11.2117 1% 0.972222 -11.212 outlines 0.972222 -11.212 formats 0.972222 -11.2121 sizes, 0.625 -11.2125 investigating 0.972222 -11.2127 suite 0.553571 -11.2131 confirm 0.482993 -11.2132 china 0.206725 -11.2133 nlp 0.836735 -11.2134 kevin 0.972222 -11.2135 responses, 0.807692 -11.2136 cp 0.972222 -11.2137 mutual k-nn 0.263052 -11.2138 whole 0.705882 -11.2138 vi 0.947368 -11.2142 “what 0.585859 -11.2142 person, 0.281755 -11.2143 a good 0.851064 -11.2144 correction, 0.972222 -11.2145 destination 0.972222 -11.2145 decode 0.627907 -11.2146 purposes. 0.972222 -11.2146 failures 0.762712 -11.2146 lmbot 0.658228 -11.2146 our approach. 0.690141 -11.2147 actions, 0.972222 -11.2147 alternations 0.675676 -11.2147 (2008). 0.972222 -11.2151 procesing 0.972222 -11.2153 drastically 0.972222 -11.2155 connectives, 0.608696 -11.2156 stored in 0.507576 -11.2157 even when 0.7 -11.2159 phosphorylation 0.285036 -11.2159 would like to 0.553571 -11.216 characterize 0.925 -11.2162 japanese. 0.947368 -11.2162 top ranked 0.459627 -11.2162 labeled with 0.972222 -11.2164 acceptance 0.925 -11.2165 prompt 0.925 -11.2165 unit, 0.723077 -11.2167 twenty 0.925 -11.2169 in both cases, 0.82 -11.217 crfs 0.608696 -11.2171 closure 0.866667 -11.2176 our proposed approach 0.762712 -11.2178 lda, 0.266667 -11.2178 majority 0.507576 -11.2178 lexicon, 0.972222 -11.218 boston 0.851064 -11.2181 20, 0.972222 -11.2184 keyboard 0.402844 -11.2188 representative 0.248201 -11.2189 (in 0.902439 -11.2192 a0 0.164038 -11.2192 four 0.972222 -11.2193 reversed 0.925 -11.2198 information extraction, 0.400943 -11.2201 an average 0.972222 -11.2204 simplifies 0.723077 -11.2204 certainly 0.781818 -11.2206 lies in 0.972222 -11.2208 inversion 0.972222 -11.2211 gsc 0.662338 -11.2213 reinforcement 0.627907 -11.2216 input: 0.662338 -11.2216 switchboard 0.18815 -11.2223 therefore 0.972222 -11.2224 perceptual 0.77193 -11.2225 different kinds 0.972222 -11.2228 metamap 0.49635 -11.2229 comprehensive 0.972222 -11.223 174 0.833333 -11.2231 (petrov et 0.972222 -11.2232 riloff, 0.332248 -11.2232 * 0.883721 -11.2234 previous works 0.238333 -11.2234 interesting 0.972222 -11.2235 compression rate 0.314869 -11.2239 reflect 0.972222 -11.224 116 0.608696 -11.224 and klein 0.851064 -11.2241 infectious 0.627907 -11.2242 members of 0.781818 -11.2242 3rd 0.972222 -11.2244 negotiation 0.259406 -11.2249 (b) 0.638554 -11.2249 reversible 0.548673 -11.225 in table 4. 0.5 -11.225 fuzzy 0.972222 -11.2252 148 0.528926 -11.2252 claim 0.866667 -11.2255 information gain 0.972222 -11.2257 f-measures 0.65 -11.2258 mechanisms 0.851064 -11.2259 n1 0.972222 -11.226 (though 0.972222 -11.226 truncated 0.972222 -11.2261 meanwhile, 0.972222 -11.2262 pause 0.54386 -11.2265 classification accuracy 0.945946 -11.2265 confidence estimation 0.435754 -11.2266 entities, 0.792453 -11.2266 tokenizer 0.566038 -11.2267 shift-reduce 0.77193 -11.227 paterns 1 -11.2273 scope. 1 -11.2273 titles. 1 -11.2273 gain. 1 -11.2273 prepositions. 1 -11.2273 4.2. 1 -11.2273 0.1. 1 -11.2273 cube pruning 0.902439 -11.2274 derivational 0.587629 -11.228 explains 0.372951 -11.228 tuned 0.75 -11.2282 overall accuracy 0.180249 -11.2286 resulting 0.75 -11.2291 tau 0.595745 -11.2291 soon 0.945946 -11.2294 3.6 0.533898 -11.2301 solely 0.847826 -11.2302 content selection 0.65 -11.2304 semiring 0.248629 -11.2306 part-of-speech 0.397196 -11.2306 took 0.971429 -11.2307 fuzzy match 0.449102 -11.2312 surrounding 0.322086 -11.2313 setup 0.883721 -11.2314 eat 0.866667 -11.2315 positive, negative 0.902439 -11.2316 accompanied 0.945946 -11.2317 follows, 0.560748 -11.2317 ngrams 0.65 -11.2321 showing that 0.971429 -11.2323 pad´o 0.833333 -11.2325 help us 0.902439 -11.2326 wang, 0.792453 -11.2328 peer-review 0.82 -11.2329 advances in 0.489362 -11.2333 not always 0.883721 -11.2333 we now describe 0.758621 -11.2335 joint probability 0.321101 -11.2338 indicator 0.883721 -11.2338 penn treebank. 0.971429 -11.2339 receiving 0.567308 -11.234 these approaches 0.945946 -11.234 word lattices 0.82 -11.2341 application, 0.54386 -11.2344 confirmed 0.503817 -11.2345 occurrence of 0.71875 -11.2347 language models, 0.311239 -11.2348 observation 0.245081 -11.2349 to evaluate 0.71875 -11.2349 error, 0.971429 -11.2351 transmission 0.489209 -11.2352 newly 0.529412 -11.2353 ranges 0.653846 -11.2354 to resolve 0.737705 -11.2354 fragments. 0.730159 -11.2357 67 0.737705 -11.2357 brackets 0.680556 -11.2358 stopping 0.595745 -11.2359 noted that 0.758621 -11.2361 simplifying 0.641975 -11.2363 others, 0.82 -11.2364 zero, 0.971429 -11.2364 freedom 0.971429 -11.2365 0.99 0.923077 -11.2367 jaccard 0.730159 -11.2369 january 0.163233 -11.2369 show that 0.334448 -11.2373 designed to 0.520325 -11.2379 indices 0.393519 -11.2379 to distinguish 0.971429 -11.238 converts 0.971429 -11.2381 information-status 0.883721 -11.2381 language pair, 0.883721 -11.2382 lexicographic 0.653846 -11.2384 probability, 0.863636 -11.2385 affecting 0.737705 -11.2385 discounting 0.666667 -11.2389 percentages 0.971429 -11.2389 pipelined 0.833333 -11.2392 an intermediate 0.730159 -11.2394 pre-test 0.393519 -11.2395 said 0.971429 -11.2397 effort, 0.971429 -11.2399 3.8 0.567308 -11.2402 revealed 0.945946 -11.2402 123 0.971429 -11.2402 methodology, 0.971429 -11.2402 trading 0.758621 -11.2404 academic 0.307042 -11.2404 was used 0.544643 -11.2407 disagreement 0.595745 -11.2407 names. 0.64557 -11.2411 services 0.923077 -11.2411 apple 0.630952 -11.2414 phrase extraction 0.671233 -11.2416 mj 0.971429 -11.2417 posed 0.520325 -11.2417 adapting 0.492754 -11.2418 continue 0.971429 -11.2419 puts 0.737705 -11.2419 (2), 0.333333 -11.2421 programming 0.745763 -11.2421 guess 0.971429 -11.2422 colors 0.520325 -11.2423 sorted 0.971429 -11.2425 crowd 0.349265 -11.2425 5, 0.712121 -11.2428 f-score. 0.971429 -11.2429 elaboration 0.135194 -11.2429 rule 0.945946 -11.2439 end up 0.971429 -11.2446 alternation 0.792453 -11.2448 mary 0.816327 -11.2448 wikipedia article 0.971429 -11.2457 forces 0.971429 -11.2457 reviewing 0.634146 -11.2458 enabling 0.971429 -11.2461 described below. 0.971429 -11.2461 simultaneous 0.847826 -11.2462 adult 0.971429 -11.2462 devoted to 0.971429 -11.2463 multivariate 0.971429 -11.2464 web1t 0.971429 -11.2465 cross-document 0.945946 -11.2468 pearson correlation 0.777778 -11.2469 regularized 0.767857 -11.247 d., 0.971429 -11.2471 matrices, 0.971429 -11.2471 fractional 0.971429 -11.2472 falling 0.971429 -11.2473 interface, 0.737705 -11.2474 ex 0.971429 -11.2477 unbalanced 0.792453 -11.2478 predicate, 0.971429 -11.2481 optimisation 0.923077 -11.2481 digits 0.767857 -11.2484 recovered 0.971429 -11.2484 hyperparameter 0.767857 -11.2488 possessive 0.971429 -11.2489 reasons: 0.671233 -11.249 8. 0.971429 -11.2494 wrote 0.767857 -11.2496 frequently used 0.272523 -11.2497 languages, 0.923077 -11.2498 context-sensitive 0.161189 -11.2499 perform 0.330033 -11.2501 is still 0.971429 -11.2501 sbj 0.945946 -11.2502 word order, 0.725806 -11.2502 word boundaries 0.971429 -11.2503 irrespective of 0.880952 -11.2504 word alignments. 0.516129 -11.2504 back to 0.863636 -11.2504 dutch, 0.606742 -11.2507 ns 0.971429 -11.2507 bag-of-word 0.737705 -11.2507 modality 0.880952 -11.2508 title, 0.492647 -11.2511 y. 0.971429 -11.2511 minimally 0.971429 -11.2511 concentrated 0.556604 -11.2513 positive, 0.233169 -11.2514 edges 0.754386 -11.2516 deriving 0.971429 -11.2516 blogs, 0.971429 -11.2516 0.09 0.863636 -11.2516 continuation 0.971429 -11.2517 combined, 0.671233 -11.2519 tradeoff 0.971429 -11.2521 demands 0.9 -11.2528 (but not 0.847826 -11.2529 usable 0.233831 -11.253 version of 0.33677 -11.2531 showing 0.971429 -11.2531 computational complexity 0.388128 -11.2531 encode 0.816327 -11.2531 em, 0.357977 -11.2532 detail 0.971429 -11.2534 prepare 0.691176 -11.2535 stimuli 0.803922 -11.2538 algorithmic 0.971429 -11.2539 bip 0.349442 -11.2543 suggested 0.971429 -11.2545 lowercased 0.971429 -11.2547 nyt 0.971429 -11.2547 roc 0.971429 -11.2548 long, 0.971429 -11.255 habash 0.64557 -11.255 activity 0.971429 -11.2553 essays 0.971429 -11.2554 blanc 0.863636 -11.2554 we omit 0.971429 -11.2556 idioms 0.685714 -11.2557 mit 0.507812 -11.2557 cache 0.662162 -11.2557 0.25 0.485714 -11.2559 going 0.803922 -11.256 log-linear model 0.829787 -11.2561 nonetheless, 0.301105 -11.2562 2010 0.880952 -11.2563 slovak 0.921053 -11.2563 telephone 0.944444 -11.2564 acquisition, 0.944444 -11.2566 dictionary-based 0.816327 -11.257 to assist 1 -11.2571 evidence. 1 -11.2571 tokens). 1 -11.2571 preposition. 1 -11.2571 accuracies. 1 -11.2571 overlap. 1 -11.2571 page. 1 -11.2571 root. 0.174721 -11.2574 (2010) 0.944444 -11.2575 decomposes 0.803922 -11.2577 second step 0.777778 -11.2577 expense of 0.235593 -11.2578 threshold 0.492537 -11.2578 predefined 0.803922 -11.258 permutations 0.421622 -11.2581 findings 0.662162 -11.2585 analysed 0.6375 -11.2586 more specifically, 0.970588 -11.2587 harmonic mean 0.944444 -11.2588 corpora: 0.575758 -11.2593 srilm 0.583333 -11.2593 (but 0.423077 -11.2593 american 0.944444 -11.2595 mining, 0.788462 -11.2597 words; 0.829787 -11.2598 experimentation 0.970588 -11.2599 transforms 0.707692 -11.2599 neighborhood 0.231908 -11.2602 run 0.944444 -11.2603 102 0.725806 -11.2605 0% 0.606742 -11.2606 failure 0.970588 -11.2606 punctuation marks 0.362903 -11.2606 bayes 0.921053 -11.2606 semcor 0.261506 -11.2608 fixed 0.944444 -11.2611 join 0.777778 -11.2611 0.33 0.609195 -11.2612 significant difference 0.336806 -11.2613 0.5 0.9 -11.2615 amt 0.754386 -11.2617 notably 0.388889 -11.2618 modifier 0.970588 -11.262 discriminatively 0.725806 -11.262 concept, 0.777778 -11.262 0.02 0.816327 -11.2623 handles 0.944444 -11.2628 mistake 0.9 -11.2628 0.59 0.676056 -11.2628 additive 0.9 -11.2629 transductive 0.944444 -11.2634 works well 0.788462 -11.2635 working on 0.844444 -11.2635 editors 0.944444 -11.2636 f2 0.25498 -11.2638 collected 0.921053 -11.2639 malay 0.844444 -11.2642 (pradhan 0.844444 -11.2644 meaning representation 0.609195 -11.2647 (iii) 0.429379 -11.2648 sent 0.970588 -11.2651 (nivre, 0.970588 -11.2653 during decoding, 0.402985 -11.2654 applicable 0.921053 -11.2654 non-parametric 0.9 -11.2656 modified version 0.396135 -11.2656 overview of 0.384615 -11.2657 agent 0.777778 -11.2657 idiosyncratic 0.763636 -11.2658 differing 0.707692 -11.2658 places 0.564356 -11.2661 proteins 0.777778 -11.2661 comma 0.970588 -11.2664 movements 0.970588 -11.2668 corpus) 0.788462 -11.2668 turkers 0.8 -11.2668 validated 0.616279 -11.2669 educational 0.777778 -11.2669 vowels 0.844444 -11.267 significantly improves 0.844444 -11.2673 disputant 0.653333 -11.2676 0.65 0.970588 -11.2677 (haghighi 0.6375 -11.2678 please 0.844444 -11.2679 formalisms 0.69697 -11.2679 objects, 0.381166 -11.2682 unseen 0.591398 -11.2683 distances 0.387097 -11.2685 finally 0.970588 -11.269 judgments, 0.676056 -11.2691 in figure 2, 0.944444 -11.2692 grammatical errors 0.970588 -11.2693 known, 0.69697 -11.2694 0.50 0.970588 -11.2695 (described 0.435294 -11.2695 desired 0.970588 -11.2696 germanic 0.829787 -11.2699 dense 0.788462 -11.27 martin 0.455128 -11.2704 am 0.714286 -11.2705 attempting 0.970588 -11.2707 given, 0.285 -11.2707 log 0.970588 -11.2711 offline 0.406091 -11.2712 introducing 0.970588 -11.2713 accepting 0.431034 -11.2717 ordered 0.970588 -11.2718 curent 0.970588 -11.2718 pedersen, 0.970588 -11.2719 posible 0.970588 -11.2721 format, 0.307918 -11.2722 empty 0.970588 -11.2725 quoted 0.970588 -11.2726 exhibited 0.970588 -11.2726 websites 0.970588 -11.2726 prohibitively 0.921053 -11.2728 rules: 0.970588 -11.2729 unfortunately 0.8 -11.2729 lie 0.970588 -11.273 pipeline, 0.860465 -11.273 d1 0.254 -11.2731 reduce 0.676056 -11.2732 operating 0.970588 -11.2733 accumulated 0.258799 -11.2736 conditional 0.970588 -11.2738 fell 0.530435 -11.2739 make it 0.9 -11.2742 customer 0.763636 -11.2742 form: 0.763636 -11.2743 varieties 0.653333 -11.2743 publication 0.714286 -11.2743 our method, 0.829787 -11.2744 mt. 0.564356 -11.2746 j, 0.921053 -11.2746 i think 0.457516 -11.2746 datasets. 0.878049 -11.2747 machine reading 0.482014 -11.2747 establish 0.970588 -11.2748 requested 0.194836 -11.2748 b 0.237762 -11.2749 (or 0.878049 -11.2751 colours 0.512195 -11.2753 margin 0.733333 -11.2755 part, 0.578947 -11.2755 strength of 0.970588 -11.2758 propositional 0.351145 -11.2759 averaged 0.176471 -11.2759 trees 0.585106 -11.276 reviews, 0.970588 -11.2762 selective 0.329966 -11.2762 hence, 0.593407 -11.2762 (2000) 0.860465 -11.2762 essay 0.970588 -11.2766 characterizing 0.641026 -11.2767 language technologies 0.32459 -11.2774 differ 0.611765 -11.2774 indexed 0.970588 -11.2775 stopped 0.878049 -11.2778 (left) 0.621951 -11.278 ca 0.970588 -11.2782 thresholding 0.315789 -11.2782 may not 0.62963 -11.2782 sentence level 0.970588 -11.2784 person’s 0.386047 -11.2785 normal 0.666667 -11.2788 800 0.970588 -11.279 has proven 0.970588 -11.279 sample, 0.666667 -11.2792 md 0.844444 -11.2795 excluded from 0.773585 -11.2795 also includes 0.970588 -11.2796 inaccurate 0.970588 -11.2798 news articles, 0.970588 -11.28 (0 0.8125 -11.28 section 3 describes 0.970588 -11.2803 hyperlinks 0.970588 -11.2805 targets, 0.602273 -11.2806 built from 0.970588 -11.2811 5we 0.428571 -11.2812 unfortunately, 0.897436 -11.2812 stance 0.970588 -11.2813 conjunctive 0.970588 -11.2814 policy, 0.254545 -11.2816 independent 0.970588 -11.2818 (when 0.8125 -11.2818 morpho-syntactic 0.970588 -11.282 i2b2 0.878049 -11.2822 0.52 0.516667 -11.2822 calculation 0.944444 -11.2822 entered 0.860465 -11.2822 period, 0.970588 -11.2823 2,000 0.970588 -11.2827 deleting 0.475177 -11.2827 school of 0.860465 -11.2827 targeting 0.741379 -11.2828 words). 0.970588 -11.2829 reflexive 0.686567 -11.2831 training sets 0.944444 -11.2832 novels 0.69697 -11.2832 part-ofspeech 0.644737 -11.2833 identification. 0.521368 -11.2834 red 0.970588 -11.2835 aspect, 0.277108 -11.2835 aspect 0.844444 -11.2836 semantic roles. 0.970588 -11.284 samt 0.8125 -11.2842 g2p 0.844444 -11.2847 cross-validation. 0.595506 -11.2848 of, 0.970588 -11.2849 sdp 0.686567 -11.2851 yet, 0.970588 -11.2851 approximation, 0.970588 -11.2852 constituents, 0.970588 -11.2852 managing 0.578947 -11.2853 two sets 0.488722 -11.2854 mcdonald 0.553398 -11.2855 large amount 0.384259 -11.2859 1.0 0.632911 -11.286 edges. 0.942857 -11.2862 135 0.202564 -11.2863 developed 0.942857 -11.2868 context: 0.553398 -11.2871 ending 0.644737 -11.2873 hashing 0.8125 -11.2878 following, 1 -11.2879 readability. 1 -11.2879 sense-based 1 -11.2879 limit. 1 -11.2879 distant supervision 1 -11.2879 clause. 1 -11.2879 random. 1 -11.2879 codes. 1 -11.2879 genres. 0.270023 -11.2879 greater 0.969697 -11.288 surface realisation 0.969697 -11.2881 message board 0.860465 -11.2883 bnc 0.969697 -11.2887 tells us 0.329897 -11.2889 relationships 0.897436 -11.2889 (either 0.897436 -11.2894 (number 0.942857 -11.2894 environment, 0.918919 -11.2898 deterministically 0.485075 -11.2899 signal 0.721311 -11.2899 rel 0.301724 -11.2902 composed 0.969697 -11.2905 stochastic gradient 0.969697 -11.2905 keep track 0.439024 -11.2905 x, 0.175 -11.2906 similar to 0.826087 -11.2908 governor 0.942857 -11.2908 thinking 0.942857 -11.291 oh 0.401015 -11.2913 obtaining 0.840909 -11.2913 began 0.721311 -11.2913 6). 0.897436 -11.2915 movie review 0.411765 -11.2916 ones. 0.784314 -11.2917 tree kernels 0.942857 -11.2919 natural language. 0.918919 -11.2922 (b), 0.621951 -11.2924 subsequently 0.516949 -11.2924 leaves 0.376682 -11.2926 assessment 0.918919 -11.2928 debates 0.918919 -11.293 dimensions, 0.331034 -11.2931 60 0.163575 -11.2935 frequency 0.357724 -11.2939 definitions 0.918919 -11.294 completeness 0.942857 -11.294 ν 0.5 -11.2941 newspaper 0.703125 -11.2944 el 0.878049 -11.2945 rumor 0.969697 -11.2945 subcategorization 0.335714 -11.2946 uk, july 0.918919 -11.2951 visit 0.217262 -11.2953 → 0.468531 -11.2954 images 0.840909 -11.2955 section 3.2. 0.918919 -11.2959 defined, 0.840909 -11.2959 challenges. 0.918919 -11.296 000 0.969697 -11.2965 impacts 0.30086 -11.2965 behavior 0.692308 -11.2965 one-to-one 0.661972 -11.2967 procedure, 0.703125 -11.2968 summaries, 0.276699 -11.2969 between two 0.759259 -11.2971 mind 0.878049 -11.2972 replaces 0.969697 -11.2972 transitions, 0.969697 -11.2972 ulisse 0.969697 -11.2974 probing 0.969697 -11.2975 felt 0.48855 -11.2975 submission 0.969697 -11.2977 day. 0.969697 -11.2978 0.37 0.969697 -11.2978 poorer 0.657534 -11.2979 of course 0.969697 -11.298 beat 0.736842 -11.2982 (also 0.141497 -11.2982 often 0.969697 -11.2983 implying 0.969697 -11.2983 engage 0.969697 -11.2984 resultant 0.243446 -11.2987 he 0.942857 -11.2987 parse tree, 0.918919 -11.2988 have ben 0.728814 -11.2989 job 0.969697 -11.299 us, 0.942857 -11.299 communication, 0.942857 -11.299 go beyond 0.237077 -11.2992 discussed 0.692308 -11.2993 issues. 0.795918 -11.2995 relatively high 0.537037 -11.2996 bag of 0.354582 -11.2996 propagation 0.969697 -11.2997 5,000 0.686567 -11.3002 implications 0.448718 -11.3005 immediately 0.826087 -11.3008 unsupervised methods 0.969697 -11.3008 levenshtein 0.332168 -11.3008 evaluation, 0.969697 -11.301 substitution, 0.784314 -11.301 motivate 0.686567 -11.3011 an empty 0.969697 -11.3011 experienced 0.969697 -11.3012 accordance 0.759259 -11.3012 transitivity 0.969697 -11.3014 clte 0.969697 -11.3016 quantitatively 0.840909 -11.3019 tj 0.969697 -11.302 barack 0.969697 -11.3023 apparently 0.268349 -11.3025 until 0.614458 -11.3025 color 0.795918 -11.3026 handwritten 0.795918 -11.3028 mod 0.604651 -11.3029 creative 0.636364 -11.303 spread 0.153962 -11.303 2. 0.0958504 -11.3033 new 0.648649 -11.3035 spanish, 0.969697 -11.3035 durations 0.969697 -11.3038 diffusion 0.942857 -11.3038 used: 0.969697 -11.3039 have, 0.897436 -11.3041 variation. 0.875 -11.3041 idf 0.574468 -11.3042 pang 0.636364 -11.3043 johnson 0.33574 -11.3043 partially 0.56 -11.3043 classification task 0.969697 -11.3046 prompts 0.840909 -11.3047 linkage 0.808511 -11.3049 financial 0.969697 -11.3049 forest, 0.969697 -11.305 zhou, 0.969697 -11.3052 isa 0.38835 -11.3053 we expect 0.453947 -11.3057 done by 0.918919 -11.3059 pairs) 0.969697 -11.3061 tesla 0.942857 -11.3061 pos tagging, 0.969697 -11.3062 foster 0.969697 -11.3062 (1997) 0.212034 -11.3062 i.e. 0.969697 -11.3063 answering, 0.918919 -11.3064 integral 0.139913 -11.3064 labeled 0.969697 -11.3065 removed, 0.969697 -11.3067 lose 0.969697 -11.3067 instrument 0.969697 -11.3068 99% 0.248031 -11.307 protein 0.969697 -11.3072 be, 0.969697 -11.3074 subsumes 0.918919 -11.3075 hypothetical 0.795918 -11.3075 full-text 0.969697 -11.3075 devised 0.636364 -11.3075 tagged with 0.969697 -11.3079 timeml 0.840909 -11.3079 substituted 0.875 -11.3079 higher. 0.969697 -11.3082 recurring 0.969697 -11.3083 analyses, 0.969697 -11.3084 violate 0.728814 -11.3085 cohn 0.698413 -11.3085 sr 0.728814 -11.3086 participants were 0.661972 -11.3087 issue, 0.969697 -11.3089 refinement 0.496063 -11.309 since they 0.64 -11.309 to expand 0.354839 -11.309 test data 0.969697 -11.3092 coordinated 0.64 -11.3094 instantiated 0.795918 -11.3095 wants to 0.969697 -11.3096 3% 0.969697 -11.3097 almost always 0.969697 -11.3099 tie 0.969697 -11.3099 precisions 0.894737 -11.3099 exist. 0.942857 -11.3099 first 0.759259 -11.31 linguistic knowledge 0.969697 -11.3101 products, 0.969697 -11.3101 isn’t 0.857143 -11.3102 perspectives 0.795918 -11.3102 baselines, 0.969697 -11.3102 english-spanish 0.840909 -11.3102 (lin, 0.969697 -11.3103 headwords 0.166964 -11.3104 measures 0.840909 -11.3105 conll shared 0.969697 -11.3106 forums 0.918919 -11.3107 most salient 0.840909 -11.3115 nulli 0.894737 -11.3115 derivation, 0.969697 -11.3117 strtm 0.969697 -11.3117 mwes, 0.133106 -11.3117 error 0.969697 -11.3118 approximating 0.431138 -11.3118 deletion 0.272727 -11.3121 benefit 0.698413 -11.3122 bounded 0.353414 -11.3123 ibm 0.412088 -11.3123 determined by 0.875 -11.3123 inside-outside 0.840909 -11.3124 informativeness 0.453333 -11.3127 satisfaction 0.969697 -11.313 erroneously 0.969697 -11.3131 tei 0.808511 -11.3132 ij 0.875 -11.3132 word segmentation, 0.969697 -11.3134 equality 0.969697 -11.3134 arranged 0.336996 -11.3135 nist 0.458904 -11.3135 show how 0.857143 -11.3137 vas 0.840909 -11.3137 test sets, 0.542857 -11.3138 it contains 0.969697 -11.314 resemble 0.241573 -11.3141 combining 0.894737 -11.3143 walks 0.857143 -11.3148 aside 0.969697 -11.3148 rounds 0.590909 -11.3152 x2 0.270588 -11.3155 time, 0.969697 -11.3156 sub-sentential 0.716667 -11.3157 non-compositional 0.6 -11.3158 cut 0.942857 -11.3159 both sides 0.969697 -11.316 reject 0.590909 -11.316 linguistic features 0.969697 -11.3161 minimization 0.408602 -11.3162 contributions 0.628205 -11.3166 1996; 0.969697 -11.3167 bi-grams 0.875 -11.3167 impressions 0.969697 -11.3171 accepts 0.857143 -11.3175 best, 0.78 -11.3176 knight 0.941176 -11.3178 advantages over 0.941176 -11.3179 late 0.78 -11.3182 etc.), 0.561224 -11.3184 modular 0.666667 -11.3185 amazon 0.969697 -11.3186 game, 0.969697 -11.3186 loss functions 0.53211 -11.3189 ben 0.916667 -11.3189 discourse connective 0.527273 -11.319 paragraph 0.941176 -11.3194 transliterated 1 -11.3196 similarly. 1 -11.3196 review. 1 -11.3196 property. 1 -11.3196 arcs. 1 -11.3196 kernel. 1 -11.3196 support. 1 -11.3196 100. 1 -11.3196 media. 0.822222 -11.3196 syntactic categories 0.822222 -11.3196 three main 0.96875 -11.3197 prec. 0.309598 -11.3198 to represent 0.941176 -11.3199 175 0.96875 -11.32 directed acyclic 0.681818 -11.3206 modifying 0.292479 -11.3207 easy 0.301775 -11.3207 into account 0.941176 -11.3207 hmm, 0.53211 -11.321 identified by 0.628205 -11.3212 0.85 0.237226 -11.3212 work on 0.941176 -11.3215 as mentioned above, 0.941176 -11.3217 108 0.372197 -11.3224 and therefore 0.192077 -11.3226 consists of 0.681818 -11.3227 generator 0.300885 -11.3227 tasks, 0.492063 -11.3227 topic model 0.875 -11.3229 phenotype 0.716667 -11.3229 samples. 0.393939 -11.323 vertices 0.745455 -11.3232 0.80 0.941176 -11.3232 river 0.808511 -11.3235 intended message 0.96875 -11.3236 counterparts 0.96875 -11.3241 vision 0.754717 -11.3244 bidirectional 0.96875 -11.3247 cluster-based 0.96875 -11.3247 q1 0.745455 -11.3248 top k 0.941176 -11.3248 response, 0.132345 -11.3251 positive 0.875 -11.3254 restrictions on 0.916667 -11.3255 tutorial dialogue 0.916667 -11.3256 zhai, 0.78 -11.3257 steps, 0.148043 -11.3258 another 0.130243 -11.3259 precision 0.916667 -11.3261 sized 0.334545 -11.3263 t. 0.822222 -11.3264 investigates 0.808511 -11.3265 author. 0.894737 -11.3268 below). 0.941176 -11.3272 115 0.96875 -11.3272 negation, 0.96875 -11.3273 evoke 0.681818 -11.3275 avoids 0.4375 -11.3277 we generate 0.382775 -11.3277 narrative 0.916667 -11.3278 recovering 0.96875 -11.328 pursue 0.754717 -11.3281 stemmed 0.837209 -11.3282 dominates 0.96875 -11.3283 hand-coded 0.504202 -11.3285 activation 0.916667 -11.3288 occurs when 0.620253 -11.3291 are more likely 0.396907 -11.3292 again, 0.941176 -11.3292 agglomerative clustering 0.543689 -11.3294 interval 0.916667 -11.3295 tended to 0.96875 -11.3295 entailing 0.643836 -11.3296 behavior. 0.657143 -11.3297 an article 0.609756 -11.3301 tokenized 0.96875 -11.3305 lenient 0.18133 -11.3306 resources 0.96875 -11.3306 roots 0.853659 -11.3309 irish 0.96875 -11.3312 polarity, 0.96875 -11.3316 while maintaining 0.96875 -11.3316 cores 0.745455 -11.3318 gigaword corpus 0.96875 -11.3319 12% 0.620253 -11.332 negative examples 0.732143 -11.3321 discourse. 0.60241 -11.3324 we conduct 0.96875 -11.3326 computers 0.96875 -11.3327 follower 0.916667 -11.3328 instantiating 0.96875 -11.3331 positive/negative 0.941176 -11.3332 lcfrs 0.402116 -11.3334 berkeley 0.96875 -11.3334 cut-off 0.96875 -11.3335 disputants 0.5 -11.3338 dialogue, 0.170152 -11.3339 above 0.96875 -11.3342 arrows 0.96875 -11.3342 modelled 0.6875 -11.3343 car 0.96875 -11.3343 muc-4 0.96875 -11.3344 actively 0.593023 -11.3347 rewrite 0.96875 -11.3354 interaction, 0.556701 -11.3355 incrementally 0.671642 -11.3355 present, 0.711864 -11.3357 repetition 0.96875 -11.3357 121 0.96875 -11.3359 (1995) 0.543689 -11.336 sentence length 0.360515 -11.3362 stop 0.96875 -11.3363 feedback, 0.96875 -11.3365 “good” 0.367713 -11.3366 reports 0.96875 -11.3366 other things 0.837209 -11.3367 f-measure. 0.871795 -11.3367 bi-gram 0.822222 -11.3367 blood 0.837209 -11.3368 lexical units 0.916667 -11.3368 counterpart 0.941176 -11.3368 113 0.154818 -11.3369 labels 0.791667 -11.3369 save 0.96875 -11.3369 pair-wise 0.804348 -11.337 currently, 0.96875 -11.3371 ment 0.78 -11.3373 huang, 0.222951 -11.3373 cost 0.96875 -11.3374 spanning tree 0.837209 -11.3375 trust 0.791667 -11.3377 causing 0.853659 -11.3377 inspection of 0.273399 -11.3378 (2006) 0.577778 -11.3379 marked with 0.754717 -11.338 15% 0.290503 -11.338 indicates that 0.96875 -11.3382 named-entity 0.470588 -11.3383 which contains 0.724138 -11.3385 johnson, 0.724138 -11.3386 semisupervised 0.58427 -11.3388 10. 0.504202 -11.3389 on top 0.96875 -11.3389 signed 0.623377 -11.3391 1.00 0.395833 -11.3392 plus 0.96875 -11.3394 0.16 0.96875 -11.3394 (positive 0.96875 -11.3395 modifiers, 0.96875 -11.34 well-defined 0.174571 -11.3401 learn 0.392857 -11.3404 70 0.754717 -11.3407 isr 0.395833 -11.3409 off 0.390863 -11.3411 respectively 0.96875 -11.3415 stops 0.96875 -11.3417 separates 0.96875 -11.3419 enrichment 0.916667 -11.342 alternating 0.853659 -11.3421 styles 0.96875 -11.3421 non-null 0.96875 -11.3422 vague 0.232558 -11.3425 optimal 0.449664 -11.3426 embedded 0.740741 -11.3426 (smith 0.96875 -11.3426 takes advantage 0.96875 -11.3427 0.5, 0.96875 -11.3429 (2001), 0.96875 -11.343 (oov) 0.661765 -11.3432 same way as 0.764706 -11.3433 streaming 0.55102 -11.3434 our system. 0.6125 -11.3435 tf-idf 0.96875 -11.3437 descending 0.693548 -11.3439 fold 0.257143 -11.344 property 0.96875 -11.3442 execute 0.711864 -11.3443 traditionally 0.287671 -11.3445 feedback 0.657143 -11.3446 intervals 0.754717 -11.3447 lex 0.96875 -11.3447 clause, 0.569892 -11.3448 sa 0.473684 -11.3448 perspective 0.96875 -11.3449 bbn 0.871795 -11.345 nt 0.96875 -11.345 scorer 0.272953 -11.3451 markov 0.891892 -11.3451 spans. 0.635135 -11.3452 0.86 0.837209 -11.3453 transcribers 0.429448 -11.3457 throughout 0.871795 -11.3457 scaled 0.277635 -11.3457 needed 0.504274 -11.3459 2nd 0.96875 -11.3464 bags of 0.732143 -11.3465 establishing 0.764706 -11.3468 ’s 0.871795 -11.3469 s: 0.647887 -11.3469 clearly, 0.161739 -11.3471 e 0.60241 -11.3472 decreasing 0.96875 -11.3473 turku 0.818182 -11.3479 an earlier 0.914286 -11.3481 very little 0.719298 -11.3484 intuitions 0.615385 -11.3485 j., 0.96875 -11.3485 103 0.56383 -11.3486 agenda 0.853659 -11.349 105 0.711864 -11.3495 0.4 0.6 0.246939 -11.3496 (2008) 0.732143 -11.3502 persons 0.96875 -11.3502 corporate 0.586207 -11.3504 center for 0.6125 -11.3505 strategy, 0.495868 -11.3505 incorporates 0.96875 -11.3506 experimental setting 0.96875 -11.3509 files, 0.96875 -11.3512 (1994) 0.115728 -11.3512 parsing 0.293948 -11.3513 for a given 0.272953 -11.3513 collection of 0.853659 -11.3513 google n-gram 0.818182 -11.3514 engineering, 0.914286 -11.3518 xij 0.711864 -11.3519 we have developed 0.6125 -11.3519 correlates 0.853659 -11.3522 induction, 0.136534 -11.3523 , 1 -11.3524 discussions. 1 -11.3524 choices. 1 -11.3524 individually. 1 -11.3524 (b). 1 -11.3524 researchers. 1 -11.3524 overall. 1 -11.3524 univ. 1 -11.3524 discussed. 1 -11.3524 coherence. 1 -11.3524 ours. 1 -11.3524 judges. 1 -11.3524 acquisition. 1 -11.3524 pronoun. 1 -11.3524 however. 0.939394 -11.3524 emphasized 0.891892 -11.3529 exact match 0.939394 -11.3529 mrr 0.967742 -11.3531 indistinguishable 0.175258 -11.3539 lower 0.292264 -11.3541 computation 0.467153 -11.3542 4). 0.579545 -11.3546 forced 0.764706 -11.3548 papers, 0.939394 -11.3549 f1-score 0.693548 -11.3552 ψ 0.626667 -11.3552 theories 0.914286 -11.3555 invited 0.871795 -11.3558 discriminative learning 0.853659 -11.3558 ∆ 0.914286 -11.3559 comparatively 0.853659 -11.3561 ancestor 0.420118 -11.3562 members 0.804348 -11.3562 bacterium 0.939394 -11.3565 bn 0.27551 -11.3565 variants 0.243513 -11.3567 representing 0.967742 -11.3568 aso 0.491803 -11.3571 segmentation, 0.421687 -11.3571 seem to 0.405556 -11.3571 interpolation 0.440789 -11.3574 learned from 0.518182 -11.3574 = argmax 0.428571 -11.3577 augmented 0.143152 -11.3577 pages 0.7 -11.3579 1995) 0.967742 -11.3579 best-first 0.638889 -11.3581 low, 0.77551 -11.3581 stages. 0.833333 -11.3582 pre 0.967742 -11.3584 this purpose. 0.345382 -11.3584 section 4 0.199198 -11.3584 identifying 0.48062 -11.3585 an improvement 0.638889 -11.3585 inference, 0.351464 -11.3585 forward 0.967742 -11.3586 reuse 0.871795 -11.3586 older 0.647887 -11.3586 clustering algorithm 0.939394 -11.3589 extraneous 0.967742 -11.3591 enforced 0.253247 -11.3591 ones 0.311475 -11.3593 it can be 0.914286 -11.3594 news articles. 0.967742 -11.3596 token-based 0.939394 -11.3596 mbr system combination 0.77551 -11.3596 paraphrased 0.939394 -11.3597 (even 0.55102 -11.36 pathway 0.366516 -11.3601 style 0.303406 -11.3602 boundary 0.588235 -11.3605 food 0.266187 -11.3608 roles 0.939394 -11.361 text: 0.871795 -11.3611 compound splitting 0.914286 -11.3614 age, 0.378641 -11.3619 a fixed 0.939394 -11.3621 (note that 0.75 -11.3622 related work. 0.939394 -11.3623 two steps: 0.597561 -11.3626 directions. 0.397849 -11.3628 european 0.967742 -11.3629 sentences), 0.727273 -11.3633 by performing 0.40678 -11.3635 verbs, 0.309446 -11.3638 analyze 0.411429 -11.3638 there was 0.967742 -11.364 nature, 0.415205 -11.3641 agree 0.727273 -11.3644 optimization problem 0.110156 -11.3645 document 0.77551 -11.3649 0.40 0.719298 -11.365 top 10 0.787234 -11.3651 wizards 0.967742 -11.3656 4.2.2 0.375 -11.366 probably 0.967742 -11.3662 methodologies 0.967742 -11.3664 1991) 0.8 -11.3665 binary classifier 0.5 -11.3665 x1 0.891892 -11.3666 “in 0.68254 -11.3666 1993), 0.787234 -11.3668 examination of 0.286908 -11.3669 co-occurrence 0.234962 -11.3671 implemented 0.128348 -11.3671 noun 0.435065 -11.3672 opposite 0.967742 -11.3672 weight, 0.967742 -11.3673 encounter 0.967742 -11.3674 domainspecific 0.652174 -11.3674 blue 0.967742 -11.3676 dynamic cache 0.939394 -11.3676 relations: 0.68254 -11.3678 2007. 0.967742 -11.3679 subject-verb 0.967742 -11.368 hand-annotated 0.967742 -11.368 coarse-to-fine 0.597561 -11.368 sentence pairs. 0.476562 -11.3682 united 0.914286 -11.3682 for, 0.967742 -11.3683 sentences; 0.164371 -11.3683 take 0.727273 -11.3684 (14) 0.967742 -11.3684 0.19 0.967742 -11.3684 0.46 0.0958307 -11.3686 three 0.967742 -11.3686 non-standard 0.85 -11.3686 s3h 0.967742 -11.3688 deduction system 0.967742 -11.3689 frames, 0.891892 -11.3689 counted as 0.967742 -11.369 lexicalization 0.76 -11.369 evaluators 0.939394 -11.3693 (16) 0.914286 -11.3695 pathways 0.818182 -11.37 both. 0.967742 -11.3701 collocational 0.85 -11.3701 detectors 0.588235 -11.3701 failed 0.967742 -11.3702 clasifier 0.967742 -11.3702 source-target 0.297872 -11.3703 training, 0.914286 -11.3703 infobox 0.833333 -11.3703 sentiment lexicon 0.967742 -11.3706 gathering 0.967742 -11.3706 33% 0.967742 -11.3707 served 0.565217 -11.3711 30% 0.787234 -11.3712 partials 0.523364 -11.3717 priors 0.476562 -11.3718 which makes 0.76 -11.3718 misspelled 0.967742 -11.3719 experimental settings 0.967742 -11.3719 indexes 0.217391 -11.372 table 1 0.967742 -11.372 2.6 0.967742 -11.3722 beter 0.967742 -11.3723 famous 0.967742 -11.3723 licensed 0.76 -11.3724 useful. 0.256757 -11.3727 detailed 0.75 -11.3728 memory. 0.967742 -11.373 preposition, 0.727273 -11.3732 word order. 0.451389 -11.3734 which uses 0.630137 -11.3735 li, 0.868421 -11.3736 interrogative 0.237911 -11.3738 sampling 0.914286 -11.3739 4.6 0.161463 -11.374 tags 0.706897 -11.3741 -1 0.85 -11.3741 left-to-right 0.967742 -11.3742 sorting 0.281081 -11.3743 4, 0.127072 -11.3745 knowledge 0.914286 -11.3749 dysarthric 0.528846 -11.3753 an argument 0.967742 -11.3755 low-frequency 0.85 -11.3755 pos-tags 0.76 -11.3758 due to space 0.967742 -11.376 descendant 0.967742 -11.3762 seemingly 0.234082 -11.3763 consistent 0.914286 -11.3763 zhao et al. 0.5 -11.3764 maintain 0.967742 -11.3765 approaches: 0.656716 -11.3765 frontier 0.967742 -11.3766 architectural 0.967742 -11.3767 recruited 0.261792 -11.3768 relations. 0.967742 -11.3769 126 0.75 -11.3773 low-level 0.967742 -11.3774 complex, 0.85 -11.3777 hotel 0.459854 -11.3777 output, 0.398907 -11.3778 series of 0.388601 -11.3779 compares 0.22 -11.378 confidence 0.27957 -11.378 ordering 0.967742 -11.3781 artificially 0.76 -11.3782 fr 0.967742 -11.3783 contrasts 0.967742 -11.3783 wmt10 0.888889 -11.3784 constraints: 0.85 -11.3786 dependency graphs 0.194301 -11.3787 objective 0.714286 -11.3788 purity 0.967742 -11.379 chunks, 0.323741 -11.3791 upper 0.787234 -11.3793 0.73 0.770833 -11.3795 hurt 0.656716 -11.3796 distance, 0.888889 -11.3797 line graphs 0.967742 -11.3797 restrictive 0.37561 -11.3797 decomposition 0.888889 -11.38 yes/no 0.36 -11.3805 voice 0.967742 -11.3805 clitics 0.437086 -11.3807 projects 0.706897 -11.3809 we call this 0.813953 -11.3811 opt 0.437086 -11.3811 dialect 0.967742 -11.3818 driving 0.868421 -11.3819 revisions 0.939394 -11.3821 (fig. 0.727273 -11.3822 raises 0.677419 -11.3825 for testing. 0.44 -11.3825 assumes 0.727273 -11.3828 alternate 0.770833 -11.3828 rhyme 0.671875 -11.3829 user, 0.967742 -11.3832 folowing 0.351931 -11.3832 2.3 0.85 -11.384 spent 0.694915 -11.3842 votes 0.9375 -11.3845 unweighted 0.888889 -11.3846 aj 0.6 -11.3847 treatment of 0.967742 -11.3848 year, 0.735849 -11.3848 algebra 0.967742 -11.385 parliament 0.9375 -11.385 trade-off between 0.677419 -11.3851 syntax, 0.447552 -11.3852 (2002) 0.323636 -11.3854 symbol 0.9375 -11.386 doing so 0.186298 -11.3861 differences 0.9375 -11.3861 regulatory 1 -11.3863 whole. 1 -11.3863 consideration. 1 -11.3863 act. 1 -11.3863 turk. 1 -11.3863 configuration. 1 -11.3863 gains. 1 -11.3863 learner. 1 -11.3863 transition. 1 -11.3863 exp. 1 -11.3863 lattices. 1 -11.3863 genre. 1 -11.3863 challenging. 0.770833 -11.3866 optimizes 0.966667 -11.3866 reading comprehension 0.85 -11.3866 phonemic 0.9375 -11.3866 136 0.9375 -11.3867 baseball 0.442177 -11.387 error analysis 0.647059 -11.3872 overlaps 0.911765 -11.3873 greedy algorithm 0.888889 -11.3874 104 0.647059 -11.3874 james 0.273438 -11.3875 share 0.472868 -11.3875 health 0.466165 -11.3881 prevent 0.813953 -11.3888 tutors 0.0942408 -11.389 dependency 0.782609 -11.3891 we are currently 0.9375 -11.3894 ftb 0.813953 -11.3895 palmer, 0.694915 -11.3897 attempted to 0.829268 -11.39 mihalcea, 0.9375 -11.3901 och, 0.9375 -11.3902 topmost 0.966667 -11.3903 hidden states 0.888889 -11.391 downloaded from 0.813953 -11.3913 by analyzing 0.966667 -11.3918 genuine 0.125548 -11.3918 specific 0.9375 -11.3918 dummy 0.966667 -11.392 1995), 0.36036 -11.3923 dual 0.770833 -11.3924 william 0.513761 -11.3927 french, 0.535354 -11.3927 sense, 0.829268 -11.3929 initiative 0.829268 -11.3934 posterior distribution 0.813953 -11.3935 backchannels 0.966667 -11.3939 randomly chosen 0.495726 -11.3939 reviews. 0.745098 -11.3941 execution 0.677419 -11.3942 mapping between 0.966667 -11.3942 practically 0.966667 -11.3944 wmt 2010 0.735849 -11.3944 moses (koehn 0.966667 -11.3946 replicate 0.813953 -11.3947 lang 0.966667 -11.3947 bearing 0.284507 -11.3948 estimation 0.966667 -11.3948 scope, 0.888889 -11.395 ambiguity. 0.735849 -11.3952 judgement 0.198895 -11.3954 already 0.735849 -11.3957 pipeline. 0.966667 -11.3958 mwe detection 0.309764 -11.3959 (with 0.217463 -11.396 part of the 0.911765 -11.3961 constraining 0.345992 -11.3961 leading 0.966667 -11.3963 research council 0.9375 -11.3963 internal nodes 0.846154 -11.3966 pattern matching 0.966667 -11.3966 machine learning, 0.61039 -11.3967 when applied 0.206231 -11.3968 properties 0.222031 -11.3969 improves 0.966667 -11.3973 jobs 0.198347 -11.3973 match 0.966667 -11.3974 /b/ 0.387435 -11.3976 σ 0.966667 -11.3977 mode, 0.966667 -11.3978 necessary, 0.541667 -11.3978 north 0.966667 -11.398 coreference, 0.782609 -11.398 challenge. 0.194702 -11.3981 expressions 0.123012 -11.3981 entity 0.20317 -11.3981 computer 0.410714 -11.3982 meaning. 0.966667 -11.3982 ultimately, 0.966667 -11.3983 perfectly 0.966667 -11.3986 pitler 0.244259 -11.3986 solution 0.846154 -11.3987 projects agency 0.966667 -11.3987 6% 0.472441 -11.3988 iterations. 0.966667 -11.3993 timelines 0.966667 -11.3994 derivations, 0.770833 -11.3995 5-gram language 0.966667 -11.3996 groupings 0.745098 -11.3996 totally 0.911765 -11.3999 neural network 0.966667 -11.3999 friend 0.431373 -11.4002 captured 0.966667 -11.4004 learners’ 0.9375 -11.4005 information status 0.722222 -11.4006 an item 0.694915 -11.4007 ctb 0.966667 -11.4007 fluency, 0.585366 -11.4008 trie 0.414634 -11.4009 effects of 0.966667 -11.4009 pitch-accent 0.421384 -11.4009 feature, 0.651515 -11.401 disambiguated 0.53 -11.401 boolean 0.911765 -11.4011 finished 0.966667 -11.4011 coreferring 0.966667 -11.4011 compensate for 0.535354 -11.4011 unless 0.592593 -11.4011 satisfies 0.666667 -11.4012 tree structure 0.46875 -11.4012 d, 0.966667 -11.4012 resolver 0.683333 -11.4013 making use 0.504505 -11.4013 semeval 0.911765 -11.4015 raise 0.911765 -11.4018 genre, 0.966667 -11.4018 remediation 0.966667 -11.4019 standardized 0.911765 -11.4021 nnlm 0.483607 -11.4021 an automatic 0.966667 -11.4022 volunteers 0.966667 -11.4022 pronounced 0.966667 -11.4022 intersections 0.633803 -11.4023 cc 0.561798 -11.4025 (to 0.966667 -11.4026 finer 0.966667 -11.4026 single, 0.782609 -11.4027 clusterings 0.864865 -11.4028 disagreements 0.966667 -11.4029 added, 0.290801 -11.4029 preliminary 0.966667 -11.403 phrase-pair 0.204713 -11.4033 predicate 0.245763 -11.4033 combine 0.829268 -11.4033 believed 0.315789 -11.4035 precision, 0.966667 -11.4035 prepositional phrase 0.966667 -11.4035 2.8 0.554348 -11.4038 divided by 0.666667 -11.4042 word clusters 0.966667 -11.4044 completing 0.966667 -11.4045 contributing 0.307692 -11.4047 image 0.279778 -11.4048 otherwise 0.966667 -11.4048 index, 0.966667 -11.4048 hub 0.966667 -11.4048 dag 0.651515 -11.4049 recognition, 0.966667 -11.4049 speed, 0.966667 -11.4049 limitations, 0.966667 -11.4049 from, 0.966667 -11.4051 references, 0.53 -11.4052 identified as 0.795455 -11.4052 fbis 0.966667 -11.4053 social networks 0.966667 -11.4054 svm-based 0.966667 -11.406 141 0.966667 -11.406 1.4 0.966667 -11.406 computation, 0.966667 -11.4061 formulations 0.267003 -11.4062 it is not 0.911765 -11.4063 mood 0.911765 -11.4063 robots 0.102252 -11.4063 following 0.864865 -11.4064 p., 0.966667 -11.4066 alongside 0.966667 -11.4068 weakly-supervised 0.651515 -11.4071 holding 0.283286 -11.4071 view 0.966667 -11.4071 t/v 0.966667 -11.4072 antonyms 0.162037 -11.4075 joint 0.683333 -11.4076 (collins 0.864865 -11.4076 g2 0.683333 -11.4077 stateof-the-art 0.966667 -11.4077 ac 0.438356 -11.4078 reader 0.911765 -11.4079 numbered 0.0665209 -11.408 their 0.554348 -11.4081 to adapt 0.966667 -11.4081 filtered out 0.966667 -11.4082 image-based 0.911765 -11.4083 finnish 0.966667 -11.4083 superset 0.277174 -11.4083 providing 0.368932 -11.4083 engineering 0.966667 -11.4085 summarizers 0.966667 -11.4085 authored 0.966667 -11.4086 avoided 0.755102 -11.4086 retraining 0.396648 -11.4086 manning, 0.165377 -11.4087 linear 0.966667 -11.4087 excellence 0.414634 -11.4089 formula 0.966667 -11.4089 underlined 0.966667 -11.4089 exists, 0.966667 -11.4091 judging 0.637681 -11.4091 lemmatized 0.199438 -11.4092 capture 0.966667 -11.4093 lasso 0.241237 -11.4094 easily 0.966667 -11.4094 grammar-based 0.362791 -11.4095 annotations. 0.139889 -11.4095 process 0.651515 -11.4097 vocabulary, 0.594937 -11.4097 attention to 0.120143 -11.4098 annotated 0.966667 -11.4101 relax 0.193122 -11.4103 j 0.966667 -11.4104 independent, 0.966667 -11.4105 word) 0.966667 -11.4106 142 0.576471 -11.4106 (at 0.795455 -11.4106 p(e) 0.346154 -11.4107 official 0.966667 -11.4107 contextually 0.219554 -11.4107 grammatical 0.911765 -11.4109 ended 0.555556 -11.411 (not 0.722222 -11.411 k1 0.0834359 -11.411 for each 0.829268 -11.4112 won 0.613333 -11.4113 (liu et 0.885714 -11.4115 2004, 0.885714 -11.4115 von 0.911765 -11.4115 table 7 shows 0.443662 -11.4117 400 0.381443 -11.4117 year 0.966667 -11.4118 scalar 0.163654 -11.4124 tasks 0.966667 -11.4131 preterminal 0.625 -11.4142 relation types 0.829268 -11.4142 125 0.651515 -11.4144 it is interesting 0.328185 -11.4145 since it 0.966667 -11.4145 disfluencies 0.829268 -11.4145 nlp. 0.966667 -11.4146 1500 0.966667 -11.4146 handcrafted 0.204478 -11.4147 variables 0.200573 -11.415 unlabeled 0.966667 -11.4151 emergency 0.966667 -11.4152 fisher 0.755102 -11.4152 very simple 0.9375 -11.4152 post, 0.795455 -11.4152 loose 0.966667 -11.4154 language: 0.966667 -11.4155 d) 0.909091 -11.4155 mg 0.809524 -11.4156 (s) 0.423077 -11.4158 recent work 0.221254 -11.4158 application 0.966667 -11.416 im 0.966667 -11.4161 clinton 0.966667 -11.4161 area, 0.966667 -11.4164 quantifying 0.561798 -11.4167 varied 0.128462 -11.4168 candidate 0.212903 -11.4169 ( 0.795455 -11.4171 children’s 0.966667 -11.4172 videos 0.966667 -11.4172 mother 0.795455 -11.4172 single-word 0.364929 -11.4173 recall, 0.809524 -11.4173 visited 0.701754 -11.4174 genetic 0.487179 -11.4174 precise 0.966667 -11.4184 issued 0.966667 -11.4184 129 0.427632 -11.4187 request 0.651515 -11.4188 callison-burch 0.809524 -11.4188 headed 0.966667 -11.4194 isolation, 0.125922 -11.4195 user 0.651515 -11.4198 keeps 0.966667 -11.4198 155 0.173118 -11.42 metric 0.935484 -11.4201 t: 0.885714 -11.4202 (17) 0.166667 -11.4203 annotations 0.935484 -11.4205 (4), 0.885714 -11.4206 slope 0.825 -11.4206 coded 0.672131 -11.4207 a comprehensive 0.555556 -11.421 p1 0.864865 -11.4212 paths. 0.935484 -11.4212 parameter tuning 1 -11.4214 help. 1 -11.4214 fixed. 1 -11.4214 comparable. 1 -11.4214 generalization. 1 -11.4214 4.1. 1 -11.4214 place. 1 -11.4214 seconds. 1 -11.4214 index. 1 -11.4214 policies. 1 -11.4214 hong kong 1 -11.4214 λ. 1 -11.4214 accurate. 1 -11.4214 experience. 1 -11.4214 preprocessing. 1 -11.4214 (resp. 1 -11.4214 be. 0.965517 -11.4215 pruning. 0.935484 -11.4215 p-values 0.935484 -11.4217 wiktionary 0.935484 -11.4218 built-in 0.909091 -11.4218 into account, 0.935484 -11.4222 sieve 0.152941 -11.4222 translations 0.0832507 -11.4222 method 0.730769 -11.4226 inter 0.276712 -11.4228 z 0.909091 -11.4229 context free 0.965517 -11.423 (turney 0.409639 -11.4233 to combine 0.864865 -11.4233 first attempt 0.324528 -11.4233 parsed 0.218058 -11.4233 designed 0.935484 -11.4235 4.7 0.777778 -11.4236 signature module 0.440559 -11.4239 most important 0.965517 -11.424 hmm. 0.641791 -11.4241 average precision 0.138167 -11.4245 length 0.65625 -11.4246 non-english 0.965517 -11.4247 arrive at 0.909091 -11.4249 frequently occurring 0.186483 -11.4249 includes 0.199426 -11.425 4. 0.965517 -11.4252 flag 0.555556 -11.4253 debate 0.563218 -11.4261 pyramid 0.571429 -11.4261 found. 0.244589 -11.4262 actually 0.965517 -11.4264 sorts 0.885714 -11.4266 plan to investigate 0.709091 -11.427 slightly different 0.825 -11.4271 marginally 0.864865 -11.4274 wk 0.965517 -11.4277 split-merge 0.909091 -11.4277 most informative 0.909091 -11.4279 graphical representation 0.965517 -11.4281 (cohn 0.935484 -11.4282 computed, 0.777778 -11.4283 slu 0.211974 -11.4284 address 0.864865 -11.4285 +1 0.965517 -11.4285 section 3.3. 0.965517 -11.429 titles, 0.965517 -11.429 reconstruction error 0.965517 -11.429 49.4 0.965517 -11.4292 too, 0.909091 -11.4295 0.13 0.14121 -11.4296 graph 0.677966 -11.4297 priori 0.641791 -11.4297 0.78 0.765957 -11.4298 0.56 0.842105 -11.4299 (english 0.935484 -11.4299 tp 0.965517 -11.43 nugues, 0.809524 -11.43 detected by 0.909091 -11.4302 reliance 0.965517 -11.4303 assumptions, 0.864865 -11.4304 vms 0.965517 -11.4306 dominating 0.965517 -11.4307 bias, 0.74 -11.4308 that maximizes 0.965517 -11.431 lexico-syntactic patterns 0.75 -11.4311 umls 0.509434 -11.4318 an interesting 0.486957 -11.4319 thread 0.909091 -11.432 random sampling 0.38587 -11.4321 title 0.909091 -11.4322 par with 0.483051 -11.4322 jj 0.396552 -11.4324 example. 0.965517 -11.4328 plans 0.809524 -11.4328 it is hard 0.965517 -11.4331 dop 0.935484 -11.4332 scl 0.765957 -11.4332 χ2 0.965517 -11.4335 chinese-to-english 0.809524 -11.4336 two versions of 0.842105 -11.434 cohen’s 0.689655 -11.434 short, 0.935484 -11.4341 document’s 0.809524 -11.4342 charniak, 0.965517 -11.4346 goldstandard 0.483051 -11.4346 yes 0.909091 -11.4347 rejection 0.157516 -11.4348 dataset 0.965517 -11.4349 decline 0.20208 -11.4349 usually 0.909091 -11.4349 chicago 0.696429 -11.4351 headlines 0.245077 -11.4352 language, 0.74 -11.4356 feature set, 0.965517 -11.4359 wu, 0.406061 -11.4362 convergence 0.65625 -11.4362 thre 0.53125 -11.4362 k, 0.182588 -11.4363 np 0.965517 -11.4367 phenomenon, 0.74 -11.4367 prosodic features 0.716981 -11.4368 compounds, 0.790698 -11.4368 featured 0.458015 -11.4369 p = 0.965517 -11.437 29% 0.965517 -11.4371 last, 0.965517 -11.4372 genpex 0.965517 -11.4373 partof 0.965517 -11.4373 style, 0.672131 -11.4373 dimensions. 0.935484 -11.4375 section 4 presents 0.935484 -11.4375 spoken dialog 0.394286 -11.4376 regulation 0.716981 -11.4377 hypotheses, 0.965517 -11.4378 pronoun, 0.130086 -11.4379 t 0.965517 -11.438 burden 0.74 -11.4381 wiebe, 0.861111 -11.4382 hierarchical clustering 0.632353 -11.4383 zhu 0.965517 -11.4384 acting 0.909091 -11.4385 u, 0.550562 -11.4385 bold 0.549451 -11.4387 requests 0.909091 -11.4388 goodman 0.478992 -11.4389 survival 0.965517 -11.4389 arrays 0.842105 -11.439 more precise 0.965517 -11.4391 california, 0.233533 -11.4392 are also 0.75 -11.4393 ed 0.965517 -11.4394 exemplar 0.965517 -11.4395 immortal 0.965517 -11.4395 target-language 0.482759 -11.4395 hit 0.885714 -11.4396 act, 0.965517 -11.4398 narratives 0.72549 -11.4401 text-based 0.965517 -11.4402 efficient, 0.608108 -11.4402 retained 0.5 -11.4405 prominent 0.550562 -11.4407 times, 0.885714 -11.4409 con 0.965517 -11.4409 both, 0.965517 -11.441 lscript1 0.965517 -11.4418 cache-based 0.72549 -11.4418 suspect that 0.965517 -11.4419 soccer 0.965517 -11.442 entropy, 0.0492646 -11.4421 using 0.965517 -11.4421 qualities 0.909091 -11.4421 more formally, 0.573171 -11.4422 25% 0.66129 -11.4423 abstraction 0.825 -11.4423 pivoting 0.619718 -11.4424 token, 0.354839 -11.4424 meanings 0.965517 -11.4425 (jiang 0.965517 -11.4426 culture 0.935484 -11.4426 parse disambiguation 0.965517 -11.4427 sought 0.777778 -11.4427 proximity 0.52 -11.4427 fifth 0.965517 -11.443 cohen, 0.965517 -11.443 dirichlet prior 0.965517 -11.443 unlabelled 0.177829 -11.4432 point 0.965517 -11.4436 severely 0.965517 -11.4436 accordingly 0.825 -11.4439 fan-out 0.4125 -11.4439 2000) 0.965517 -11.444 identifiable 0.965517 -11.4442 20,000 0.965517 -11.4443 whilst 0.471545 -11.4443 patient 0.677966 -11.4444 important. 0.842105 -11.4445 peak 0.611111 -11.4446 we show how 0.965517 -11.4446 decompositions 0.804878 -11.4447 very close 0.965517 -11.4447 parts-of-speech 0.965517 -11.4447 qualitative analysis 0.909091 -11.4447 kim, 0.66129 -11.4452 elaborate 0.965517 -11.4452 romanian 0.75 -11.4454 consequently 0.965517 -11.4455 multi-source 0.6 -11.4455 utterance, 0.564706 -11.4456 ongoing 0.72549 -11.4457 one’s 0.271003 -11.4459 determined 0.965517 -11.446 involved, 0.965517 -11.4461 understanding, 0.861111 -11.4461 capitalized 0.156334 -11.4463 opinion 0.965517 -11.4463 drive 0.965517 -11.4463 own, 0.75 -11.4465 conceptually 0.965517 -11.4466 correctly, 0.174107 -11.4466 reported 0.842105 -11.4468 graphic 0.72549 -11.4471 price 0.132515 -11.4471 measure 0.909091 -11.4472 randomly select 0.290123 -11.4472 two different 0.134487 -11.4473 grammar 0.790698 -11.4473 ss 0.301003 -11.4474 remove 0.75 -11.4477 zi 0.636364 -11.4479 sentiment. 0.790698 -11.448 reviewed 0.935484 -11.4483 117 0.777778 -11.4486 barzilay, 0.965517 -11.4488 prlm 0.684211 -11.4492 differences. 0.935484 -11.4495 wordnet synsets 0.4125 -11.4495 details of 0.261307 -11.4497 tools 0.650794 -11.4499 cj 0.804878 -11.45 (1996) 0.965517 -11.45 turker 0.965517 -11.45 4.0 0.75 -11.4501 likewise, 0.76087 -11.4502 george 0.0932991 -11.451 do 0.53125 -11.451 government 0.804878 -11.4511 we discard 0.965517 -11.4511 military 0.321839 -11.4511 game 0.197406 -11.4511 adding 0.514851 -11.4512 correct, 0.965517 -11.4513 (yamada 0.790698 -11.4513 once, 0.218695 -11.4517 direct 0.965517 -11.4517 standalone 0.235656 -11.4519 shared task 0.428571 -11.4519 to express 0.623188 -11.4519 plots 0.965517 -11.452 deviance 0.965517 -11.4521 processes, 0.72549 -11.4523 0) 0.965517 -11.4524 render 0.965517 -11.4524 win 0.395349 -11.4525 vertex 0.684211 -11.4525 knn 0.965517 -11.4525 lets 0.965517 -11.4528 c-value 0.935484 -11.453 moses. 0.790698 -11.4532 most effective 0.965517 -11.4536 including: 0.804878 -11.4537 maintained 0.965517 -11.4537 locative 0.90625 -11.4537 optimistic 0.804878 -11.4538 seen, 0.297386 -11.4539 translate 0.861111 -11.4539 uima 0.90625 -11.4539 demand 0.965517 -11.4545 acuracy 0.965517 -11.4549 born 0.90625 -11.4552 related work, 0.790698 -11.4553 lowbow 0.933333 -11.4554 configuration, 0.965517 -11.4557 winning 0.820513 -11.4557 tokens) 0.804878 -11.4558 deletions 0.933333 -11.4558 obtained, 0.407407 -11.4561 6 conclusion 0.804878 -11.4564 bl 0.115863 -11.4564 alignment 0.0986985 -11.4564 how 0.933333 -11.4569 p0 0.933333 -11.4572 breaking 0.467213 -11.4576 as discussed 0.861111 -11.4576 flight 0.276353 -11.4578 helpful 1 -11.4578 hits. 1 -11.4578 parent. 1 -11.4578 interpretations. 1 -11.4578 monte carlo 1 -11.4578 dynamic programming. 1 -11.4578 described. 1 -11.4578 f1. 1 -11.4578 excluded. 1 -11.4578 so. 1 -11.4578 options. 1 -11.4578 alternatives. 1 -11.4578 key. 1 -11.4578 good. 1 -11.4578 efficiently. 1 -11.4578 matches. 1 -11.4578 workers. 1 -11.4578 rates. 0.203077 -11.4578 true 0.964286 -11.4579 reported. 0.964286 -11.4579 overfitting. 0.614286 -11.458 best result 0.666667 -11.4584 eventually 0.90625 -11.4585 section 3 presents 0.882353 -11.4587 0.38 0.933333 -11.4588 auxiliaries 0.790698 -11.4588 arity 0.308511 -11.4589 come 0.804878 -11.4591 50.0 0.804878 -11.4591 more fine-grained 0.804878 -11.4592 saying 0.842105 -11.4594 rule’s 0.684211 -11.4594 9, 0.882353 -11.4596 long-term 0.861111 -11.4597 builds on 0.368687 -11.4599 × 0.650794 -11.4602 adaptation. 0.626866 -11.4602 reordering. 0.336134 -11.4602 pp 0.219424 -11.4603 there is a 0.820513 -11.4603 exist, 0.933333 -11.4605 tabular 0.357143 -11.4605 connective 0.804878 -11.4606 rule: 0.861111 -11.4609 history, 0.575 -11.4611 preserve 0.623188 -11.4611 partitioning 0.772727 -11.4611 syntactic dependencies 0.690909 -11.4611 inc. 0.861111 -11.4613 160 0.790698 -11.4614 118 0.964286 -11.4616 syntactical 0.247706 -11.4616 actual 0.39645 -11.4618 effect on 0.90625 -11.4618 litman 0.126718 -11.462 class 0.820513 -11.4623 act as 0.544444 -11.4629 translation probabilities 0.882353 -11.463 dialogue management 0.703704 -11.4633 green 0.614286 -11.4636 hypernym 0.861111 -11.4637 whom 0.933333 -11.4638 133 0.933333 -11.4639 image, 0.772727 -11.464 argmin 0.933333 -11.4642 usage, 0.933333 -11.4642 distinguishing between 0.882353 -11.4643 iwslt 0.193548 -11.4644 concept 0.804878 -11.4647 english-german 0.772727 -11.4647 activated 0.666667 -11.4648 entity types 0.76087 -11.4649 children. 0.933333 -11.4651 opennlp 0.861111 -11.4651 large numbers 0.614286 -11.4655 cluster, 0.964286 -11.4656 loosely 0.964286 -11.466 epsilon 0.520408 -11.4661 yang 0.964286 -11.4662 supervision, 0.575 -11.4662 long-distance 0.964286 -11.4663 stevenson, 0.964286 -11.4664 qualitatively 0.933333 -11.4665 check whether 0.744681 -11.4666 lexicons, 0.103448 -11.4668 is not 0.882353 -11.4669 came from 0.325397 -11.4673 − 0.861111 -11.4674 ω 0.964286 -11.4675 justify 0.509804 -11.4677 intelligent 0.690909 -11.4677 a bit 0.76087 -11.468 eliminated 0.129052 -11.4681 uses 0.478261 -11.4682 simplify 0.360976 -11.4682 log-linear 0.964286 -11.4684 frame-semantic 0.90625 -11.4685 recommend 0.744681 -11.4687 quadratic 0.933333 -11.4688 145 0.90625 -11.4689 plurals 0.90625 -11.469 funny 0.861111 -11.4691 processes. 0.187583 -11.4691 known 0.964286 -11.4692 nbest 0.964286 -11.4692 human-annotated 0.772727 -11.4692 user interface 0.515152 -11.4694 article, 0.772727 -11.4696 liwc 0.90625 -11.4696 bilingual dictionaries 0.882353 -11.4697 comon 0.482456 -11.4697 displayed 0.122599 -11.4697 bleu 0.467213 -11.4698 0 0 . 0.933333 -11.4698 (our 0.964286 -11.47 kim et al. 0.964286 -11.47 multidocument 0.933333 -11.4702 distances between 0.964286 -11.4704 limited, 0.90625 -11.4707 lsh 0.711538 -11.4708 i’m 0.772727 -11.4709 interleaving 0.837838 -11.4709 hal 0.60274 -11.471 characters, 0.605634 -11.471 ~ 0.804878 -11.4712 couple of 0.964286 -11.4715 probabilistically 0.964286 -11.4715 ambiguous, 0.566265 -11.4715 vast 0.933333 -11.4719 algorithm: 0.429577 -11.472 greatly 0.882353 -11.472 switching 0.837838 -11.4723 greedily 0.837838 -11.4723 nlf 0.964286 -11.4725 realizing 0.964286 -11.4732 assistance 0.391813 -11.4732 appears in 0.772727 -11.4732 in some cases 0.90625 -11.4732 morph 0.964286 -11.4733 agreements 0.567901 -11.4733 much higher 0.734694 -11.4736 positional 0.882353 -11.474 cpu 0.804878 -11.474 0.87 0.964286 -11.474 early modern 0.964286 -11.4741 conjuncts 0.964286 -11.4744 baldwin 0.605634 -11.4748 conditions, 0.933333 -11.4748 process: 0.964286 -11.4749 trigger, 0.785714 -11.475 searching for 0.964286 -11.4751 intentions 0.964286 -11.4751 sentencelevel 0.576923 -11.4757 smallest 0.820513 -11.4759 0.49 0.171111 -11.4759 articles 0.882353 -11.4761 ssrs 0.605634 -11.4761 prefixes 0.526316 -11.4762 neighbor 0.964286 -11.4762 content-based 0.470588 -11.4764 contiguous 0.575 -11.4764 data points 0.964286 -11.4766 adaptation, 0.90625 -11.4766 denotation 0.90625 -11.4766 e) 0.785714 -11.4767 deliver 0.964286 -11.4768 e-rater 0.964286 -11.4768 preand 0.964286 -11.4768 drawbacks 0.964286 -11.4771 responding 0.133603 -11.4775 p 0.655738 -11.4776 instantiation 0.0907956 -11.4777 lexical 0.964286 -11.4779 differed 0.964286 -11.4779 difficult, 0.964286 -11.4779 table 1 summarizes 0.285714 -11.478 posterior 0.964286 -11.478 practices 0.785714 -11.4781 centered 0.698113 -11.4782 dictionaries. 0.964286 -11.4783 broad-coverage 0.964286 -11.4783 sub-words 0.964286 -11.4784 haiti 0.394118 -11.4788 write 0.661017 -11.4792 s., 0.144094 -11.4796 words, 0.964286 -11.4797 customized 0.964286 -11.4797 fast, 0.964286 -11.4798 cardinal 0.403727 -11.4798 clusters. 0.964286 -11.48 feeling 0.964286 -11.48 table 1 lists 0.532609 -11.48 improvements over 0.744681 -11.48 exploration of 0.964286 -11.4806 134 0.964286 -11.4806 (snow 0.964286 -11.4808 167 0.964286 -11.4808 transition, 0.273504 -11.4808 entry 0.964286 -11.4809 tens 0.837838 -11.4809 in turn, 0.8 -11.481 semantic change 0.964286 -11.4811 concise 0.964286 -11.4811 more, 0.630769 -11.4811 network, 0.755556 -11.4812 w/ 0.964286 -11.4813 mean, 0.964286 -11.4814 inspiration 0.964286 -11.4814 specialised 0.605634 -11.4815 it appears 0.711538 -11.4818 rule markov 0.964286 -11.482 vsm 0.698113 -11.482 syllables 0.350467 -11.4823 extraction, 0.878788 -11.4823 targeted self-training 0.837838 -11.4824 95% confidence 0.90625 -11.4824 binned 0.964286 -11.4825 nell’s 0.878788 -11.4825 parallel corpora, 0.576923 -11.4826 only a small 0.191074 -11.4827 matrix 0.878788 -11.4827 unlexicalized 0.219557 -11.4828 evidence 0.90625 -11.4828 gr 0.785714 -11.483 an iterative 0.698113 -11.483 itg 0.429577 -11.4834 cn 0.964286 -11.4838 0.14 0.964286 -11.484 collectively 0.964286 -11.4842 modularity 0.90625 -11.4842 in section 6, 0.964286 -11.4845 0.05, 0.964286 -11.4845 paradigm, 0.964286 -11.4845 imposing 0.8 -11.4846 we conclude that 0.0711257 -11.4847 feature 0.857143 -11.485 squared 0.964286 -11.4851 pra 0.772727 -11.4851 putting 0.964286 -11.4853 optimise 0.567901 -11.4853 to induce 0.837838 -11.4853 translationese 0.837838 -11.4854 adjective, 0.964286 -11.4856 addressee 0.24594 -11.4857 2006) 0.212069 -11.4858 svm 0.586667 -11.4859 expressed as 0.964286 -11.4861 economy 0.446154 -11.4862 baselines. 0.964286 -11.4864 bringing 0.655738 -11.4864 two kinds 0.964286 -11.4868 results; 0.630769 -11.4871 det 0.964286 -11.4872 (corresponding 0.964286 -11.4874 230 0.526882 -11.4874 inventory 0.339207 -11.4876 working 0.443609 -11.4877 in eq. 0.443609 -11.4879 inference. 0.878788 -11.4881 lattice, 0.645161 -11.4882 tree kernel 0.964286 -11.4883 encyclopedia 0.878788 -11.4884 figure 5 shows 0.785714 -11.4886 cu 0.72 -11.4886 gildea 0.837838 -11.4886 world, 0.50495 -11.4888 seek 0.785714 -11.4891 class-based 0.605634 -11.4891 deceptive 0.594595 -11.4893 results reported 0.964286 -11.4894 derives 0.933333 -11.4895 mild 0.72 -11.4895 spectrum 0.933333 -11.4897 (shown 0.50495 -11.4897 mt evaluation 0.903226 -11.49 indexed by 0.964286 -11.4902 campaign 0.256345 -11.4904 we believe 0.645161 -11.4905 brings 0.546512 -11.4905 sentence: 0.964286 -11.4906 8% 0.560976 -11.4906 should have 0.744681 -11.4907 (och, 2003) 0.785714 -11.491 pls 0.495238 -11.4911 additionally 0.837838 -11.4914 scopes 0.442748 -11.4915 pick 0.964286 -11.4915 131 0.755556 -11.4915 attribute selection 0.878788 -11.4915 contrary, 0.8 -11.4916 monologue 0.785714 -11.4919 pedersen 0.815789 -11.4926 c: 0.903226 -11.4931 disk 0.964286 -11.4932 h: 0.964286 -11.4933 facebook 0.560976 -11.4933 beliefs 0.878788 -11.4933 sh 0.520833 -11.4933 our model, 0.8 -11.4936 we shall 0.964286 -11.4937 count, 0.56962 -11.4937 acronym 0.964286 -11.4938 functions: 0.931034 -11.4939 kl-divergence 0.931034 -11.4941 we have: 0.931034 -11.4941 configurations, 0.8 -11.4941 recal 0.903226 -11.4945 significantly improved 0.350711 -11.4948 the former 0.931034 -11.4952 cp-net 0.578947 -11.4952 impression 0.931034 -11.4953 ta 0.72 -11.4953 results are presented 0.402516 -11.4953 besides 0.785714 -11.4954 effective, 0.698113 -11.4954 attractive 1 -11.4955 corrections. 1 -11.4955 toolkit. 1 -11.4955 subtrees. 1 -11.4955 end. 1 -11.4955 relevance. 1 -11.4955 tables. 1 -11.4955 exists. 1 -11.4955 framenet. 1 -11.4955 cache. 1 -11.4955 child. 0.962963 -11.4956 etc). 0.173862 -11.4957 v 0.878788 -11.496 0.92 0.72 -11.4962 our experimental results 0.815789 -11.4964 measurements 0.578947 -11.4964 reference translation 0.785714 -11.4966 off-the-shelf 0.188446 -11.4968 improvements 0.390533 -11.4968 obvious 0.931034 -11.497 119 0.878788 -11.497 (banko 0.645161 -11.4971 (li et 0.931034 -11.4971 exemplars 0.931034 -11.4978 can’t 0.521277 -11.4978 maintaining 0.755556 -11.4982 speaking, 0.302491 -11.4983 subjects 0.439394 -11.4987 semantics. 0.634921 -11.4992 we implement 0.729167 -11.4993 formulated as 0.0571176 -11.4993 at 0.327801 -11.4994 composition 0.103263 -11.4995 scores 0.375691 -11.4995 needed to 0.815789 -11.4995 il 0.466102 -11.4996 aid 0.0686328 -11.4996 both 0.931034 -11.4997 5-grams 0.903226 -11.4997 confirmation 0.962963 -11.4999 posterior probabilities 0.214669 -11.5 participants 0.103793 -11.5002 human 0.705882 -11.5002 2 background 0.586667 -11.5003 and eisner, 0.962963 -11.5004 simplex 0.931034 -11.5006 ∩ 0.678571 -11.5006 resource, 0.962963 -11.5006 etzioni 0.698113 -11.5007 preserving 0.878788 -11.5007 wfst 0.50495 -11.5009 meant 0.729167 -11.5011 conventions 0.8 -11.5012 section 7. 0.878788 -11.5013 phrase table, 0.767442 -11.5014 systemt 0.521277 -11.5015 tutorial 0.289902 -11.5015 adjective 0.473684 -11.5017 2.4 0.857143 -11.5018 truthful 0.8 -11.5018 being able 0.931034 -11.502 γ, 0.931034 -11.502 education, 0.5625 -11.5021 expressed by 0.903226 -11.5021 l) 0.903226 -11.5021 adequately 0.962963 -11.5021 (blunsom 0.366492 -11.5022 synonyms 0.833333 -11.5023 feature sets, 0.73913 -11.5024 cognates 0.878788 -11.5026 operations, 0.962963 -11.5026 5.0 0.903226 -11.5027 deception 0.214286 -11.5028 an example 0.931034 -11.5029 news article 0.597222 -11.5031 adj 0.8 -11.5031 supplementary 0.931034 -11.5033 actually, 0.878788 -11.5034 hour 0.533333 -11.5035 7, 0.962963 -11.5037 naturalness 0.931034 -11.5038 n-gram, 0.962963 -11.5039 descriptions, 0.962963 -11.504 power, 0.878788 -11.5042 dyer 0.433824 -11.5043 biological 0.621212 -11.5044 much less 0.903226 -11.5044 b1 0.962963 -11.5045 50,000 0.685185 -11.5049 indefinite 0.6 -11.505 queue 0.266667 -11.5051 reduced 0.931034 -11.5051 diff 0.833333 -11.5051 line graph 0.385965 -11.5054 body 0.430657 -11.5056 explanation 0.578947 -11.5058 normalize 0.815789 -11.5059 re-scoring 0.442748 -11.5061 input, 0.903226 -11.5063 df 0.246445 -11.5063 attribute 0.962963 -11.5065 cws 0.931034 -11.5066 sentences: 0.6 -11.5066 cross-domain 0.103234 -11.5067 event 0.962963 -11.5067 relevant. 0.65 -11.5068 relating 0.962963 -11.5068 succeeding 0.170045 -11.5069 d 0.138889 -11.5069 sequence 0.815789 -11.5072 tokenization, 0.554217 -11.5072 broken 0.755556 -11.5075 didn’t 0.962963 -11.5079 executed 0.204918 -11.5081 count 0.410596 -11.5082 technologies 0.931034 -11.5083 (around 0.510204 -11.5083 proficiency 0.350962 -11.5084 files 0.962963 -11.5085 (pedersen 0.931034 -11.5086 122 0.833333 -11.5088 0.12 0.547619 -11.5091 mined 0.65 -11.5091 oil 0.292929 -11.5091 light 0.931034 -11.5092 values: 0.685185 -11.5097 term. 0.962963 -11.5103 mildly 0.962963 -11.5104 devise 0.102166 -11.5105 very 0.767442 -11.5108 multi-word expressions 0.962963 -11.5108 inventories 0.931034 -11.5109 (words 0.962963 -11.5109 ner, 0.8 -11.5109 exemplified 0.364583 -11.511 pronoun 0.187845 -11.511 novel 0.473684 -11.5111 occurs in 0.962963 -11.5113 (1999), 0.425532 -11.5114 investigation 0.962963 -11.5115 3.1) 0.903226 -11.5117 cardie, 0.625 -11.5118 to guide 0.962963 -11.512 unsupervised, 0.342593 -11.5121 signature 0.903226 -11.5122 african 0.962963 -11.5122 claimed 0.962963 -11.5122 memory, 0.589041 -11.5123 surprisingly, 0.962963 -11.5123 strube 0.903226 -11.5124 schedule 0.962963 -11.5127 comparably 0.962963 -11.5127 mrda 0.962963 -11.5128 discovers 0.581081 -11.5129 collins, 0.555556 -11.5129 on-line 0.931034 -11.5131 meronymy 0.852941 -11.5131 server 0.962963 -11.5131 abundant 0.962963 -11.5131 regularity 0.445312 -11.5132 proved 0.815789 -11.5133 long-range 0.852941 -11.5136 three kinds of 0.962963 -11.5137 functor 0.714286 -11.5139 to augment 0.547619 -11.5139 recursively 0.962963 -11.5139 lightweight 0.73913 -11.5141 evolution 0.273529 -11.5141 international 0.878788 -11.5142 carries 0.833333 -11.5143 proposed, 0.962963 -11.5143 (2000), 0.962963 -11.5144 justified 0.962963 -11.5144 retrain 0.833333 -11.5145 1992). 0.815789 -11.5146 proposed. 0.263014 -11.5147 correction 0.962963 -11.5148 outliers 0.412162 -11.5149 topic, 0.962963 -11.5149 nenkova 0.962963 -11.5149 112 0.685185 -11.5151 brought 0.284345 -11.5152 result in 0.962963 -11.5152 excludes 0.672727 -11.5153 0.06 0.962963 -11.5153 ratio, 0.962963 -11.5154 spreading 0.634921 -11.5155 ∪ 0.655172 -11.5155 katakana 0.780488 -11.5156 comprised of 0.477477 -11.5157 connected to 0.284345 -11.5159 identical 0.962963 -11.5159 certainty 0.852941 -11.5162 interpretable 0.962963 -11.5162 imposes 0.962963 -11.5163 169 0.547619 -11.5166 to retrieve 0.962963 -11.5166 hyper-parameters 0.903226 -11.5167 csr 0.319838 -11.5167 dimensions 0.473214 -11.5168 small, 0.962963 -11.5168 speeds 0.962963 -11.517 cognition 0.5625 -11.5171 each pair of 0.815789 -11.5171 context window 0.903226 -11.5172 ef 0.962963 -11.5172 intentionally 0.852941 -11.5174 chan 0.833333 -11.5174 interfaces 0.73913 -11.5175 text simplification 0.465517 -11.5175 availability of 0.61194 -11.5175 propose a novel 0.962963 -11.5176 preparing 0.591549 -11.5177 string-to-tree 0.625 -11.5179 wsi 0.931034 -11.518 faust 0.5625 -11.5182 facial 0.815789 -11.5182 107 0.780488 -11.5183 xue 0.903226 -11.5185 port 0.54023 -11.5186 word-aligned 0.903226 -11.5186 132 0.73913 -11.5187 lastly, 0.962963 -11.5187 mentioned above 0.962963 -11.5187 ∅ 0.962963 -11.5188 infoboxes 0.962963 -11.519 information) 0.903226 -11.5192 drugs 0.903226 -11.5194 lhs 0.692308 -11.5195 opinion analysis 0.962963 -11.5195 ls 0.815789 -11.5196 parameter vector 0.962963 -11.5196 composition, 0.962963 -11.5197 assembled 0.75 -11.5198 riedel 0.903226 -11.5199 tf 0.666667 -11.5201 2006. 0.962963 -11.5201 init 0.931034 -11.5203 most promising 0.589041 -11.5204 minimizing 0.962963 -11.5205 159 0.833333 -11.5208 general-purpose 0.833333 -11.521 for each sentence, 0.962963 -11.5214 tree-to-tree 0.962963 -11.5214 patterns: 0.962963 -11.5214 gui 0.815789 -11.5215 stephen 0.931034 -11.5218 book requests 0.324895 -11.5219 more likely 0.852941 -11.5219 adverbs, 0.962963 -11.5221 3.7 0.903226 -11.5223 (1992) 0.61194 -11.5223 training corpus, 0.962963 -11.5223 markers, 0.962963 -11.5224 liked 0.962963 -11.5228 (a), 0.903226 -11.5229 costs. 0.962963 -11.5229 lth 0.723404 -11.5234 1; 0.75 -11.5235 arabic. 0.591549 -11.5237 ma 0.962963 -11.5237 manipulate 0.962963 -11.5239 constrains 0.962963 -11.5239 pscfg 0.962963 -11.5239 temporally 0.714286 -11.5241 7.1 0.962963 -11.5246 goodness 0.547619 -11.5247 replaced with 0.962963 -11.5247 suport 0.358974 -11.5248 colour 0.794872 -11.5254 czech, 0.258667 -11.5254 we assume 0.852941 -11.5256 performs well 0.962963 -11.5257 synsets, 0.962963 -11.5257 347 0.692308 -11.5258 an extended 0.833333 -11.5259 motion 0.962963 -11.5259 do, 0.852941 -11.5262 hierarchies 0.794872 -11.5262 task-based 0.962963 -11.5262 surprise 0.962963 -11.5263 interesting, 0.962963 -11.5267 compile 0.347826 -11.5272 tags, 0.7 -11.5273 computational approaches 0.672727 -11.5274 personalized 0.52809 -11.5275 600 0.12844 -11.5275 x 0.962963 -11.5275 charts 0.962963 -11.5275 svm, 0.75 -11.5277 backchannel 0.833333 -11.5278 cleaning 0.875 -11.5278 previously reported 0.655172 -11.5279 is composed of 0.962963 -11.528 contention 0.602941 -11.5281 typically, 0.469027 -11.5282 5% 0.387879 -11.5283 suppose 0.931034 -11.5284 cotton 0.692308 -11.5285 different, 0.962963 -11.5285 iterates 0.962963 -11.5286 2a 0.931034 -11.5287 mwu 0.9 -11.5287 spellings 0.962963 -11.5288 decoded 0.655172 -11.529 columbia 0.962963 -11.5291 snomed 0.875 -11.5295 opens 0.962963 -11.5295 sucre 0.962963 -11.5296 165 0.319672 -11.5298 we discuss 0.274924 -11.5301 transfer 0.505155 -11.5301 syntactic information 0.534884 -11.5301 lf 0.672727 -11.5301 sentence) 0.639344 -11.5303 output: 0.962963 -11.5304 ellipses 0.962963 -11.5304 phases 0.962963 -11.5305 commonsense 0.49505 -11.5307 query, 0.962963 -11.5308 non-overlapping 0.469027 -11.5311 tested on 0.962963 -11.5312 (row 0.5 -11.5314 test, 0.692308 -11.5315 macaon 0.220273 -11.5318 constructed 0.875 -11.5319 toutanova 0.564103 -11.532 vertical 0.11742 -11.5321 output 0.962963 -11.5322 trace 0.928571 -11.5323 assumptions about 0.406667 -11.5324 factored 0.348039 -11.5326 inter-annotator 0.413793 -11.5327 0.7 0.761905 -11.5327 segmentor 0.962963 -11.5327 spans, 0.852941 -11.533 death 0.962963 -11.533 quote 0.672727 -11.5332 constraint. 0.179355 -11.5333 amount of 0.564103 -11.5333 procedings of the 0.9 -11.5334 trec 0.9 -11.5338 system) 0.810811 -11.5338 maximally 0.75 -11.534 claim that 0.810811 -11.534 strongest 0.780488 -11.534 vignettes 0.9 -11.5341 kitchen 0.438462 -11.5342 appeared 0.173285 -11.5344 vectors 0.472727 -11.5345 neighboring 0.780488 -11.5347 applicability of 1 -11.5348 abstracts. 1 -11.5348 czech. 1 -11.5348 submission. 1 -11.5348 solutions. 1 -11.5348 trained. 1 -11.5348 moves. 1 -11.5348 occurrences. 1 -11.5348 reduced. 1 -11.5348 lemma. 1 -11.5348 differently. 1 -11.5348 boundary. 1 -11.5348 letters. 1 -11.5348 ontologies. 1 -11.5348 machine. 1 -11.5348 difficulty. 1 -11.5348 either. 1 -11.5348 date. 1 -11.5348 similarities. 1 -11.5348 axioms. 0.961538 -11.5349 let’s go 0.961538 -11.5349 redundancy. 0.961538 -11.5349 power. 0.875 -11.5349 stay 0.402597 -11.5355 conduct 0.447154 -11.5359 transcripts 0.852941 -11.5359 incoherent 0.564103 -11.536 the berkeley parser 0.481132 -11.5361 times. 0.9 -11.5362 immediately after 0.75 -11.5362 transcriptions 0.875 -11.5364 overhead 0.469027 -11.5369 acquire 0.852941 -11.5372 gathered from 0.928571 -11.5372 factoid 0.331858 -11.5379 derive 0.928571 -11.5381 spontaneous speech 0.344498 -11.5382 liu 0.928571 -11.5391 evaluation: 0.733333 -11.5392 0.26 0.794872 -11.5392 slovene 0.7 -11.5393 solely on 0.304511 -11.5401 contribution of 0.181818 -11.5402 increase 0.928571 -11.5404 138 0.810811 -11.5405 this paper proposes 0.961538 -11.5405 srl-aware scfg 0.928571 -11.5407 monotonically 0.9 -11.5411 lvc 0.606061 -11.5413 indian 0.928571 -11.5413 ridge 0.961538 -11.5413 72.5 0.961538 -11.5414 forced alignment 0.75 -11.5415 history. 0.928571 -11.5422 ha 0.54878 -11.5423 unification 0.852941 -11.5423 replicated 0.928571 -11.5423 settings: 0.928571 -11.5424 adults 0.54878 -11.5428 predictions. 0.9 -11.5428 organisation 0.5 -11.5428 other languages 0.556962 -11.5429 corpus consists of 0.75 -11.543 compounds. 0.485714 -11.543 tackle 0.9 -11.5435 non-linear 0.644068 -11.5436 head-dependents 0.708333 -11.5438 collections. 0.810811 -11.5439 august 0.184979 -11.544 effective 0.363636 -11.5442 naive 0.928571 -11.5443 conference, 0.472727 -11.5446 corrections 0.961538 -11.5449 records, 0.9 -11.545 substantial improvement 0.961538 -11.545 shortcoming 0.961538 -11.545 underspecification 0.267442 -11.5454 word sense 0.292683 -11.5454 allowed 0.961538 -11.5454 interactions, 0.341232 -11.5457 attempts 0.629032 -11.5459 approximated 0.679245 -11.5461 target language, 0.7 -11.5463 experiments reported 0.794872 -11.5464 f) 0.9 -11.5465 norm 0.18306 -11.5465 create 0.300366 -11.5467 were used 0.629032 -11.5468 lagrangian 0.875 -11.5469 score) 0.928571 -11.5471 0.21 0.733333 -11.5472 stacked 0.606061 -11.5472 comprising 0.961538 -11.5472 human-written 0.761905 -11.5473 future research. 0.7 -11.5475 accordingly, 0.128967 -11.5475 phrases 0.961538 -11.5476 internally 0.961538 -11.5479 65% 0.961538 -11.5479 eliminates 0.103613 -11.5481 tree 0.961538 -11.5482 low-resource 0.315789 -11.5483 instead, 0.961538 -11.5485 hwdep 0.928571 -11.5487 document d, 0.476636 -11.5488 keyphrases 0.679245 -11.5493 average length 0.961538 -11.5495 entail 0.649123 -11.5495 wa 0.961538 -11.5496 pre-signature 0.961538 -11.5496 come up 0.961538 -11.5496 sense-tagged 0.875 -11.5497 recordings 0.9 -11.5498 thesis 0.961538 -11.55 choice, 0.961538 -11.5502 neutral, 0.0679685 -11.5503 corpus 0.961538 -11.5505 rules) 0.9 -11.5506 [1] 0.660714 -11.5506 reorder 0.0924253 -11.5509 about 0.961538 -11.5509 libsvm 0.292683 -11.5514 purpose 0.961538 -11.5515 incorrect, 0.961538 -11.5515 wider range 0.708333 -11.5515 scan 0.606061 -11.5517 markables 0.961538 -11.5517 anaphors 0.961538 -11.5517 holistic 0.313253 -11.5517 occurrence 0.388889 -11.5518 participant 0.875 -11.552 bootstrapping. 0.9 -11.5521 0.93 0.542169 -11.5521 a brief 0.12322 -11.5521 generated 0.535714 -11.5522 phrasebased 0.510638 -11.5522 eliminate 0.417266 -11.5522 otherwise, 0.154599 -11.5523 (2) 0.516484 -11.5523 balance 0.961538 -11.5523 emphasizes 0.375723 -11.5528 events, 0.961538 -11.553 calculus 0.961538 -11.5532 137 0.961538 -11.5534 optimality 0.848485 -11.5535 hedge 0.828571 -11.5536 can effectively 0.961538 -11.5536 c3 0.158763 -11.5537 next 0.141455 -11.5539 describe 0.961538 -11.5539 regional 0.961538 -11.5539 graphbased 0.961538 -11.554 97% 0.494949 -11.5541 sure 0.333333 -11.5541 scores, 0.961538 -11.5541 madnani 0.464286 -11.5543 “ 0.666667 -11.5544 geographical 0.9 -11.5547 (a) (b) 0.7 -11.5547 had no 0.9 -11.5548 vg 0.9 -11.5548 aer 0.45 -11.5549 (9) 0.585714 -11.5549 table 6. 0.828571 -11.555 gildea, 0.472222 -11.555 lattices 0.961538 -11.5551 seeds, 0.19076 -11.5552 (3) 0.575342 -11.5552 impossible 0.301115 -11.5553 in this work, 0.961538 -11.5555 19th 0.961538 -11.5556 13% 0.961538 -11.5557 206 0.961538 -11.5557 slang 0.961538 -11.5557 smaller, 0.961538 -11.5557 partition function 0.810811 -11.556 improved. 0.679245 -11.5561 south 0.961538 -11.5561 rename 0.961538 -11.5561 lemmas. 0.961538 -11.5562 copyright 0.828571 -11.5563 design, 0.9 -11.5563 september 0.371429 -11.5563 extension of 0.961538 -11.5564 forward-backward 0.961538 -11.5564 commands 0.362162 -11.5565 record 0.928571 -11.5565 objective function, 0.961538 -11.5566 researcher 0.961538 -11.5568 interpolating 0.961538 -11.5569 shift, 0.961538 -11.557 0.22 0.24105 -11.5572 (4) 0.9 -11.5572 to) 0.961538 -11.5572 delivering 0.961538 -11.5573 images, 0.928571 -11.5574 amber 0.961538 -11.5574 controversial 0.49 -11.5574 (line 0.278846 -11.5579 (a 0.828571 -11.5582 bars 0.186514 -11.5586 forms 0.961538 -11.5586 bernoulli 0.686275 -11.5587 shifts 0.961538 -11.5587 attitudes 0.22547 -11.559 (i.e. 0.324675 -11.559 clause 0.848485 -11.5592 last row 0.848485 -11.5592 ’ 0.406897 -11.5596 semantic similarity 0.686275 -11.5597 anaphor 0.810811 -11.5597 0.27 0.961538 -11.5597 necessity 0.961538 -11.5599 readable 0.159832 -11.5599 weights 0.828571 -11.5601 ssa 0.961538 -11.5602 catch 0.775 -11.5603 fes 0.666667 -11.5604 less likely 0.789474 -11.5605 du 0.3875 -11.5605 moves 0.961538 -11.5606 2n 0.575342 -11.5611 exist in 0.961538 -11.5615 154 0.128855 -11.5616 provide 0.961538 -11.5616 inductive 0.961538 -11.5616 english) 0.828571 -11.5618 adjectival 0.961538 -11.562 ood 0.961538 -11.5622 exhaustively 0.649123 -11.5622 z. 0.828571 -11.5623 fung 0.529412 -11.5623 to enable 0.961538 -11.5623 (2010a) 0.708333 -11.5624 rejected 0.414286 -11.5625 to keep 0.961538 -11.5625 polish 0.744186 -11.5632 argued that 0.928571 -11.5632 chunking, 0.609375 -11.5632 76 0.961538 -11.5633 pov 0.961538 -11.5633 environmental 0.870968 -11.5634 collaborators 0.346535 -11.5634 indicated 0.961538 -11.5635 naturally, 0.928571 -11.5636 (so 0.961538 -11.5636 section 2.2. 0.961538 -11.5639 enforcing 0.606061 -11.564 scripts 0.789474 -11.5641 0.34 0.961538 -11.5641 repeating 0.870968 -11.5644 subtasks 0.961538 -11.5644 martin, 0.961538 -11.5647 anywhere 0.961538 -11.5647 177 0.961538 -11.5649 changes, 0.577465 -11.5651 this indicates that 0.522727 -11.5652 ends 0.961538 -11.5652 idea, 0.775 -11.5653 word; 0.120576 -11.5654 values 0.449153 -11.566 comment 0.666667 -11.5662 may contain 0.619048 -11.5663 vote 0.449153 -11.5663 counting 0.775 -11.5665 √ 0.961538 -11.5666 actionable 0.961538 -11.5666 career 0.744186 -11.5669 (over 0.961538 -11.567 recursion 0.666667 -11.5671 horizontal 0.193199 -11.5671 outperforms 0.848485 -11.5672 toy 0.961538 -11.5673 accessibility 0.961538 -11.5674 lia 0.789474 -11.5676 bears 0.961538 -11.5677 root, 0.637931 -11.5682 detects 0.961538 -11.5686 svo 0.961538 -11.5688 propagation, 0.870968 -11.569 berg-kirkpatrick 0.870968 -11.569 phenotypes 0.535714 -11.569 spontaneous 0.686275 -11.5691 (sec. 0.266272 -11.5693 spelling 0.649123 -11.5695 working with 0.585714 -11.5696 in cases where 0.961538 -11.5697 clicked 0.961538 -11.5697 unsegmented 0.961538 -11.57 imperfect 0.424242 -11.5701 of interest 0.870968 -11.5703 i,j 0.424242 -11.5704 cross-validation 0.744186 -11.5704 resampling 0.961538 -11.5705 russian, 0.961538 -11.5705 mr. 0.346734 -11.5706 vary 0.654545 -11.5707 fragments, 0.775 -11.5708 remained 0.775 -11.5709 span, 0.961538 -11.5712 grey 0.352332 -11.5714 projected 0.961538 -11.5714 observable 0.654545 -11.5716 ng, 0.585714 -11.5716 7 conclusions 0.756098 -11.5721 native language 0.717391 -11.5724 may lead 0.756098 -11.5728 our findings 0.896552 -11.5728 samples, 0.775 -11.5729 radio 0.686275 -11.5732 contrasting 0.248705 -11.5732 indicating 0.243176 -11.5733 interest 0.828571 -11.5734 discovered by 0.925926 -11.5736 dogs 0.961538 -11.5736 domain experts 0.805556 -11.5739 mccallum 0.180217 -11.5744 annotators 0.789474 -11.5744 nell 0.523256 -11.5748 µ 0.925926 -11.5752 158 0.870968 -11.5753 assignment, 0.622951 -11.5754 subgraph 0.334906 -11.5755 overall, 0.421053 -11.5756 labeled as 1 -11.5756 needs. 1 -11.5756 (right). 1 -11.5756 (1999). 1 -11.5756 2009)). 1 -11.5756 folds. 1 -11.5756 dependent. 1 -11.5756 taxonomy. 1 -11.5756 ratio. 1 -11.5756 updates. 1 -11.5756 slot. 1 -11.5756 transformations. 1 -11.5756 dev. 1 -11.5756 cf. 1 -11.5756 scoring. 1 -11.5756 larger. 1 -11.5756 api. 1 -11.5756 association. 1 -11.5756 indicators. 1 -11.5756 areas. 1 -11.5756 scene. 1 -11.5756 extractions. 0.460177 -11.5756 marcu, 0.609375 -11.5757 the results are shown in 0.96 -11.5757 run. 0.505376 -11.576 to establish 0.96 -11.5762 10-fold cross-validation 0.480392 -11.5765 contexts, 0.6 -11.5765 an increase 0.693878 -11.5767 sampling. 0.756098 -11.5767 source-language 0.848485 -11.5767 source-context 0.56 -11.5768 drop in 0.896552 -11.5769 investment 0.925926 -11.577 has attracted 0.511111 -11.577 the availability of 0.896552 -11.5773 property, 0.334906 -11.5773 figures 0.485149 -11.5776 some of these 0.96 -11.5777 text-to-text generation 0.271605 -11.578 over all 0.325893 -11.5783 study, 0.925926 -11.5788 masculine 0.96 -11.5788 feng, 0.870968 -11.5791 relationships, 0.96 -11.5792 clls 0.480392 -11.5793 reveals 0.96 -11.5793 degraded 0.099014 -11.5796 example 0.96 -11.5796 fluency ranking 0.96 -11.5798 gold standards 0.925926 -11.5799 more comprehensive 0.896552 -11.5799 mterater 0.588235 -11.58 observing 0.132789 -11.5801 combination 0.925926 -11.5806 pursued 0.848485 -11.5806 word-pair 0.283784 -11.5806 interface 0.870968 -11.5808 give us 0.925926 -11.5813 irregular 0.870968 -11.5814 (henceforth 0.870968 -11.5814 moore 0.805556 -11.5817 centre for 0.452174 -11.5818 probabilities. 0.622951 -11.582 an empirical 0.789474 -11.582 extractors 0.925926 -11.5821 (grant 0.96 -11.5821 tac 2008 0.925926 -11.5822 clas 0.523256 -11.5822 states that 0.96 -11.5822 60.0 0.769231 -11.5828 solver 0.96 -11.5829 stand-alone 0.96 -11.5832 guo 0.805556 -11.5832 128 0.925926 -11.5835 256 0.925926 -11.5835 treebanks. 0.896552 -11.584 “he 0.756098 -11.5842 files. 0.121495 -11.5845 speech 0.756098 -11.585 third-order 0.896552 -11.5851 50, 0.68 -11.5853 mt08 0.361111 -11.5855 formulation 0.896552 -11.5857 lt 0.206463 -11.5857 elements 0.183239 -11.5858 (2009) 0.654545 -11.5858 > 0 0.925926 -11.5862 inference axioms 0.523256 -11.5862 truth 0.448276 -11.5865 only. 0.789474 -11.5866 35% 0.896552 -11.5866 walker 0.459459 -11.5866 ours 0.612903 -11.5866 demonstrates that 0.471698 -11.5867 we tested 0.96 -11.5867 (given 0.57971 -11.5867 that: 0.0470281 -11.5869 all 0.132396 -11.587 selected 0.577465 -11.587 to verify 0.96 -11.5872 homogeneity 0.925926 -11.5873 allowed us 0.896552 -11.5874 paths, 0.96 -11.5875 chapters 0.702128 -11.5877 relations among 0.96 -11.5878 models; 0.96 -11.5878 exchanges 0.96 -11.5878 resembles 0.5 -11.5878 which means 0.96 -11.588 referents 0.870968 -11.5881 parameter values 0.654545 -11.5881 recommendation 0.96 -11.5882 levin 0.96 -11.5886 agrees with 0.96 -11.5891 rbc 0.96 -11.5892 registered 0.96 -11.5892 larger, 0.320175 -11.5893 dialog 0.896552 -11.5895 fo 0.769231 -11.5897 quantities 0.96 -11.5898 note, however, 0.96 -11.5899 rendering 0.823529 -11.59 light verbs 0.823529 -11.5902 inferior 0.370588 -11.5902 email 0.264095 -11.5902 effects 0.445378 -11.5904 (we 0.361111 -11.5905 klein 0.654545 -11.5905 this experiment, we 0.96 -11.5906 progressive 0.96 -11.5907 16% 0.896552 -11.5909 crossvalidation 0.413043 -11.591 (2008), 0.805556 -11.591 kbp 0.96 -11.5911 viewing 0.96 -11.5911 memory-based 0.317597 -11.5912 required to 0.96 -11.5913 extraction (ie) 0.363636 -11.5914 giza++ 0.96 -11.5915 interpretations, 0.3125 -11.5916 learning. 0.642857 -11.5916 adjacency 0.870968 -11.5917 church 0.738095 -11.5917 visible 0.727273 -11.5919 earthquake 0.738095 -11.592 (f) 0.96 -11.5922 part-of-speech tags, 0.96 -11.5922 co-occurs 0.96 -11.5922 syntactico-semantic 0.769231 -11.5924 route 0.896552 -11.5924 related, 0.6 -11.5924 zhang, 0.756098 -11.5925 (2003). 0.612903 -11.5925 graduate 0.702128 -11.5928 deals 0.96 -11.5929 nominals 0.870968 -11.593 350 0.68 -11.5932 in bold. 0.205776 -11.5932 towards 0.96 -11.5933 biomolecular 0.96 -11.5934 empty, 0.727273 -11.5935 hoc 0.68 -11.5936 chang 0.756098 -11.5936 coercion 0.96 -11.5937 moses, 0.480392 -11.5939 for evaluating 0.115604 -11.5939 dialogue 0.544304 -11.5939 prune 0.423077 -11.594 2001) 0.673077 -11.5941 surprisal 0.896552 -11.5941 (n = 0.896552 -11.5941 french–english 0.727273 -11.5942 co 0.896552 -11.5944 (1993) 0.327189 -11.5944 seems to 0.96 -11.5944 monitoring 0.96 -11.5947 dk 0.896552 -11.595 csd 0.405594 -11.5951 verbs. 0.96 -11.5951 voice, 0.896552 -11.5953 0.28 0.96 -11.5953 spaces, 0.96 -11.5953 55% 0.84375 -11.5957 multimedia 0.896552 -11.5959 enumerate 0.823529 -11.5962 paraphrasing rules 0.96 -11.5963 success, 0.96 -11.5964 opening 0.96 -11.5964 vocabularies 0.96 -11.5966 constantly 0.96 -11.5967 frequent, 0.96 -11.5968 yes, 0.96 -11.5968 enforces 0.642857 -11.597 missed 0.187406 -11.597 2, 0.96 -11.5971 delivered 0.96 -11.5973 (baseline) 0.727273 -11.5974 palmer 0.68 -11.5974 phenomena, 0.357143 -11.5977 examined 0.96 -11.5977 multi-modal 0.727273 -11.5981 to overcome this 0.267692 -11.5981 translation model 0.96 -11.5982 multidimensional 0.6 -11.5986 disjoint 0.896552 -11.5987 tv 0.410072 -11.5988 to create a 0.96 -11.5989 we, 0.96 -11.599 dynamic-programming 0.896552 -11.5991 cutting 0.96 -11.5997 informational 0.769231 -11.6001 augmentation 0.590909 -11.6001 mckeown, 0.590909 -11.6002 smt. 0.84375 -11.6004 inference procedure 0.783784 -11.6005 -> 0.334951 -11.6007 figure 3 0.96 -11.6007 visualize 0.642857 -11.6008 square6 0.296578 -11.6009 examples. 0.823529 -11.601 sentiment classification. 0.84375 -11.601 iq 0.925926 -11.6011 clustered into 0.419847 -11.6011 algorithm 1 0.711111 -11.6011 proves 0.68 -11.6013 180 0.96 -11.6013 browsing 0.866667 -11.6016 remarkable 0.896552 -11.6018 recognizes 0.517241 -11.6018 reordered 0.187879 -11.6019 utterances 0.896552 -11.602 accessing 0.642857 -11.602 microsoft research 0.866667 -11.6021 particles 0.96 -11.6022 solutions, 0.805556 -11.6022 ga 0.866667 -11.6024 actor 0.96 -11.6027 cleaned 0.823529 -11.6028 bare 0.96 -11.6029 conveys 0.925926 -11.6031 (18) 0.896552 -11.6031 averaged across 0.96 -11.6031 material, 0.727273 -11.6032 two-step 0.96 -11.6033 (baseline 0.96 -11.6034 syntactic-semantic 0.96 -11.6034 commas 0.783784 -11.6034 retrieved from 0.96 -11.6037 desire 0.631579 -11.6037 mc 0.96 -11.6038 aphasia 0.530864 -11.6038 petrov 0.96 -11.604 intelligibility 0.404255 -11.6041 guide 0.96 -11.6041 some, 0.648148 -11.6042 75% 0.896552 -11.6043 1.0. 0.603175 -11.6043 headword 0.896552 -11.6045 lda. 0.783784 -11.6048 or not, 0.96 -11.6049 m&l 0.96 -11.6052 witten, 0.96 -11.6055 2par 0.96 -11.6057 iterate 0.96 -11.606 reserved 0.616667 -11.606 $ 0.218557 -11.6061 taking 0.96 -11.6061 acknowledged 0.642857 -11.6061 reverb 0.438017 -11.6062 proposes 0.866667 -11.6063 physics 0.925926 -11.6064 baseline: 0.96 -11.6064 writer’s 0.925926 -11.6071 system-level 0.96 -11.6071 semeval-2007 0.642857 -11.6071 clark, 0.96 -11.6075 ellipsis 0.0296535 -11.6077 this 0.925926 -11.6079 best performing system 0.96 -11.6079 games, 0.823529 -11.608 contributions. 0.96 -11.6083 refining 0.866667 -11.6084 mapping, 0.866667 -11.6084 ironic 0.96 -11.6085 remember 0.96 -11.6088 scientific papers 0.648148 -11.6088 suppose that 0.96 -11.6089 anyone 0.96 -11.609 proces 0.892857 -11.609 could potentially 0.769231 -11.6091 parentheses 0.75 -11.6092 querying 0.96 -11.6092 alpha 0.96 -11.6092 sentence’s 0.571429 -11.6093 as long 0.711111 -11.6094 2005, 0.823529 -11.6097 logic, 0.8 -11.6101 expectation maximization 0.257143 -11.6103 widely 0.642857 -11.6103 high accuracy 0.48 -11.6104 kendall’s 0.8 -11.6105 lavie 0.8 -11.6106 segment, 0.96 -11.6107 the, 0.925926 -11.6108 ter (snover 0.177112 -11.6108 l 0.96 -11.611 (each 0.258721 -11.6111 occurs 0.96 -11.6117 lots 0.8 -11.6117 contact 0.866667 -11.6118 um 0.96 -11.6118 preprocessing, 0.96 -11.6122 listen 0.8 -11.6122 dative 0.84375 -11.6123 as soon 0.136731 -11.6124 step 0.378882 -11.6124 4.4 0.711111 -11.6127 parallel corpus, 0.96 -11.6128 chunkers 0.96 -11.6129 backbone 0.96 -11.613 c@1 0.24173 -11.613 call 0.738095 -11.6131 approach: 0.866667 -11.6131 alias 0.530864 -11.6132 were selected 0.590909 -11.6133 tensor 0.84375 -11.6136 λ1 0.783784 -11.6136 (poon 0.376543 -11.6136 possible. 0.475248 -11.6137 (2001) 0.96 -11.6137 (p) 0.96 -11.6139 formal definition 0.823529 -11.614 partof-speech 0.75 -11.614 distributional semantic 0.866667 -11.6141 identical, 0.96 -11.6142 cf 0.8 -11.6145 hungarian 0.8 -11.6148 feature representations 0.447368 -11.6149 clusters, 0.75 -11.6149 confirm that 0.413534 -11.6149 are usually 0.156051 -11.615 y 0.738095 -11.6152 training instance 0.923077 -11.6154 favored 0.84375 -11.6155 t-score 0.96 -11.6155 wrongly 0.280822 -11.6156 presence of 0.505618 -11.616 our model. 0.923077 -11.6161 line, 0.892857 -11.6162 pre-ordering 0.631579 -11.6162 0.81 0.75 -11.6164 very good 0.84375 -11.6168 body of work 0.84375 -11.6169 twss 0.923077 -11.617 log linear 0.6875 -11.6173 we seek 0.711111 -11.6177 spatial information 0.923077 -11.6178 learning: 0.866667 -11.6179 background knowledge 0.648148 -11.6179 syntax. 0.923077 -11.618 impressive 0.923077 -11.618 finer-grained 0.84375 -11.6181 couples 1 -11.6181 hold. 1 -11.6181 images. 1 -11.6181 2.1. 1 -11.6181 substitution. 1 -11.6181 aligned. 1 -11.6181 ratings. 1 -11.6181 parentheses. 1 -11.6181 books. 1 -11.6181 features). 1 -11.6181 minutes. 1 -11.6181 appropriate. 1 -11.6181 id. 1 -11.6181 morphemes. 1 -11.6181 streams. 1 -11.6181 variations. 1 -11.6181 del rey, 1 -11.6181 (3). 1 -11.6181 dependency. 1 -11.6181 priors. 0.958333 -11.6183 ibm model-1 0.475248 -11.6183 analyzer 0.923077 -11.6184 challenged 0.436975 -11.6186 n = 0.666667 -11.6188 this problem. 0.427419 -11.6188 trying 0.462264 -11.6189 excluded 0.958333 -11.6191 naturally occurring 0.783784 -11.6193 “how 0.738095 -11.6196 begin by 0.892857 -11.6198 collocated 0.783784 -11.6199 ‘to 0.52439 -11.6207 packed 0.8 -11.6208 d2 0.84375 -11.6209 ref 0.208333 -11.6211 describes 0.484536 -11.6213 examining 0.616667 -11.6213 lu 0.62069 -11.6219 needed. 0.5 -11.6219 digital 0.62069 -11.6222 implementation, 0.56338 -11.6223 simulate 0.958333 -11.6225 arabic-to-english 0.866667 -11.6226 0.32 0.711111 -11.6228 auto 0.178025 -11.6228 points 0.305785 -11.6229 reduces 0.8 -11.623 to do this, 0.695652 -11.6231 0.54 0.923077 -11.6231 trick 0.783784 -11.6233 mwe. 0.631579 -11.6235 task: 0.818182 -11.6237 0.17 0.249315 -11.6238 accuracy. 0.958333 -11.624 colored 0.866667 -11.624 nuclei 0.923077 -11.6243 lus 0.554054 -11.6244 m, 0.958333 -11.6247 vector-based 0.958333 -11.6247 work: 0.923077 -11.6248 groups: 0.958333 -11.625 leads us 0.8 -11.6252 version, 0.763158 -11.6255 responds 0.866667 -11.6256 92% 0.364706 -11.6257 2000). 0.554054 -11.6258 falls 0.695652 -11.6258 can capture 0.892857 -11.6258 these results suggest 0.763158 -11.626 numberof 0.573529 -11.6262 stores 0.666667 -11.6263 our experiments show 0.892857 -11.6266 wl 0.892857 -11.6267 eleven 0.958333 -11.6268 osborne, 0.958333 -11.6268 supplemented 0.151822 -11.6269 experiment 0.205993 -11.6273 estimated 0.958333 -11.6274 parallelism 0.673469 -11.6274 location. 0.454545 -11.6275 seeds 0.958333 -11.6278 bacteria biotope 0.364706 -11.628 composed of 0.0882555 -11.6282 score 0.958333 -11.6285 relationships among 0.958333 -11.6285 alternatives, 0.555556 -11.6285 salience 0.283688 -11.6287 cross 0.763158 -11.6287 subsection 0.866667 -11.6287 margin. 0.631579 -11.6287 solid 0.818182 -11.6288 ams 0.554054 -11.6288 this issue 0.958333 -11.6291 guidance 0.21371 -11.6291 resolution 0.958333 -11.6292 revealing 0.958333 -11.6293 speaking rate 0.56338 -11.6293 written by 0.866667 -11.6296 h2 0.923077 -11.6297 unknown. 0.554054 -11.6298 helpful for 0.923077 -11.6298 111 0.19187 -11.6302 right 0.763158 -11.6303 fertility 0.923077 -11.6306 definitions, 0.137615 -11.6308 generate 0.892857 -11.6309 discriminate between 0.923077 -11.631 f(x) 0.958333 -11.6312 common practice 0.958333 -11.6312 tiles 0.194305 -11.6312 having 0.8 -11.6312 recent work has 0.226757 -11.6314 tested 0.892857 -11.6314 z) 0.958333 -11.6315 (carlson 0.958333 -11.632 inadequate 0.958333 -11.6322 information-theoretic 0.439655 -11.6322 we conducted 0.546667 -11.6322 needed for 0.892857 -11.6322 shi 0.923077 -11.6322 grandsibling 0.958333 -11.6323 20 newsgroups 0.958333 -11.6324 (bergsma 0.958333 -11.6324 itspoke 0.958333 -11.6324 2this 0.892857 -11.6324 advice 0.325472 -11.6325 received 0.958333 -11.6327 frequencies, 0.268371 -11.6329 generates 0.958333 -11.633 bootstrapping, 0.866667 -11.6331 apertium 0.75 -11.6332 chemical 0.596774 -11.6332 matching. 0.101134 -11.6333 single 0.72093 -11.6334 encourages 0.75 -11.6335 mt, 0.923077 -11.6335 spring 0.958333 -11.6336 history-based 0.892857 -11.6336 loss, 0.958333 -11.6341 dvd 0.818182 -11.6342 earlier work 0.958333 -11.6344 jones 0.958333 -11.6346 boldface 0.364706 -11.6346 matrices 0.308017 -11.6346 occur in 0.958333 -11.6348 cook 0.958333 -11.6348 linguistically-motivated 0.72093 -11.6349 probabilistic models 0.958333 -11.635 bridging 0.0810968 -11.6358 algorithm 0.958333 -11.636 sub-system 0.59375 -11.636 further improve 0.958333 -11.6361 cycles 0.72093 -11.6361 feature sets. 0.892857 -11.6361 143 0.958333 -11.6363 meanings, 0.958333 -11.6363 mwus 0.72093 -11.6364 algorithm, which 0.958333 -11.6367 premises 0.958333 -11.6367 intrinsically 0.818182 -11.6368 eric 0.763158 -11.6368 substrings 0.958333 -11.6371 labor 0.335 -11.6371 ∑ 0.958333 -11.6374 comparable, 0.8 -11.6374 ein 0.958333 -11.6375 higher, 0.289474 -11.6375 approximation 0.958333 -11.6378 artifact 0.892857 -11.6378 ireland 0.866667 -11.6379 situations, 0.573529 -11.638 calculated using 0.923077 -11.6381 determining whether 0.958333 -11.6382 1991; 0.442478 -11.6383 to analyze 0.818182 -11.6384 reservoir 0.573529 -11.6386 their work 0.358382 -11.6388 starts 0.83871 -11.6388 standard deviations 0.8 -11.6392 wp 0.958333 -11.6397 worthwhile 0.923077 -11.6397 organized into 0.958333 -11.6398 presented, 0.417323 -11.6399 creates 0.958333 -11.6401 levels: 0.958333 -11.6402 rulebased 0.958333 -11.6402 boxer 0.32243 -11.6403 neither 0.958333 -11.6404 enterprise 0.731707 -11.6404 0.48 0.958333 -11.6405 included, 0.958333 -11.6406 capitalization, 0.958333 -11.6406 f1, 0.958333 -11.6409 non-asr 0.958333 -11.6409 homogeneous 0.958333 -11.6409 darpa gale 0.958333 -11.6412 data preparation 0.818182 -11.6418 sentiment classification, 0.8 -11.6419 asd 0.187994 -11.642 a simple 0.372671 -11.6421 list. 0.958333 -11.6422 id, 0.958333 -11.6424 joy 0.958333 -11.6425 formulae 0.958333 -11.6425 divides 0.958333 -11.6426 dramatic 0.539474 -11.6431 koo 0.958333 -11.6432 delta 0.958333 -11.6433 monods 0.958333 -11.6435 table 2 lists 0.610169 -11.6435 left-hand 0.105473 -11.6435 second 0.42623 -11.6436 held 0.32243 -11.6437 addressed 0.236181 -11.644 obtained by 0.653846 -11.644 failed to 0.958333 -11.6441 partial parses 0.958333 -11.6441 segments, 0.109043 -11.6443 extracted 0.293436 -11.6444 0. 0.72093 -11.6444 r., 0.958333 -11.6445 advantageous 0.24147 -11.6445 α 0.958333 -11.6446 (v) 0.5 -11.6447 co-reference 0.565217 -11.6448 condition, 0.958333 -11.6449 i−1 0.2173 -11.645 3, 0.958333 -11.6451 spell 0.575758 -11.6451 class label 0.958333 -11.6452 losses 0.892857 -11.6454 bayesian inference 0.892857 -11.6455 prague 0.83871 -11.6456 reported here 0.958333 -11.6458 mediated 0.596774 -11.6458 as explained 0.958333 -11.6459 wildcards 0.958333 -11.6462 heuristics, 0.958333 -11.6463 2.7 0.113755 -11.6464 high 0.958333 -11.6466 o(n2) 0.818182 -11.6466 most appropriate 0.958333 -11.6475 made, 0.892857 -11.6478 nations 0.923077 -11.6478 t) 0.892857 -11.6479 si, 0.958333 -11.6482 abstracts, 0.294118 -11.6484 we perform 0.518072 -11.6486 haghighi 0.641509 -11.6487 two separate 0.731707 -11.6488 averages 0.958333 -11.6489 149 0.958333 -11.6491 explaining 0.958333 -11.6491 question-answering 0.958333 -11.6491 predicative 0.731707 -11.6491 item, 0.958333 -11.6495 166 0.958333 -11.6495 (map) 0.958333 -11.6495 (2011). 0.147783 -11.6495 means 0.274576 -11.6496 dev 0.923077 -11.6496 brazilian 0.958333 -11.6499 171 0.958333 -11.6501 case) 0.547945 -11.6501 sentiwordnet 0.958333 -11.6502 (set 0.958333 -11.6504 1.5% 0.862069 -11.6504 2006, 0.958333 -11.6506 139 0.5 -11.6506 (koehn, 0.763158 -11.6507 confidence. 0.279152 -11.6508 training. 0.958333 -11.6509 assertions 0.958333 -11.6509 october 0.557143 -11.6509 vb 0.818182 -11.651 nomenclature 0.483871 -11.6511 in our experiments, we 0.958333 -11.6515 daughter 0.306383 -11.6516 easy to 0.414062 -11.6516 setting. 0.731707 -11.6517 column shows 0.222472 -11.6518 marked 0.958333 -11.6518 iterating 0.862069 -11.6519 section 5 presents 0.83871 -11.6521 weighting scheme 0.958333 -11.6522 metagrammar 0.958333 -11.6522 attaches 0.457143 -11.6523 requirement 0.133226 -11.6525 among 0.958333 -11.6526 mere 0.958333 -11.6526 2.2.1 0.958333 -11.6528 diagnosis 0.958333 -11.6528 human-like 0.958333 -11.6528 italics 0.888889 -11.653 mn 0.794118 -11.653 to accommodate 0.958333 -11.6531 compromise 0.958333 -11.6532 reimplementation 0.862069 -11.6533 report results on 0.958333 -11.6534 162 0.449541 -11.6534 mitchell 0.268852 -11.6535 community 0.958333 -11.6537 paying 0.794118 -11.6537 appended 0.958333 -11.6538 search engine. 0.958333 -11.6539 point-wise 0.958333 -11.6539 top-10 0.818182 -11.6539 pst 0.565217 -11.654 correspondences 0.272727 -11.6542 (2005) 0.193878 -11.6543 setting 0.958333 -11.6548 illustrative 0.958333 -11.6548 referee 0.345946 -11.6549 (cf. 0.680851 -11.6549 1994) 0.958333 -11.655 scoring, 0.958333 -11.655 downloads 0.958333 -11.655 173 0.557143 -11.6551 roth 0.151295 -11.6552 correlation 0.181682 -11.6553 together 0.862069 -11.6553 extractive summarization 0.958333 -11.6555 constructed, 0.74359 -11.6556 jurafsky, 0.237245 -11.6559 sources 0.376623 -11.6559 curve 0.3125 -11.6564 observations 0.958333 -11.6565 slm 0.958333 -11.6565 simstring 0.270903 -11.6565 were not 0.862069 -11.6567 import 0.958333 -11.6568 although, 0.958333 -11.6568 (because 0.862069 -11.6568 in many cases, 0.251429 -11.6569 different from 0.958333 -11.6573 predictability 0.888889 -11.6576 static cache 0.135495 -11.6581 improvement 0.505882 -11.6583 arise 0.74359 -11.6583 lr 0.625 -11.6583 in this study 0.958333 -11.6584 panel 0.66 -11.6588 spirit 0.958333 -11.659 misunderstanding 0.270903 -11.659 2004) 0.333333 -11.659 scenario 0.958333 -11.6591 i0 0.127393 -11.6592 abstract 0.66 -11.6592 s0 0.92 -11.6593 essentially, 0.283088 -11.6594 successfully 0.862069 -11.6594 repair 0.363636 -11.6597 half of 0.376623 -11.6604 sophisticated 0.414062 -11.6604 if there is 0.614035 -11.6605 url 0.958333 -11.6606 appropriate, 0.625 -11.6607 mixture model 0.74359 -11.6609 located in 0.777778 -11.661 goldwater 0.187801 -11.6611 comparing 0.547945 -11.6613 we keep 0.610169 -11.6613 bi 0.704545 -11.6614 induced from 0.66 -11.6618 transitions. 0.92 -11.6623 wn++ 0.505882 -11.6624 more specific 0.777778 -11.6625 state items 1 -11.6626 reduction. 1 -11.6626 retrieved. 1 -11.6626 internet. 1 -11.6626 scales. 1 -11.6626 explanation. 1 -11.6626 production. 1 -11.6626 assigned. 1 -11.6626 paraphrasing. 1 -11.6626 relationship. 1 -11.6626 beneficial. 1 -11.6626 achieved. 1 -11.6626 viz. 1 -11.6626 cross lingual 1 -11.6626 devices. 1 -11.6626 1990). 1 -11.6626 modifications. 1 -11.6626 rank. 1 -11.6626 counterparts. 1 -11.6626 coreferent. 1 -11.6626 propositions. 1 -11.6626 diversity. 1 -11.6626 investigated. 1 -11.6626 compared. 1 -11.6626 age. 1 -11.6626 versions. 1 -11.6626 detected. 1 -11.6626 screen. 0.74359 -11.6627 relief 0.862069 -11.6628 tweet, 0.92 -11.6629 trade 0.704545 -11.6629 usa. 0.625 -11.6629 guarantees 0.794118 -11.6629 given by: 0.532468 -11.663 each individual 0.83871 -11.6631 earliest 0.6 -11.6631 ranked by 0.731707 -11.6633 infinitive 0.92 -11.6633 simulating 0.956522 -11.6635 if any, 0.74359 -11.6636 parameter, 0.92 -11.6636 empirically, 0.92 -11.6637 156 0.666667 -11.664 abduction 0.862069 -11.6641 sempos 0.406015 -11.6642 expectations 0.956522 -11.6645 tensor product 0.666667 -11.6645 leftmost 0.92 -11.6645 bigram, 0.62963 -11.6645 statistical models 0.92 -11.6647 σ, 0.92 -11.6648 perl 0.92 -11.6648 staff 0.92 -11.665 contextual polarity 0.62963 -11.665 empirical results 0.5 -11.6651 probabilistic model 0.641509 -11.6653 paul 0.704545 -11.6654 removal of 0.83871 -11.6657 s3 0.5 -11.6657 consider the following 0.956522 -11.666 paired t-test 0.92 -11.6665 ave 0.680851 -11.6665 matrix-space 0.525641 -11.6665 annotators were 0.92 -11.6666 extraction: 0.83871 -11.6666 emerging 0.92 -11.6668 c2, 0.83871 -11.6668 indicates whether 0.862069 -11.667 recently proposed 0.92 -11.6671 worse, 0.227488 -11.6672 technology 0.714286 -11.6672 wong 0.83871 -11.6672 sen 0.74359 -11.6675 metaphorical 0.92 -11.6676 insertion, 0.956522 -11.6679 writing style 0.714286 -11.668 questions about 0.257669 -11.6682 the effect of 0.549296 -11.6682 liang 0.956522 -11.6684 turns, 0.19646 -11.6684 object 0.45283 -11.6685 enhance 0.777778 -11.6687 novelty 0.794118 -11.6687 experts, 0.307359 -11.6687 readability 0.92 -11.6689 intel 0.756757 -11.6691 ted 0.92 -11.6692 run, 0.0574018 -11.6692 other 0.888889 -11.6694 people’s 0.888889 -11.6694 replicating 0.614035 -11.6695 monolingual data 0.794118 -11.6697 0.23 0.956522 -11.6698 compressive 0.862069 -11.6699 neg 0.293651 -11.6701 linking 0.92 -11.6703 (feature 0.862069 -11.6704 horn 0.956522 -11.6705 cdec 0.956522 -11.6705 68.0 0.92 -11.6706 syntaxbased 0.66 -11.6706 hu 0.956522 -11.6709 bj¨orne 0.614035 -11.6709 suggestion 0.756757 -11.6709 higher level 0.956522 -11.6712 method’s 0.956522 -11.6712 thus far 0.124026 -11.6713 significant 0.6 -11.6714 match, 0.862069 -11.6715 uppercase 0.688889 -11.6716 vital 0.92 -11.6716 (2010), who 0.956522 -11.6717 caching 0.525641 -11.6717 woz 0.888889 -11.6718 spot 0.956522 -11.6721 inevitably 0.714286 -11.6721 intensity 0.956522 -11.6722 rhyming 0.92 -11.6723 25, 0.29249 -11.6727 iteration 0.956522 -11.673 33.3 0.74359 -11.673 graph construction 0.74359 -11.6731 dropping 0.956522 -11.6733 document-specific 0.590164 -11.6735 solution. 0.92 -11.6735 killed 0.92 -11.6735 dependence on 0.888889 -11.6738 flu 0.888889 -11.674 pbt 0.956522 -11.6741 markedly 0.3 -11.6741 (2004) 0.888889 -11.6744 tamil 0.558824 -11.6745 much lower 0.956522 -11.6746 error propagation 0.956522 -11.6747 nonetheless 0.182236 -11.6747 description 0.956522 -11.6749 aiming at 0.74359 -11.675 promise 0.92 -11.6751 (20) 0.956522 -11.6756 subgraphs 0.956522 -11.6756 faces 0.956522 -11.6759 ought 0.888889 -11.6759 su 0.956522 -11.6763 bve 0.92 -11.6765 (t 0.92 -11.6766 audience 0.888889 -11.6767 intensive 0.92 -11.6768 investigation. 0.956522 -11.6769 matchings 0.956522 -11.6772 inspecting 0.756757 -11.6772 in summary, 0.549296 -11.6773 communications 0.956522 -11.6774 today, 0.956522 -11.6778 built, 0.956522 -11.678 fruitful 0.888889 -11.6782 everyone 0.956522 -11.6783 characteristics, 0.714286 -11.6783 dictionaries, 0.166453 -11.6783 component 0.230769 -11.6785 item 0.956522 -11.6787 miscellaneous 0.956522 -11.6787 relative clause 0.473684 -11.6787 augment 0.956522 -11.679 az 0.92 -11.679 147 0.525641 -11.679 flow 0.888889 -11.6792 pitman-yor 0.207171 -11.6792 slightly 0.956522 -11.6793 tree-based 0.956522 -11.6794 te, 0.590164 -11.6795 dependence 0.390071 -11.6795 operator 0.956522 -11.6796 syntactic/semantic 0.956522 -11.6796 depict 0.8125 -11.6796 2007, 0.956522 -11.68 indomain 0.725 -11.68 106 0.394161 -11.6801 at the end 0.956522 -11.6802 publish 0.8125 -11.6802 7) 0.725 -11.6803 vanderwende, 0.888889 -11.6805 69.2 0.956522 -11.6806 symmetrized 0.756757 -11.6807 numerical expressions 0.221461 -11.6808 contained 0.956522 -11.6809 vanilla 0.956522 -11.6811 random, 0.956522 -11.6811 other’s 0.956522 -11.6812 18% 0.666667 -11.6812 0.5. 0.603448 -11.6812 more frequent 0.956522 -11.6814 ultimate 0.578125 -11.6814 limiting 0.725 -11.6814 avenue 0.0960549 -11.6814 we have 0.833333 -11.6817 mln 0.956522 -11.6818 boosts 0.956522 -11.6819 follow-up 0.666667 -11.6819 mccallum, 0.666667 -11.6819 personality 0.956522 -11.6821 rescore 0.956522 -11.6822 compete 0.74359 -11.6822 adjunct 0.714286 -11.6823 path, 0.92 -11.6823 wordnet synset 0.569231 -11.6824 principled 0.888889 -11.6826 visualisation 0.369427 -11.6826 are given in 0.956522 -11.6827 framenet, 0.956522 -11.6828 190 0.956522 -11.6828 collapse 0.888889 -11.6829 description length 0.956522 -11.683 (charniak 0.956522 -11.6831 descendants 0.794118 -11.6831 eye 0.8125 -11.6831 filtering, 0.956522 -11.6833 reading-style 0.603448 -11.6833 posted 0.421488 -11.6834 machines 0.888889 -11.6834 requirements. 0.771429 -11.6838 specially 0.956522 -11.6839 accompanying 0.956522 -11.6839 hidden variables 0.956522 -11.6839 serving 0.888889 -11.684 heard 0.956522 -11.6845 titov 0.688889 -11.6845 aligns 0.714286 -11.6847 that incorporates 0.956522 -11.6848 (top) 0.956522 -11.6849 p< 0.956522 -11.6849 alpino 0.956522 -11.6852 sole 0.956522 -11.6852 most prominent 0.956522 -11.6854 analogy 0.956522 -11.6861 67.8 0.956522 -11.6862 matroid 0.956522 -11.6862 pruning, 0.8125 -11.6863 pushed 0.771429 -11.6864 bucket 0.956522 -11.6865 variations, 0.956522 -11.6865 tions 0.956522 -11.6865 background, 0.956522 -11.6867 cns 0.956522 -11.6868 customization 0.956522 -11.6871 influencing 0.956522 -11.6873 conveying 0.771429 -11.6873 anger 0.833333 -11.6874 marton 0.8125 -11.6874 press, 0.956522 -11.6876 3d 0.956522 -11.6876 instances) 0.956522 -11.6876 215 0.956522 -11.6876 arrangement 0.8125 -11.6876 ibm-1 0.956522 -11.6877 maximized 0.956522 -11.6877 forest-based 0.135813 -11.6877 (1) 0.956522 -11.6879 fundamentally 0.5125 -11.6884 precision recall 0.92 -11.6884 (out 0.116456 -11.6884 classifier 0.956522 -11.6886 skills 0.956522 -11.6886 abstract, 0.618182 -11.6887 city, 0.688889 -11.6888 an action 0.956522 -11.6889 emotions, 0.756757 -11.6889 in bold 0.321951 -11.689 type, 0.329897 -11.6891 practice 0.117647 -11.6892 less 0.771429 -11.6892 degrade 0.673913 -11.6896 disagree 0.956522 -11.69 argues 0.833333 -11.69 nps, 0.956522 -11.6901 (1998), 0.653061 -11.6901 situation, 0.857143 -11.6904 width 0.956522 -11.6904 speedup 0.409449 -11.6904 inclusion 0.956522 -11.6906 loading 0.956522 -11.6906 foot 0.956522 -11.6912 text; 0.590164 -11.6913 argumentation 0.956522 -11.6914 i: 0.956522 -11.6917 conflicts 0.956522 -11.6917 word-by-word 0.956522 -11.692 synset, 0.956522 -11.692 (t) 0.956522 -11.6923 (2 0.956522 -11.6925 functionalities 0.756757 -11.6925 we have chosen 0.956522 -11.6926 services, 0.956522 -11.6926 taxonomic 0.956522 -11.6926 sliding 0.634615 -11.6928 unlabeled data. 0.956522 -11.6929 msr 0.526316 -11.693 similarity score 0.956522 -11.6931 tam 0.956522 -11.6932 0.98 0.46875 -11.6932 i) 0.603448 -11.6933 starting point 0.833333 -11.6933 0.44 0.666667 -11.6934 program, 0.473684 -11.6934 concerns 0.956522 -11.6938 permuted 0.956522 -11.6938 sparse, 0.956522 -11.6938 rumors 0.725 -11.6938 are. 0.956522 -11.6941 porter 0.956522 -11.6944 renders 0.308036 -11.6945 assumed 0.857143 -11.6946 v2 0.956522 -11.6947 unpruned 0.455446 -11.6948 to remove 0.956522 -11.6949 avenues 0.857143 -11.695 ξ 0.8125 -11.695 protein-protein 0.426087 -11.6952 identity 0.8125 -11.6952 iterated 0.956522 -11.6953 subtracting 0.409449 -11.6953 they can be 0.956522 -11.6955 k2 0.956522 -11.6956 metonymy 0.956522 -11.6958 sptk 0.857143 -11.6959 (english) 0.956522 -11.6959 q-queries 0.956522 -11.6959 accessed 0.956522 -11.6961 young, 0.956522 -11.6965 that’s 0.956522 -11.6967 evaluated, 0.956522 -11.6967 griffiths 0.194643 -11.6967 percentage 0.298319 -11.6967 choice of 0.310502 -11.6968 material 0.956522 -11.6971 260 0.956522 -11.6973 phrases) 0.956522 -11.6973 seminar 0.8125 -11.6975 projectable 0.956522 -11.6977 subclasses 0.618182 -11.6977 tesla-m 0.956522 -11.698 humans, 0.956522 -11.698 editorial 0.59322 -11.6981 frequencies. 0.956522 -11.6982 01 0.92 -11.6982 traverse 0.787879 -11.6987 to cope 0.956522 -11.6989 sv 0.956522 -11.6989 optimization, 0.707317 -11.699 de la 0.653061 -11.699 separately, 0.956522 -11.6991 lattices, 0.8125 -11.6992 jordan 0.154106 -11.6993 parameter 0.473118 -11.6994 question, 0.956522 -11.6995 170 0.956522 -11.6997 efficiency, 0.272085 -11.6998 are used to 0.92 -11.6999 (wilson et 0.771429 -11.6999 lp-mert 0.8125 -11.6999 inquiry 0.8125 -11.7001 managed 0.338798 -11.7002 antecedent 0.956522 -11.7004 conditions: 0.884615 -11.7008 sentence-aligned 0.8125 -11.701 distractors 0.37931 -11.7014 consequently, 0.956522 -11.7015 land 0.956522 -11.7016 labels: 0.494118 -11.7016 not included 0.263158 -11.7017 synset 0.771429 -11.7017 sample. 0.956522 -11.7019 benchmarks 0.321782 -11.702 cell 0.519481 -11.7021 to align 0.956522 -11.7025 least, 0.473118 -11.7026 ” 0.519481 -11.7026 f-scores 0.956522 -11.7027 poss 0.653061 -11.7027 priority 0.46875 -11.7032 we implemented 0.787879 -11.7033 pivot translation 0.569231 -11.7036 to disambiguate 0.884615 -11.7037 .30 0.956522 -11.7038 rand 0.358025 -11.7038 listed in 0.259615 -11.7038 parsing. 0.956522 -11.7039 increase, 0.956522 -11.7039 social power 0.673913 -11.7039 propagate 0.367742 -11.7041 correspondence 0.673913 -11.7042 etc., 0.64 -11.7042 0.74 0.571429 -11.7043 corpus: 0.580645 -11.7044 grammars, 0.237838 -11.7045 used. 0.64 -11.7047 targets. 0.857143 -11.7048 nivre, 0.725 -11.7048 −1 0.550725 -11.7049 was developed 0.956522 -11.705 adjuncts 0.956522 -11.705 stemmer 0.13012 -11.705 argument 0.857143 -11.7051 [...] 0.707317 -11.7051 is available at 0.884615 -11.7051 69.9 0.833333 -11.7051 (x,y) 0.560606 -11.7051 operates 0.956522 -11.7053 (instead of 0.956522 -11.7053 127 0.884615 -11.7054 google translate 0.436364 -11.7055 inputs 0.64 -11.7055 was designed 0.956522 -11.7056 look up 0.697674 -11.7056 dividing 0.771429 -11.7056 conflict 0.506173 -11.7057 lists. 0.916667 -11.7057 child’s 0.916667 -11.7057 (recall 0.659574 -11.706 command 0.241667 -11.7061 needs 0.884615 -11.7063 discrepancies 0.956522 -11.7065 patb 0.956522 -11.7065 (they 0.916667 -11.7067 gray 0.560606 -11.7067 repository 0.333333 -11.7068 gibbs 0.956522 -11.7069 bodies 0.64 -11.7072 more informative 0.857143 -11.7074 role labeling. 0.916667 -11.7075 pretty 0.916667 -11.7076 overt 0.659574 -11.7077 runs. 0.884615 -11.7079 duffy, 0.916667 -11.7079 mixtures 0.64 -11.7082 ct 0.857143 -11.7083 hyponym 0.653061 -11.7084 israel 0.916667 -11.7085 relatively few 0.210084 -11.7085 ratio 0.707317 -11.7086 spaces. 0.884615 -11.7086 window-based 0.736842 -11.7087 cats 0.175516 -11.7088 get 0.916667 -11.7089 1m 0.173661 -11.709 relatively 1 -11.7091 latter. 1 -11.7091 estimate. 1 -11.7091 tac 2010 workshop. 1 -11.7091 subsets. 1 -11.7091 methodology. 1 -11.7091 (1997). 1 -11.7091 paradigms. 1 -11.7091 design. 1 -11.7091 external referees 1 -11.7091 site. 1 -11.7091 negation. 1 -11.7091 life. 1 -11.7091 1988). 1 -11.7091 chunking. 1 -11.7091 perplexity. 1 -11.7091 collocations. 1 -11.7091 complete. 1 -11.7091 700 800 1 -11.7091 holds. 1 -11.7091 science. 1 -11.7091 trials. 1 -11.7091 bacillus subtilis 1 -11.7091 request. 1 -11.7091 book. 1 -11.7091 services. 1 -11.7091 xi. 1 -11.7091 srl. 1 -11.7091 satisfied. 1 -11.7091 assumptions. 1 -11.7091 mentioned. 1 -11.7091 coarse grained 1 -11.7091 shift. 0.916667 -11.7092 namely: 0.954545 -11.7092 website. 0.455446 -11.7094 connecting 0.954545 -11.7094 trento, italy 0.954545 -11.7095 morphological analyser 0.833333 -11.71 estimator 0.916667 -11.7101 part-of-speech tag 0.954545 -11.7101 open ie 0.681818 -11.7102 mcclosky 0.143849 -11.7103 directly 0.681818 -11.7103 translator 0.954545 -11.7104 edit distance, 0.833333 -11.7104 notation. 0.634615 -11.7105 rb 0.916667 -11.7111 4.3.1 0.954545 -11.7111 machine-learning 0.954545 -11.7112 tsujii, 0.4 -11.7113 prior to 0.806452 -11.7114 structured role 0.857143 -11.7115 201 0.328125 -11.7117 track 0.542857 -11.712 analogous 0.0675087 -11.712 over 0.771429 -11.7121 convergence, 0.954545 -11.7121 bias towards 0.857143 -11.7122 penalty, 0.519481 -11.7122 these features are 0.806452 -11.7123 0.63 0.954545 -11.7124 expressive power 0.833333 -11.7125 ibm models 0.884615 -11.7127 commitment 0.787879 -11.7127 revised 0.954545 -11.7127 11. 0.673913 -11.7128 z, 0.653061 -11.7128 are consistent with 0.833333 -11.7129 imputed 0.916667 -11.713 simplified version 0.583333 -11.7131 strings. 0.64 -11.7132 em. 0.884615 -11.7134 bibliome 0.141618 -11.7135 experimental 0.243553 -11.7136 importance 0.560606 -11.7136 directly. 0.2 -11.7136 third 0.884615 -11.7139 aql 0.954545 -11.7139 platforms 0.833333 -11.7139 structuring 0.954545 -11.7141 unavailable 0.857143 -11.7142 typology 0.857143 -11.7145 lexical entry 0.707317 -11.7145 1992; 0.64 -11.7145 table 6 shows 0.916667 -11.7147 suffix, 0.916667 -11.7149 enumerating 0.681818 -11.7152 (m) 0.787879 -11.7154 wij 0.916667 -11.7155 clarity 0.210638 -11.7156 frame 0.857143 -11.7157 dataset consists of 0.806452 -11.7158 tsl 0.954545 -11.7158 chances 0.916667 -11.7159 cyk 0.954545 -11.7159 sparser 0.494118 -11.716 th 0.681818 -11.7164 factor, 0.22549 -11.7166 demonstrate 0.681818 -11.7166 for our experiments, 0.787879 -11.7166 misc 0.534247 -11.7168 neighbors 0.527027 -11.7173 which includes 0.806452 -11.7174 85% 0.64 -11.7175 parenleftbig 0.884615 -11.7175 y 0.806452 -11.7177 high confidence 0.286853 -11.7179 assignment 0.787879 -11.7179 backward language 0.349112 -11.7185 views 0.428571 -11.7185 or even 0.736842 -11.7186 automatic metrics 0.884615 -11.7188 intertranscriber 0.787879 -11.7188 rewards 0.916667 -11.7188 alors 0.954545 -11.719 jelinek, 0.954545 -11.719 discussions, 0.916667 -11.7191 pj 0.954545 -11.7191 formulating 0.954545 -11.7196 61% 0.20985 -11.72 factors 0.75 -11.7201 understanding. 0.916667 -11.7203 motivations 0.954545 -11.7204 documentation 0.954545 -11.7204 bayes, 0.954545 -11.7204 learners, 0.954545 -11.7204 choices, 0.884615 -11.7206 election 0.136775 -11.7206 shared 0.75 -11.721 preliminary results 0.954545 -11.721 dates, 0.954545 -11.721 elaborated 0.707317 -11.7211 important, 0.736842 -11.7211 and pereira, 0.954545 -11.7211 archives 0.75 -11.7212 system architecture 0.787879 -11.7212 calibrated 0.806452 -11.7213 ak 0.717949 -11.7214 future research 0.954545 -11.7214 traits 0.707317 -11.7215 manually. 0.954545 -11.7217 error rate. 0.954545 -11.7219 manipulation 0.607143 -11.7223 i will 0.954545 -11.7224 moore, 0.954545 -11.7224 codes, 0.884615 -11.7225 differentiating 0.596491 -11.723 (nivre et 0.954545 -11.723 tedious 0.954545 -11.7231 nissim 0.954545 -11.7231 explicitly, 0.884615 -11.7232 anchored 0.377622 -11.7236 an entity 0.884615 -11.7236 red, 0.954545 -11.7236 line) 0.857143 -11.7237 elicit 0.717949 -11.7237 anaphoricity 0.5 -11.7237 high precision 0.717949 -11.7238 dr. 0.787879 -11.7239 4we 0.954545 -11.7239 populate 0.954545 -11.7239 door 0.916667 -11.724 omitting 0.954545 -11.724 75.5 0.954545 -11.724 overtly 0.736842 -11.7241 song 0.806452 -11.7241 predominant 0.857143 -11.7242 80.0 0.954545 -11.7242 method: 0.857143 -11.7244 harvest 0.954545 -11.7244 high dimensional 0.827586 -11.7245 most cases, 0.527027 -11.7246 weka 0.368421 -11.7246 bit 0.954545 -11.7247 special characters 0.954545 -11.7248 tes 0.916667 -11.725 language modeling. 0.954545 -11.7253 assessment, 0.954545 -11.7254 high-frequency 0.916667 -11.7255 t∈t 0.954545 -11.7256 83.2 0.512821 -11.7259 null1 0.954545 -11.726 rebuttal 0.954545 -11.7262 speculate 0.954545 -11.7262 pragmatics 0.954545 -11.7262 populated 0.488372 -11.7263 a maximum entropy 0.954545 -11.7264 opposing 0.954545 -11.7264 excerpts 0.954545 -11.7264 inter-rater 0.884615 -11.7265 yarowsky 0.954545 -11.7268 behavioral 0.707317 -11.7269 dependency parser. 0.954545 -11.727 listening 0.916667 -11.7271 content shift 0.954545 -11.7273 kaf 0.884615 -11.7275 setups 0.253125 -11.7276 objects 0.954545 -11.7276 populations 0.954545 -11.7276 parsim 0.954545 -11.7276 traversing 0.954545 -11.7276 bacterial 0.954545 -11.7277 constant, 0.954545 -11.7281 successor 0.75 -11.7282 bush 0.916667 -11.7283 examples: 0.827586 -11.7284 conducted experiments 0.243478 -11.7285 (as 0.954545 -11.7287 loopy belief 0.544118 -11.7288 complexity, 0.954545 -11.729 dissimilarity 0.954545 -11.7293 kernel, 0.954545 -11.7294 below) 0.954545 -11.7296 arising from 0.954545 -11.7296 theirs 0.0535889 -11.7297 only 0.827586 -11.7297 reference translation. 0.954545 -11.7298 corect 0.335165 -11.7299 chunk 0.806452 -11.73 )) 0.52 -11.73 tagger, 0.75 -11.7302 gb 0.477778 -11.7303 2001), 0.954545 -11.7304 nation 0.954545 -11.7304 (mann 0.596491 -11.7305 words: 0.916667 -11.7305 symposium 0.954545 -11.7305 kernels, 0.827586 -11.7306 enriched with 0.827586 -11.7306 day, 0.717949 -11.7307 3.2, 0.787879 -11.7307 festival 0.916667 -11.7307 vextendsinglevextendsingle 0.954545 -11.7307 prosodic breaks 0.954545 -11.7308 compactness 0.75 -11.7309 0.76 0.221687 -11.731 assume that 0.477778 -11.731 berkeley parser 0.116634 -11.7311 us 0.827586 -11.7312 psychosentiwordnet 0.954545 -11.7313 cuts 0.482759 -11.7314 could also 0.384058 -11.7316 whenever 0.197343 -11.7316 generative 0.954545 -11.7316 fitted 0.806452 -11.7317 30k 0.104952 -11.732 search 0.827586 -11.732 terp 0.681818 -11.7321 problem: 0.369128 -11.7322 grammar, 0.884615 -11.7324 two-tailed 0.954545 -11.7324 66.7 0.954545 -11.7325 steedman 0.231579 -11.7327 the remaining 0.26087 -11.7327 exploit 0.884615 -11.7328 control, 0.787879 -11.733 2011 shared task 0.954545 -11.733 q&a 0.954545 -11.733 switzerland 0.482759 -11.733 agreement between 0.954545 -11.7333 paraphrastic 0.681818 -11.7334 once. 0.954545 -11.7335 z-mert 0.954545 -11.7335 firm 0.954545 -11.7335 ontologies, 0.666667 -11.7339 effects on 0.954545 -11.7339 reproduced 0.954545 -11.7339 8.2 0.954545 -11.7339 lda-based 0.513158 -11.734 maximization 0.627451 -11.734 faced 0.806452 -11.734 incident 0.884615 -11.7341 lexicalised 0.954545 -11.7341 spend 0.954545 -11.7342 @x 0.954545 -11.7342 coupling 0.954545 -11.7342 terminology, 0.764706 -11.7344 park 0.0648949 -11.7344 the same 0.52 -11.7344 developed by 0.954545 -11.7346 thrax 0.5625 -11.7347 thing 0.527778 -11.7347 training instances 0.681818 -11.7347 acc. 0.438095 -11.7347 revision 0.954545 -11.7347 chain, 0.954545 -11.7347 participates 0.954545 -11.7347 surveys 0.954545 -11.7347 pcfgs 0.75 -11.7349 finish 0.596491 -11.7349 11: 0.26 -11.735 2002). 0.266667 -11.7351 variant 0.884615 -11.7354 antonymous 0.493976 -11.7355 tagged as 0.954545 -11.7356 annotators agreed 0.137291 -11.7358 relevant 0.764706 -11.7361 prior. 0.645833 -11.7361 prefers 0.954545 -11.7361 cyclic 0.954545 -11.7361 entering 0.183471 -11.7362 randomly 0.488095 -11.7363 growth 0.137546 -11.7364 ) 0.954545 -11.7367 tackling 0.954545 -11.737 reconstructing 0.954545 -11.7372 trained, 0.954545 -11.7372 diagrams 0.884615 -11.7373 idiom 0.954545 -11.7374 clear, 0.954545 -11.7375 consult 0.884615 -11.7377 converged 0.806452 -11.7377 as before, 0.645833 -11.7377 distinguishes 0.954545 -11.7377 literacy 0.764706 -11.7378 transition matrix 0.954545 -11.7378 releases 0.954545 -11.7378 extrinsic evaluation 0.954545 -11.7378 concluded that 0.806452 -11.7381 svd 0.421053 -11.7383 our method is 0.954545 -11.7383 6the 0.827586 -11.7383 jason 0.884615 -11.7384 dobj 0.851852 -11.7384 dg 0.954545 -11.7385 among others. 0.954545 -11.7385 68.7 0.827586 -11.7385 tightly 0.666667 -11.7388 watson 0.954545 -11.7388 2for 0.343023 -11.7389 integer 0.764706 -11.7389 rush 0.666667 -11.7389 visually 0.645833 -11.7391 interactions. 0.262799 -11.7394 λ 0.954545 -11.7394 duplicates 0.954545 -11.7394 triggers, 0.851852 -11.7394 ' 0.690476 -11.7395 previous approaches 0.5 -11.7395 generally, 0.954545 -11.7397 commissive 0.536232 -11.7399 source sentence. 0.954545 -11.7399 interruptions 0.358974 -11.74 research on 0.535211 -11.7402 lexical, 0.954545 -11.7402 attribute, 0.954545 -11.7403 communicative functions 0.433962 -11.7407 member 0.884615 -11.7407 transcripts. 0.827586 -11.7408 conversations, 0.717949 -11.7409 two major 0.764706 -11.741 fixation 0.954545 -11.7413 god 0.827586 -11.7413 0.51 0.645833 -11.7418 relate to 0.536232 -11.7419 centroid 0.954545 -11.742 non-linguistic 0.57377 -11.7421 character-level 0.954545 -11.7422 designated 0.954545 -11.7425 pi, 0.916667 -11.7426 opinion. 0.954545 -11.743 repeats 0.954545 -11.743 resorting 0.806452 -11.7431 medieval 0.954545 -11.7436 sometimes, 0.851852 -11.7437 ney 0.4 -11.7439 realized 0.954545 -11.7439 analysing 0.0511232 -11.744 also 0.806452 -11.744 boy 0.851852 -11.7441 b+l 0.954545 -11.7444 answers, 0.954545 -11.7444 information graphics 0.764706 -11.7445 speakers’ 0.319797 -11.7446 format 0.47191 -11.7447 0.60 0.954545 -11.7447 coordination, 0.954545 -11.7449 grounds 0.954545 -11.7449 pauses 0.851852 -11.7449 may seem 0.954545 -11.745 reports, 0.954545 -11.7452 stepwise 0.954545 -11.7452 modes 0.954545 -11.7457 hpb 0.764706 -11.7458 networks, 0.954545 -11.7458 cps 0.827586 -11.7459 language’s 0.954545 -11.7461 146 0.235457 -11.7461 3.3 0.88 -11.7461 wrong. 0.851852 -11.7461 eliciting 0.954545 -11.7464 xing 0.954545 -11.7464 consequent 0.954545 -11.7466 consider. 0.954545 -11.7466 axioms, 0.954545 -11.7468 multi-domain 0.954545 -11.7469 joining 0.954545 -11.7469 lucene 0.57377 -11.7469 α = 0.954545 -11.7471 corpusbased 0.827586 -11.7472 bioscope 0.954545 -11.7474 10th 0.851852 -11.7477 (marked 0.645833 -11.7478 multiplication 0.954545 -11.748 message, 0.954545 -11.7482 experience, 0.78125 -11.7483 phrase pairs, 0.6 -11.7483 crf model 0.396825 -11.7485 incoming 0.827586 -11.7485 mouse 0.851852 -11.7486 of: 0.674419 -11.7488 analyzes 0.954545 -11.7493 wrong, 0.954545 -11.7493 sov 0.632653 -11.7493 our method. 0.47191 -11.7494 recall that 0.954545 -11.7494 differential 0.627451 -11.7495 charles 0.916667 -11.7496 reorderings. 0.916667 -11.7496 frequent term 0.690476 -11.7497 wanted to 0.144341 -11.7498 w 0.954545 -11.7499 (2010b) 0.7 -11.7501 english-french 0.433962 -11.7501 aligning 0.954545 -11.7502 7we 0.72973 -11.7505 combinatory 0.138756 -11.7505 here 0.954545 -11.7507 calling 0.312195 -11.7508 enable 0.72973 -11.7509 sports 0.78125 -11.7512 pc 0.645833 -11.7512 required. 0.954545 -11.7513 trusted 0.717949 -11.7515 propagation. 0.954545 -11.7515 categorizing 0.88 -11.7516 language modelling 0.851852 -11.7517 .33 0.277992 -11.7519 annotate 0.545455 -11.7519 stimulus 0.827586 -11.7519 courses 0.954545 -11.752 beats 0.286885 -11.7521 worse 0.954545 -11.7523 posterior regularization 0.851852 -11.7524 mrf 0.513158 -11.7527 differences are 0.954545 -11.7528 5.2.1 0.78125 -11.7529 0.94 0.576271 -11.7529 lookahead 0.954545 -11.7532 corrective 0.954545 -11.7534 sentiments. 0.851852 -11.7534 huber 0.913043 -11.7537 developed, 0.0607847 -11.754 semantic 0.827586 -11.754 (ilp) 0.913043 -11.7542 left-most 0.954545 -11.7542 character n-grams 0.0810367 -11.7544 best 0.35443 -11.7546 alignments. 0.172619 -11.7547 empirical 0.954545 -11.7548 attained 0.913043 -11.7549 163 0.72973 -11.7549 5), 0.493827 -11.755 but, 0.347561 -11.7551 scored 0.466667 -11.7551 throughout the 0.764706 -11.7554 1990) 0.827586 -11.7556 was originally 0.764706 -11.7557 isi 0.913043 -11.7557 said, 0.586207 -11.7558 5-fold 0.188277 -11.7558 correctly 0.72973 -11.7559 domain specific 0.341176 -11.7562 each other 0.88 -11.7566 an opportunity 0.913043 -11.7568 3-grams 0.396825 -11.7571 ways of 0.0502208 -11.7578 features 0.742857 -11.7579 edge weights 0.324468 -11.7579 implies 1 -11.7579 library. 1 -11.7579 t2. 1 -11.7579 base gaz 1 -11.7579 focus. 1 -11.7579 scripts. 1 -11.7579 rankings. 1 -11.7579 threads. 1 -11.7579 γ. 1 -11.7579 ontonotes. 1 -11.7579 valid. 1 -11.7579 23. 1 -11.7579 state-of-the 1 -11.7579 top. 1 -11.7579 expectations. 1 -11.7579 2.3. 1 -11.7579 1991). 1 -11.7579 play. 1 -11.7579 entropy. 1 -11.7579 reaction. 1 -11.7579 explicitly. 1 -11.7579 need. 1 -11.7579 section). 1 -11.7579 β. 1 -11.7579 phases. 1 -11.7579 transformation. 1 -11.7579 (5). 1 -11.7579 approximation. 1 -11.7579 activity. 1 -11.7579 done. 1 -11.7579 tested. 1 -11.7579 asr. 1 -11.7579 sparseness. 1 -11.7579 outcomes. 1 -11.7579 impact. 1 -11.7579 voice. 1 -11.7579 processed. 1 -11.7579 (1995). 1 -11.7579 none. 1 -11.7579 survey. 0.545455 -11.758 dog 0.764706 -11.7581 claire 0.952381 -11.7581 co-reference resolution 0.952381 -11.7581 misleading 0.952381 -11.7581 decomposition. 0.952381 -11.7581 vroomen 0.114851 -11.7584 higher 0.462366 -11.7584 are listed 0.186736 -11.7586 expression 0.913043 -11.7587 biasing 0.493827 -11.7592 (2005), 0.553846 -11.7593 geometric 0.764706 -11.7594 ambiguities. 0.827586 -11.7594 wam 0.913043 -11.7595 architecture, 0.143299 -11.7595 maximum 0.337143 -11.7595 are considered 0.493827 -11.7597 differentiate 0.952381 -11.7598 majority vote 0.952381 -11.76 suzuki 0.952381 -11.7603 (wu, 0.952381 -11.7604 gaussian mixture 0.764706 -11.7605 coreference. 0.742857 -11.7607 we can compute 0.88 -11.7607 king 0.952381 -11.7608 inner product 0.246914 -11.7609 involved 0.482353 -11.7611 exploits 0.952381 -11.7611 carrying out 0.952381 -11.7611 rdm 0.913043 -11.7615 1t 0.72973 -11.7615 mountain 0.952381 -11.7616 .26 0.742857 -11.7619 produced. 0.952381 -11.7622 boosting 0.913043 -11.7624 sense disambiguation, 0.952381 -11.7625 performed, 0.255814 -11.7626 this paper is 0.952381 -11.763 tag suggestion 0.214953 -11.763 entropy 0.425926 -11.7631 0.9 0.952381 -11.7632 fed into 0.952381 -11.7632 column) 0.324468 -11.7634 support vector 0.589286 -11.7634 dialogues. 0.952381 -11.7635 maccartney 0.8 -11.7636 rbmt 0.5 -11.7636 figure 3 shows 0.764706 -11.7638 w.r.t. 0.827586 -11.7642 communities. 0.952381 -11.7643 69% 0.913043 -11.7646 this: 0.576271 -11.7647 branching 0.8 -11.7647 (although 0.72973 -11.765 c., 0.88 -11.765 retains 0.913043 -11.7652 tobi 0.506494 -11.7652 to reflect 0.107855 -11.7654 s 0.952381 -11.7654 oov detection 0.913043 -11.7655 child, 0.78125 -11.7656 an internal 0.476744 -11.7656 (2004), 0.952381 -11.7656 by, 0.0987167 -11.7659 input 0.5 -11.7659 acknowledge 0.952381 -11.766 misclassifications 0.952381 -11.7662 lectures 0.201646 -11.7663 we define 0.952381 -11.7664 higher-order 0.520548 -11.7666 more effective 0.952381 -11.7668 row/column 0.851852 -11.767 enumeration 0.173252 -11.7671 models, 0.72973 -11.7672 details about 0.88 -11.7672 mitchell, 0.307692 -11.7675 no. 0.913043 -11.7675 lake 0.72973 -11.7677 melm 0.952381 -11.7678 blei 0.952381 -11.7678 lee said 0.432692 -11.7678 dependency parser 0.88 -11.7678 higher-level 0.952381 -11.768 colour, 0.383459 -11.768 we obtained 0.952381 -11.7681 kaplan-meier 0.545455 -11.7681 dependency parse 0.652174 -11.7681 base, 0.506494 -11.7682 neural 0.742857 -11.7683 correct answers 0.710526 -11.7684 sentiment polarity 0.952381 -11.7686 elhadad, 0.952381 -11.7686 (hmm) 0.461538 -11.7687 were asked 0.952381 -11.7688 yarowsky, 0.952381 -11.7689 verb’s 0.952381 -11.7689 cohn, 0.952381 -11.7689 content assessment 0.603774 -11.7691 i-th 0.952381 -11.7691 soul 0.125701 -11.7691 meaning 0.324324 -11.7691 to define 0.288136 -11.7691 to investigate 0.952381 -11.7692 versa 0.444444 -11.7693 thematic 0.157828 -11.7697 therefore, 0.88 -11.7698 rerank 0.952381 -11.77 any) 0.88 -11.7701 0.43 0.952381 -11.7702 corank 0.88 -11.7703 f0 0.952381 -11.7704 suffered 0.952381 -11.7705 feed 0.913043 -11.7707 emoticon 0.952381 -11.7707 conversion, 0.88 -11.7707 mcnemar’s 0.952381 -11.771 dozen 0.952381 -11.771 dumais, 0.227513 -11.771 denotes 0.638298 -11.7711 precedes 0.8 -11.7711 coreferences 0.952381 -11.7712 (bj¨orne 0.952381 -11.7713 trials, 0.952381 -11.7713 pilot study 0.952381 -11.7713 appointment 0.8 -11.7713 calculations 0.952381 -11.7715 hyperlink 0.952381 -11.7715 14% 0.952381 -11.7715 levy 0.576271 -11.7718 0.79 0.487805 -11.7718 stand 0.177706 -11.7722 authors 0.821429 -11.7724 human judgments. 0.952381 -11.7725 76.4 0.952381 -11.7726 communities, 0.952381 -11.7726 66.0 0.603774 -11.7727 correlation coefficient 0.952381 -11.7728 52% 0.521127 -11.7729 not just 0.150987 -11.7729 who 0.952381 -11.7733 run-time 0.471264 -11.7734 properties. 0.913043 -11.7735 meaning representations 0.913043 -11.7735 (i, 0.952381 -11.7736 orthography 0.632653 -11.7736 searn 0.528571 -11.7737 target, 0.952381 -11.7739 collocations, 0.952381 -11.7739 classification-based 0.952381 -11.7742 laisr 0.576271 -11.7743 different. 0.913043 -11.7744 semi-automatic 0.952381 -11.7744 noncompositional 0.952381 -11.7744 assessors 0.8 -11.7744 opinion detection 0.952381 -11.7746 88.5 0.821429 -11.7748 ko 0.952381 -11.7749 assesses 0.88 -11.7749 182 0.952381 -11.775 widely-used 0.952381 -11.7752 iii, 0.652174 -11.7753 selected, 0.281633 -11.7754 we can see 0.952381 -11.7754 most importantly, 0.952381 -11.7755 62.5 0.952381 -11.7755 inability 0.603774 -11.7757 to determine which 0.952381 -11.7758 32.2 0.952381 -11.7758 procedure: 0.682927 -11.776 relationships. 0.8 -11.776 dual supervision 0.167873 -11.7762 add 0.757576 -11.7762 ∨ 0.913043 -11.7763 where: 0.913043 -11.7763 stacks 0.821429 -11.7763 syntactic processing 0.555556 -11.7765 an exact 0.952381 -11.7765 task) 0.952381 -11.7765 optimally 0.913043 -11.7766 explored. 0.0744199 -11.7767 evaluation 0.757576 -11.7767 jeffrey 0.952381 -11.7768 64.2 0.88 -11.7768 (mi 0.952381 -11.7773 cost-sensitive 0.952381 -11.7775 succeed 0.521127 -11.7776 chiang, 0.851852 -11.7776 very short 0.422018 -11.7777 u.s. 0.555556 -11.7778 condition. 0.952381 -11.7779 71.4 0.952381 -11.7781 confidence-weighted 0.952381 -11.7781 progressively 0.218137 -11.7784 retrieval 0.851852 -11.7784 left. 0.851852 -11.7784 helpfulness. 0.8 -11.7785 was conducted 0.452632 -11.7786 5.4 0.952381 -11.7788 (walker 0.88 -11.7788 similes 0.952381 -11.7789 type-based sampling 0.952381 -11.7789 emerge 0.952381 -11.7789 german-to-english 0.652174 -11.7791 instantiate 0.78125 -11.7794 runs, 0.952381 -11.7794 rates, 0.821429 -11.7794 glue rules 0.371429 -11.7795 la 0.952381 -11.7796 (he 0.8 -11.7796 word pairs, 0.952381 -11.7797 50k 0.193798 -11.7798 variable 0.292035 -11.78 depth 0.471264 -11.78 or, 0.0913749 -11.7802 related 0.952381 -11.7805 s1, 0.197211 -11.7807 (section 0.913043 -11.7809 during decoding 0.603774 -11.781 out, 0.161725 -11.781 we propose 0.952381 -11.781 (mccallum 0.952381 -11.781 voca 0.952381 -11.781 responsibility 0.682927 -11.7812 translators 0.952381 -11.7814 expresion 0.952381 -11.7815 supply 0.178689 -11.7815 scheme 0.88 -11.7816 152 0.952381 -11.7818 decomposing 0.952381 -11.7818 asserted 0.952381 -11.7822 nmi 0.952381 -11.7823 formalization 0.952381 -11.7823 correlation, 0.952381 -11.7823 abbreviations, 0.742857 -11.7824 “not 0.307317 -11.7827 separately 0.821429 -11.7827 students, 0.710526 -11.7827 would result 0.742857 -11.7828 our system’s 0.952381 -11.7828 endings 0.210526 -11.7829 is used to 0.952381 -11.783 lexica 0.168116 -11.783 key 0.952381 -11.7831 vector-space 0.546875 -11.7831 boundaries, 0.8 -11.7831 orientations 0.187956 -11.7831 turn 0.88 -11.7833 edition 0.710526 -11.7833 quirk 0.952381 -11.7835 operators, 0.592593 -11.7836 we try to 0.952381 -11.7838 strengthen 0.88 -11.784 subclass 0.88 -11.784 a’s 0.638298 -11.7841 mi, 0.513889 -11.7841 hundreds 0.8 -11.7842 weaknesses of 0.952381 -11.7843 hatzivassiloglou, 0.952381 -11.7844 leveraged 0.952381 -11.7846 argumentative 0.952381 -11.7846 usage-based 0.952381 -11.7848 susceptible to 0.952381 -11.7849 stanza 0.821429 -11.7849 v-measure 0.88 -11.7849 jedi 0.88 -11.7849 wildcard 0.952381 -11.7851 185 0.913043 -11.7853 wi, 0.821429 -11.7853 78.8 0.19798 -11.7854 the entire 0.88 -11.7854 documented 0.659091 -11.7855 nlp, 0.422018 -11.7855 thought 0.638298 -11.7859 (clark and 0.952381 -11.7859 understandable 0.506667 -11.7859 yi 0.578947 -11.7859 allophonic 0.952381 -11.7862 hypergraphs 0.952381 -11.7864 loglinear 0.952381 -11.7866 hopefully 0.952381 -11.7867 unifying 0.952381 -11.7867 62% 0.592593 -11.7867 terminals 0.506667 -11.7868 event type 0.952381 -11.7869 divergence, 0.8 -11.7869 2-way 0.722222 -11.787 eu 0.952381 -11.7871 died 0.371429 -11.7871 6.1 0.952381 -11.7872 chart, 0.952381 -11.7875 offsets 0.368794 -11.7875 decoding. 0.329545 -11.7876 art 0.952381 -11.7877 machinery 0.952381 -11.7877 tense, 0.952381 -11.7877 taxonomies 0.742857 -11.7881 demonstrative 0.952381 -11.7882 industry 0.167883 -11.7883 derived 0.952381 -11.7884 mutation 0.821429 -11.7884 wb 0.952381 -11.7887 enough, 0.952381 -11.7887 marketing 0.952381 -11.7889 st’09 0.722222 -11.789 authors’ 0.8 -11.7893 heterogeneity 0.710526 -11.7893 αi 0.8 -11.7894 ambiguous. 0.952381 -11.7895 fda 0.952381 -11.7895 long time 0.846154 -11.7896 an on-line 0.952381 -11.7897 potentials 0.952381 -11.7897 golden 0.952381 -11.7898 multiplied by 0.952381 -11.79 cells, 0.952381 -11.79 responded 0.521127 -11.7902 vector machine 0.375 -11.7902 compatible 0.952381 -11.7905 visualizations 0.952381 -11.7905 constants 0.5 -11.7905 continues 0.952381 -11.7907 rational 0.952381 -11.7907 93% 0.952381 -11.7908 demo 0.952381 -11.7911 english-chinese 0.506667 -11.7911 learning algorithms 0.952381 -11.7915 pinpoint 0.952381 -11.7915 marginalized 0.821429 -11.7915 micro 0.952381 -11.7916 generalization, 0.757576 -11.7918 radev, 0.8 -11.7919 pdr 0.952381 -11.792 top-1 0.682927 -11.7923 very well 0.513889 -11.7923 induces 0.846154 -11.7925 precedence 0.952381 -11.7925 latent variables. 0.774194 -11.7925 bengali 0.952381 -11.793 superlative 0.952381 -11.7931 lascarides 0.316062 -11.7931 what is 0.952381 -11.7933 ia 0.625 -11.7933 pbsmt 0.821429 -11.7935 reviewer 0.363014 -11.7935 these methods 0.666667 -11.7936 exploratory 0.952381 -11.7936 iwsd 0.757576 -11.7936 exposed 0.592593 -11.7936 joint models 0.414414 -11.7937 gathered 0.952381 -11.7938 referenced 0.952381 -11.7938 lekbot 0.952381 -11.7938 exhibiting 0.952381 -11.7939 45% 0.465909 -11.7941 covariance 0.163866 -11.7941 achieve 0.952381 -11.7941 supplement 0.952381 -11.7943 adjustment 0.952381 -11.7943 world’s 0.322404 -11.7946 dt 0.952381 -11.7948 anticipate 0.682927 -11.7949 and matsumoto, 0.162983 -11.7949 gives 0.952381 -11.7951 negations 0.952381 -11.7951 emotion, 0.952381 -11.7951 (left 0.952381 -11.7951 striking 0.952381 -11.7953 sbar 0.952381 -11.7954 compress 0.952381 -11.7954 stk 0.952381 -11.7954 chi-square 0.722222 -11.7955 accurate, 0.952381 -11.7956 udop 0.952381 -11.7956 useless 0.913043 -11.7958 penalizing 0.0914236 -11.796 those 0.952381 -11.7962 dsms 0.952381 -11.7962 senti-features 0.757576 -11.7965 demonstration 0.952381 -11.7966 productivity 0.952381 -11.7971 l(u) 0.757576 -11.7971 shaded 0.846154 -11.7972 hapax 0.952381 -11.7972 tags: 0.231198 -11.7974 can also 0.774194 -11.7975 repetitions 0.952381 -11.7976 reading, 0.774194 -11.7976 the same, 0.546875 -11.7976 and knight, 0.952381 -11.7979 data), 0.952381 -11.7982 sproat 0.952381 -11.7982 fractions 0.644444 -11.7983 degrades 0.682927 -11.7983 we do not use 0.521739 -11.7983 models: 0.757576 -11.7983 federal 0.568966 -11.7986 1994; 0.385827 -11.7989 channel 0.431373 -11.7989 comparison. 0.913043 -11.799 wprime 0.821429 -11.799 0.41 0.821429 -11.799 football 0.875 -11.799 word-aligned parallel 0.659091 -11.7992 computation. 0.431373 -11.7993 improvement. 0.913043 -11.7996 characterization of 0.147225 -11.7996 probabilities 0.952381 -11.7996 coarsely 0.722222 -11.7997 experimental evaluation 0.692308 -11.7998 convention 0.3 -11.8 structures. 0.846154 -11.8001 ea 0.913043 -11.8001 records. 0.952381 -11.8001 moschitti 0.952381 -11.8001 subscript 0.952381 -11.8001 slots, 0.100368 -11.8001 even 0.952381 -11.8002 aspects: 0.952381 -11.8002 misclassified as 0.952381 -11.8002 fresh 0.821429 -11.8003 now turn 0.209677 -11.8004 clear 0.952381 -11.8004 4.3.2 0.952381 -11.8004 week, 0.286957 -11.8006 adopt 0.757576 -11.8006 to gather 0.846154 -11.8007 aaai 0.952381 -11.8007 microblogs 0.952381 -11.8007 standing 0.952381 -11.8007 subword 0.465909 -11.8008 future, we 0.27381 -11.8008 exists 0.37594 -11.8008 restriction 0.952381 -11.8012 coordinates 0.774194 -11.8012 ciphers 0.578947 -11.8013 outgoing 0.952381 -11.8014 anotated 0.692308 -11.8015 more reliable 0.952381 -11.8017 politics 0.952381 -11.8017 tree: 0.821429 -11.8017 <= 0.952381 -11.802 hyponymy 0.952381 -11.802 here: 0.952381 -11.8024 as regards 0.952381 -11.8024 balancing 0.493506 -11.8031 et al., 2002), 0.513889 -11.8031 sentence selection 0.212411 -11.8032 extend 0.846154 -11.8034 thoracic 0.846154 -11.8034 created, 0.692308 -11.8034 0.95 0.952381 -11.8035 partial patterns 0.846154 -11.8037 (sections 0.644444 -11.804 tables, 0.952381 -11.804 (typically 0.952381 -11.8042 replies 0.952381 -11.8042 more detail. 0.952381 -11.8045 gradients 0.952381 -11.8045 princeton 0.281513 -11.8046 largest 0.378788 -11.8046 ranks 0.735294 -11.8046 only slightly 0.952381 -11.8047 an in-house 0.952381 -11.8049 poster 0.952381 -11.805 temperature 0.952381 -11.805 continuously 0.909091 -11.805 crawl 0.952381 -11.8052 tsg 0.952381 -11.8052 more) 0.557377 -11.8053 use, 0.192233 -11.8053 active 0.909091 -11.8054 dp-tsg 0.757576 -11.8056 confirming 0.846154 -11.8057 agreeing 0.909091 -11.8057 compound, 0.821429 -11.8057 f., 0.952381 -11.8059 wilson, 0.952381 -11.806 opinion, 0.909091 -11.806 monitor 0.909091 -11.8062 text-to-speech 0.460674 -11.8063 dialogue systems 0.952381 -11.8064 183 0.875 -11.8064 subsequences 0.875 -11.8064 n-best list. 0.952381 -11.8067 gene names 0.909091 -11.8069 reply 0.557377 -11.8069 to tackle 0.506849 -11.8072 attitude 0.821429 -11.8075 when, 0.644444 -11.8078 to constrain 0.793103 -11.8078 nber 0.909091 -11.8079 relata 0.596154 -11.8081 switch 0.666667 -11.8081 road 0.846154 -11.8088 pennacchiotti, 1 -11.8092 english). 1 -11.8092 co-cited 1 -11.8092 σ. 1 -11.8092 grammaticality. 1 -11.8092 209). 1 -11.8092 portuguese. 1 -11.8092 mechanism. 1 -11.8092 worse. 1 -11.8092 2010)). 1 -11.8092 estimated. 1 -11.8092 above). 1 -11.8092 tractable. 1 -11.8092 ugaritic 1 -11.8092 emotion. 1 -11.8092 applicable. 1 -11.8092 give-2 1 -11.8092 ground truth 1 -11.8092 2001. 1 -11.8092 snippets. 1 -11.8092 s2. 1 -11.8092 coefficient. 1 -11.8092 suggestion. 1 -11.8092 non-learner 1 -11.8092 fact. 1 -11.8092 50. 1 -11.8092 pairs). 1 -11.8092 queue. 1 -11.8092 co-citation 1 -11.8092 contribution. 1 -11.8092 genes. 1 -11.8092 untagged 1 -11.8092 engineering. 1 -11.8092 now. 1 -11.8092 apply. 1 -11.8092 citations. 1 -11.8092 met. 1 -11.8092 much. 1 -11.8092 sect. 1 -11.8092 regions. 1 -11.8092 substantially. 0.95 -11.8094 (0.1) 0.95 -11.8094 normalization. 0.27027 -11.8096 fit 0.446809 -11.8098 1993) 0.95 -11.8099 weak supervision 0.95 -11.8099 time consuming 0.95 -11.81 appendix a. 0.735294 -11.8102 head-driven 0.909091 -11.8106 thirdly, 0.909091 -11.8106 (k 0.909091 -11.8106 5.8 0.95 -11.8109 discourse connectives. 0.355705 -11.8113 grouped 0.846154 -11.8115 (after 0.95 -11.8115 closer look at 0.846154 -11.8117 textual entailment, 0.95 -11.8117 arithmetic mean 0.846154 -11.8118 efect 0.821429 -11.8118 dominate 0.909091 -11.8119 recall) 0.493506 -11.8119 acc 0.95 -11.812 task; 0.581818 -11.812 (r) 0.32948 -11.812 recognize 0.0936052 -11.8121 terms 0.875 -11.8121 evaluativity 0.95 -11.8122 wan et al. 0.95 -11.8124 littman, 0.821429 -11.8125 chart. 0.821429 -11.8125 fields, 0.398305 -11.8126 allow for 0.95 -11.8127 recognised 0.612245 -11.8127 bank 0.846154 -11.8129 (denoted as 0.420561 -11.8129 we evaluate our 0.521739 -11.8129 modeling. 0.95 -11.813 as mentioned earlier, 0.692308 -11.813 tackled 0.875 -11.813 experimental results. 0.95 -11.8132 boxes 0.390244 -11.8134 for extracting 0.95 -11.8137 dirt 0.909091 -11.8137 treatments 0.909091 -11.8138 144 0.774194 -11.814 k-nearest 0.95 -11.8142 validating 0.95 -11.8145 logical form, 0.530303 -11.8146 dependents 0.909091 -11.8148 sequence kernels 0.95 -11.8149 (ng 0.666667 -11.815 pos. 0.793103 -11.8153 med 0.95 -11.8154 68.4 0.909091 -11.8154 passive voice 0.875 -11.8154 church, 0.875 -11.8154 74.7 0.846154 -11.8155 opinion summarization 0.909091 -11.8156 edus 0.95 -11.8157 lascarides, 0.235294 -11.8158 conducted 0.581818 -11.8159 there are some 0.846154 -11.816 (spitkovsky 0.846154 -11.8161 cmc 0.909091 -11.8161 company, 0.95 -11.8162 26% 0.793103 -11.8162 apache 0.284483 -11.8166 creation 0.909091 -11.8167 124 0.15903 -11.8168 simply 0.539683 -11.8168 test data, 0.666667 -11.817 augmented-loss 0.630435 -11.817 feature extraction 0.644444 -11.8171 simulations 0.493506 -11.8171 proven 0.630435 -11.8173 noticed that 0.651163 -11.8174 second language 0.702703 -11.8174 large collection 0.774194 -11.8174 ad-hoc 0.846154 -11.8176 consider, 0.95 -11.8176 27% 0.95 -11.8177 seconds, 0.793103 -11.8178 see whether 0.548387 -11.8181 œ 0.95 -11.8181 23% 0.95 -11.8181 m`arquez, 0.343949 -11.8183 should not 0.875 -11.8183 mcdonald et al. 0.394958 -11.8188 by taking 0.702703 -11.8188 sj 0.95 -11.8189 saving 0.95 -11.8189 book request 0.95 -11.8189 mistakes, 0.909091 -11.819 industrial 0.95 -11.8191 governing 0.95 -11.8191 nenkova, 0.487179 -11.8191 life 0.625 -11.8192 attachment score 0.722222 -11.8193 k., 0.487179 -11.8193 bleu. 0.875 -11.8197 ward 0.722222 -11.8199 (r 0.722222 -11.82 sf 0.875 -11.82 8.7 0.644444 -11.8202 mechanism for 0.875 -11.8202 2010): 0.630435 -11.8202 and mooney 0.170015 -11.8203 distribution of 0.106602 -11.8203 corresponding 0.95 -11.8204 0.5% 0.909091 -11.8204 contained within 0.735294 -11.8206 identified, 0.95 -11.8206 stoyanov 0.666667 -11.8208 (akkaya et 0.95 -11.8208 guiding 0.95 -11.8209 2010); 0.104801 -11.8209 parser 0.3125 -11.821 closer 0.735294 -11.8211 optimization. 0.95 -11.8211 duration, 0.95 -11.8211 goals, 0.95 -11.8213 pdf 0.774194 -11.8214 throw 0.793103 -11.8215 backtracking 0.95 -11.8216 pick up 0.493506 -11.8219 matter 0.875 -11.8222 meat 0.95 -11.8223 context-dependent 0.95 -11.8224 cubic 0.875 -11.8225 newstest2009 0.95 -11.8226 stop list 0.95 -11.8226 (goodman, 0.0470116 -11.8227 words 0.46988 -11.8228 this task, 0.95 -11.8228 65.4 0.95 -11.8228 persons, 0.875 -11.8228 fellbaum, 0.95 -11.823 connotation lexicon 0.909091 -11.823 henceforth 0.75 -11.8231 some interesting 0.95 -11.8231 annotator’s 0.95 -11.8231 associations, 0.666667 -11.8232 subsequence 0.875 -11.8233 1986; 0.875 -11.8234 hter 0.875 -11.8234 at&t 0.95 -11.8235 4.8 0.666667 -11.8236 because, 0.95 -11.8236 non-redundant 0.95 -11.8236 machine, 0.95 -11.8236 approaching 0.45977 -11.8236 specificity 0.774194 -11.8237 segmentations. 0.507042 -11.8238 strategy. 0.95 -11.824 76.0 0.95 -11.824 (snyder 0.95 -11.824 verbs) 0.95 -11.824 thesaurus. 0.95 -11.8241 reused 0.793103 -11.8245 measures: 0.46988 -11.8246 d = 0.95 -11.8248 manipulated 0.584906 -11.8248 tail 0.846154 -11.825 cross-domain sentiment 0.95 -11.825 representational 0.875 -11.825 9% 0.774194 -11.825 nj 0.548387 -11.8251 diagram 0.702703 -11.8251 indicated that 0.692308 -11.8252 realize 0.95 -11.8253 redirect 0.465116 -11.8254 1995). 0.814815 -11.8254 decoding algorithms 0.95 -11.8255 geoquery 0.814815 -11.8255 12, 0.774194 -11.8258 jane 0.139853 -11.8259 weight 0.666667 -11.8261 adam 0.846154 -11.8262 vbg 0.774194 -11.8262 incidence 0.47561 -11.8263 10% of 0.793103 -11.8264 digit 0.702703 -11.8264 gap-fill 0.367647 -11.8264 comparison to 0.95 -11.8265 milion 0.95 -11.8265 5.7 0.95 -11.8267 70.0 0.95 -11.8268 fly 0.415094 -11.8268 for generating 0.442105 -11.8269 engines 0.320442 -11.827 in turn 0.95 -11.827 (wsd) 0.361702 -11.827 paradigm 0.846154 -11.8271 chinese restaurant 0.95 -11.8272 correspondences between 0.95 -11.8275 ...” 0.95 -11.8275 connective, 0.514286 -11.8277 picked 0.814815 -11.8277 “a” 0.27686 -11.8277 arcs 0.95 -11.828 error-rate 0.814815 -11.828 jumps 0.793103 -11.828 show that, 0.522388 -11.8282 in this way 0.493333 -11.8283 (lin and 0.539683 -11.8283 converge 0.793103 -11.8284 by optimizing 0.376923 -11.8284 we build 0.875 -11.8284 reads 0.95 -11.8285 snapshots 0.875 -11.8286 diferences 0.95 -11.8287 semafor 0.95 -11.8287 akin 0.95 -11.8291 sunday 0.814815 -11.8292 (hovy et 0.875 -11.8292 confidence scores 0.454545 -11.8293 shows how 0.735294 -11.8293 investigate whether 0.95 -11.8294 81% 0.95 -11.8294 feelings 0.514286 -11.8296 human evaluation 0.95 -11.8296 zheng 0.95 -11.8296 normalization, 0.735294 -11.8297 policies, 0.95 -11.8297 moses toolkit 0.774194 -11.8297 state item 0.875 -11.8298 repositories 0.95 -11.8299 71.0 0.2 -11.8299 learner 0.55 -11.83 nucleus 0.95 -11.8301 was, 0.793103 -11.8301 jurafsky 0.0920693 -11.8303 null 0.46988 -11.8306 begins 0.95 -11.8311 optionally 0.353741 -11.8313 suggesting 0.95 -11.8313 semantic, 0.95 -11.8313 161 0.220472 -11.8314 difficult to 0.909091 -11.8319 d: 0.95 -11.832 topological 0.95 -11.832 ∞ 0.95 -11.8321 prolific 0.95 -11.8321 aligners 0.909091 -11.8321 brown clusters 0.323864 -11.8321 dimension 0.95 -11.8325 tf*idf 0.793103 -11.8325 0.42 0.0390561 -11.8327 or 0.4 -11.8328 list, 0.95 -11.8328 more detail, 0.875 -11.8328 conceptnet 0.814815 -11.833 vogel 0.127067 -11.8331 k 0.95 -11.8333 30.8 0.793103 -11.8333 christian 0.814815 -11.8334 reference: 0.95 -11.8337 questioning 0.793103 -11.8337 cross-entity 0.95 -11.8338 instant 0.875 -11.8339 code. 0.95 -11.834 to” 0.793103 -11.8341 114 0.95 -11.8343 151 0.95 -11.8343 encyclopedic 0.95 -11.8345 developer 0.714286 -11.8346 iii) 0.714286 -11.8347 very important 0.282609 -11.8347 choosing 0.675 -11.8347 ours, 0.307692 -11.8348 columns 0.95 -11.8349 closing 0.95 -11.8352 lines, 0.493333 -11.8355 source sentences 0.95 -11.8355 stem, 0.95 -11.8355 srl, 0.95 -11.8355 mccarthy 0.95 -11.8357 64.4 0.95 -11.8357 somehow 0.95 -11.8357 subtractive 0.95 -11.8359 methodological 0.75 -11.836 target-dependent 0.793103 -11.836 london 0.714286 -11.8361 gamon 0.450549 -11.8361 6 conclusions 0.875 -11.8361 mt06 0.95 -11.8362 expertise 0.95 -11.8362 childes 0.95 -11.8362 t-test, 0.814815 -11.8362 performance gain 0.793103 -11.8363 transfer rules 0.909091 -11.8363 means, 0.95 -11.8364 part-of 0.875 -11.8364 -2 0.325581 -11.8365 techniques. 0.95 -11.8366 corpus-level 0.95 -11.8366 incurred 0.814815 -11.8366 various methods 0.95 -11.8367 safely 0.95 -11.8367 sea 0.95 -11.8367 language) 0.361702 -11.8368 implemented in 0.95 -11.8369 uncovered 0.814815 -11.8369 long. 0.75 -11.8371 korea 0.95 -11.8371 (chan 0.95 -11.8373 71% 0.95 -11.8373 nominative 0.875 -11.8374 software, 0.95 -11.8376 voiced 0.84 -11.8377 512 0.84 -11.8379 2k 0.84 -11.8379 “is 0.95 -11.8379 indicators, 0.95 -11.8379 recognizers 0.95 -11.8379 metasearch 0.909091 -11.838 summarizer 0.95 -11.8383 a4 0.95 -11.8385 scarce 0.617021 -11.8386 hgfc 0.493333 -11.8387 constructed by 0.95 -11.8388 bibliographic 0.95 -11.8388 thompson, 0.359155 -11.8388 we select 0.684211 -11.8389 draws 0.814815 -11.8389 automatic identification 0.43299 -11.8389 emission 0.95 -11.839 robust, 0.95 -11.8391 conform 0.814815 -11.8393 homophonic 0.651163 -11.8393 labeling, 0.95 -11.8395 characterised 0.95 -11.8395 telling 0.159944 -11.8395 results. 0.793103 -11.8398 confirmed that 0.630435 -11.8401 we describe how 0.95 -11.8402 .20 0.814815 -11.8402 set) 0.95 -11.8403 subset, 0.84 -11.8405 noun-noun 0.95 -11.8405 (due 0.95 -11.8407 altered 0.493333 -11.8408 it provides 0.493333 -11.8408 label, 0.27897 -11.8409 dialogues 0.393162 -11.841 denoting 0.814815 -11.8411 note also 0.486842 -11.8411 expressive 0.612245 -11.8411 1we 0.95 -11.8412 clarify 0.612245 -11.8413 thorough 0.735294 -11.8415 initially, 0.95 -11.8415 replication 0.651163 -11.8415 penalizes 0.909091 -11.8416 dashed lines 0.675 -11.8417 para 0.95 -11.8419 top-k 0.445652 -11.8419 iii 0.224377 -11.842 method, 0.95 -11.8421 transitionbased 0.793103 -11.8421 inference rules 0.95 -11.8422 i’ve 0.55 -11.8423 studies. 0.348993 -11.8424 updated 0.875 -11.8424 nonterminals. 0.95 -11.8426 challenges, 0.95 -11.8426 only; 0.909091 -11.8429 mert, 0.658537 -11.8429 graph g 0.84 -11.8431 .21 0.95 -11.8433 complete, 0.658537 -11.8433 accomplish 0.514706 -11.8437 cambridge 0.95 -11.8438 somewhere 0.909091 -11.8439 probability: 0.95 -11.8441 clda 0.702703 -11.8441 (np) 0.814815 -11.8442 gen 0.46988 -11.8443 specification 0.84 -11.8444 approximated by 0.339623 -11.8444 we explore 0.95 -11.8445 movies 0.143665 -11.8447 arguments 0.95 -11.8448 support, 0.218182 -11.845 performance on 0.814815 -11.8452 mse 0.362319 -11.8454 computed by 0.95 -11.8455 frequency counts 0.95 -11.8455 pps 0.95 -11.8455 employment 0.184701 -11.8455 denote 0.95 -11.8459 simplistic 0.253521 -11.8459 motivated 0.5 -11.8459 reconstruction 0.95 -11.8464 sitting 0.793103 -11.8465 national institute of 0.540984 -11.8467 by incorporating 0.95 -11.8471 (5 0.814815 -11.8471 doc 0.95 -11.8472 .16 0.714286 -11.8475 sent to 0.523077 -11.8475 proceeds 0.75 -11.8477 an implicit 0.95 -11.8478 protein, 0.766667 -11.8478 when evaluating 0.617021 -11.8479 parallel text 0.714286 -11.8479 and litman, 0.95 -11.8479 asociated with 0.95 -11.8479 overwhelming 0.117917 -11.848 role 0.95 -11.8481 179 0.793103 -11.8482 etzioni, 0.182994 -11.8484 second, 0.95 -11.8485 asynchronous 0.95 -11.8485 help, 0.658537 -11.8485 promote 0.211823 -11.8486 o 0.766667 -11.8487 i don’t 0.909091 -11.8487 source/target 0.6 -11.8488 rec 0.814815 -11.8489 steedman, 0.95 -11.849 221 0.95 -11.8491 bu 0.55 -11.8493 learnability 0.95 -11.8495 feasible, 0.95 -11.8495 retrieves 0.909091 -11.8495 production rules 0.507246 -11.8497 t-test 0.143341 -11.8497 containing 0.95 -11.8498 hd 0.371212 -11.85 lexical features 0.95 -11.8502 vp, 0.95 -11.8504 wsd, 0.684211 -11.8505 good, 0.95 -11.8505 service, 0.766667 -11.8505 rcv1 0.523077 -11.8506 token. 0.95 -11.8507 68.2 0.766667 -11.851 dark 0.766667 -11.851 .60 0.714286 -11.8511 discourse, 0.95 -11.8511 confidence, 0.95 -11.8511 directionality 0.23676 -11.8514 window 0.658537 -11.8515 rule based 0.636364 -11.8516 division 0.238994 -11.8517 2008), 0.909091 -11.8519 intelligence, 0.95 -11.8521 classification: 0.684211 -11.8523 correlation coefficients 0.95 -11.8523 turkish, 0.385246 -11.8525 we decided 0.588235 -11.8528 0.45 0.869565 -11.8529 lo 0.869565 -11.853 in such cases, 0.266667 -11.8531 our proposed 0.909091 -11.8532 conclusion. 0.909091 -11.8532 proteins. 0.814815 -11.8534 comments, 0.584906 -11.8534 say, 0.95 -11.8535 assignments, 0.684211 -11.8537 a principled 0.561404 -11.8538 program. 0.32 -11.8538 comparable to 0.714286 -11.8539 observed, 0.95 -11.854 preliminaries 0.588235 -11.854 office 0.684211 -11.8542 fs 0.75 -11.8543 jdmwe 0.814815 -11.8543 season 0.95 -11.8544 (5), 0.95 -11.8545 letter, 0.95 -11.8545 ppr 0.869565 -11.8546 woz data 0.95 -11.8547 row, 0.40367 -11.8551 boundaries. 0.95 -11.8551 loaded 0.584906 -11.8551 grammar rules 0.95 -11.8552 ej 0.95 -11.8554 hits, 0.814815 -11.8555 ye 0.359712 -11.8556 evaluation results 0.449438 -11.8558 edits 0.75 -11.8558 automatic detection 0.95 -11.8561 fb 0.95 -11.8563 (similar 0.6 -11.8563 amazon’s 0.75 -11.8564 joined 0.95 -11.8565 n best list 0.95 -11.8565 peaks 0.95 -11.8565 vocalization 0.95 -11.8566 abstract away 0.869565 -11.8567 pan et al. 0.95 -11.8568 czeng 0.95 -11.8568 brain 0.95 -11.8568 3this 0.271605 -11.8568 “the 0.43617 -11.8568 initialize 0.95 -11.857 hospital 0.75 -11.8572 pleasantness 0.95 -11.8572 ratings, 0.95 -11.8572 piggyback 0.814815 -11.8573 mined from 0.869565 -11.8577 ch 0.95 -11.8579 (2010)) 0.95 -11.8579 birch 0.84 -11.8583 deverbal 0.346667 -11.8586 other, 0.95 -11.8587 pb 0.904762 -11.8588 exclusion 0.200445 -11.8589 tagger 0.904762 -11.859 significantly, 0.904762 -11.859 hashed 0.658537 -11.859 regression. 0.127602 -11.8591 useful 0.904762 -11.8593 stems, 0.95 -11.8594 conll-x shared 0.869565 -11.8596 dang, 0.417476 -11.8596 hypothesis, 0.904762 -11.8597 rendered 0.463415 -11.8597 and finally 0.84 -11.86 february 0.904762 -11.8602 3.1.1 0.904762 -11.8603 singletons 0.306878 -11.8603 proposed method 0.904762 -11.8607 ...) 0.658537 -11.8607 students. 0.540984 -11.8608 abstract we present a 0.617021 -11.861 returning 0.714286 -11.8613 null hypothesis 0.904762 -11.8613 fear 0.588235 -11.8617 we will show 0.684211 -11.8619 straight 0.379032 -11.8624 best performance 0.766667 -11.8625 mathematics 0.540984 -11.8626 as far as 0.588235 -11.8626 was created 0.904762 -11.8627  0.540984 -11.8628 subject, 0.904762 -11.8632 gtm 1 -11.8633 (2010)). 1 -11.8633 markers. 1 -11.8633 participant. 1 -11.8633 signature. 1 -11.8633 problematic. 1 -11.8633 goals. 1 -11.8633 p´olya urn 1 -11.8633 template. 1 -11.8633 satisfaction. 1 -11.8633 colours. 1 -11.8633 members. 1 -11.8633 agenda. 1 -11.8633 mean. 1 -11.8633 sufficient. 1 -11.8633 kullback-leibler divergence 1 -11.8633 have. 1 -11.8633 reader. 1 -11.8633 confusing 1 -11.8633 disagreement. 1 -11.8633 predicted. 1 -11.8633 implemented. 1 -11.8633 degree. 1 -11.8633 proportional hazards 1 -11.8633 text). 1 -11.8633 lit. 1 -11.8633 friends. 1 -11.8633 script. 1 -11.8633 abstract. 1 -11.8633 repository. 1 -11.8633 right). 1 -11.8633 simplicity. 1 -11.8633 half. 1 -11.8633 wsj. 1 -11.8633 propbank. 1 -11.8633 reliability. 1 -11.8633 unsupervised. 1 -11.8633 subjectivity. 1 -11.8633 hours. 1 -11.8633 judgements. 1 -11.8633 4.2). 0.947368 -11.8635 quantifier scope 0.947368 -11.8635 referent. 0.947368 -11.8635 ci. 0.947368 -11.8635 often. 0.904762 -11.8635 compression rates 0.203233 -11.8636 reading 0.360294 -11.8637 r2 0.604167 -11.8639 right, 0.563636 -11.8639 bus 0.947368 -11.8642 status. 0.904762 -11.8643 codewords 0.869565 -11.8646 evidence, 0.453488 -11.8646 displays 0.604167 -11.8647 four different 0.540984 -11.8652 development data. 0.666667 -11.8653 ist 0.947368 -11.8654 experimental set-up 0.116578 -11.8654 space 0.604167 -11.8654 rwth 0.904762 -11.8655 borders 0.869565 -11.8657 federico 0.869565 -11.8662 particularly useful 0.694444 -11.8662 text processing 0.636364 -11.8664 cd 0.617021 -11.8665 7.2 0.947368 -11.8666 scopal 0.947368 -11.8668 words’ 0.904762 -11.8669 done, 0.869565 -11.867 scfgs 0.515152 -11.867 triggering 0.127737 -11.8671 term 0.666667 -11.8671 guidelines, 0.947368 -11.8673 (mccarthy 0.904762 -11.8675 animals 0.269388 -11.8675 observe that 0.904762 -11.8679 1z 0.947368 -11.868 vertices, 0.947368 -11.868 otherwise 0. 0.463415 -11.8681 reflected 0.947368 -11.8682 atis 0.904762 -11.8682 navigli 0.947368 -11.8684 elegant 0.293269 -11.8685 integrate 0.947368 -11.8687 (section 3.2). 0.658537 -11.869 tokyo, 0.947368 -11.8691 no matter 0.43956 -11.8691 detector 0.642857 -11.8691 mcdonald, 0.947368 -11.8693 (ir) 0.636364 -11.8696 sparsity. 0.947368 -11.8698 manageable 0.947368 -11.8698 detrimental 0.35461 -11.8699 summarizes 0.563636 -11.8699 the aforementioned 0.904762 -11.8702 8th 0.947368 -11.8703 banko 0.229851 -11.8703 2007), 0.112938 -11.8703 representation 0.869565 -11.8704 multiplying 0.869565 -11.8704 nmf 0.424242 -11.8705 connection 0.904762 -11.8709 1-gram 0.84 -11.871 0.29 0.947368 -11.871 all) 0.947368 -11.871 governed 0.947368 -11.8712 .52 0.694444 -11.8713 exceptions 0.869565 -11.8715 readings, 0.947368 -11.8716 machines, 0.947368 -11.8716 section 5.1. 0.947368 -11.8716 vastly 0.947368 -11.8716 .85 0.947368 -11.8717 77.6 0.727273 -11.8718 non-empty 0.727273 -11.8718 (it 0.947368 -11.8721 non-contiguous 0.727273 -11.8722 systems’ 0.785714 -11.8723 lemon, 0.947368 -11.8723 76% 0.380165 -11.8723 serves 0.947368 -11.8728 (xiong 0.246575 -11.8729 a certain 0.551724 -11.8731 with. 0.947368 -11.8733 task completion 0.251799 -11.8733 beam 0.904762 -11.8734 negra 0.0851324 -11.8734 types 0.947368 -11.8735 simulates 0.947368 -11.8735 relations) 0.947368 -11.8735 remarkably 0.262745 -11.8736 explained 0.666667 -11.8737 huffman 0.260536 -11.8738 try 0.947368 -11.8739 helps us 0.947368 -11.8739 vignet 0.947368 -11.874 drs 0.947368 -11.874 from scratch 0.515152 -11.8741 careful 0.622222 -11.8743 addressed by 0.212821 -11.8744 reduction 0.532258 -11.8746 annotated. 0.216359 -11.8747 the whole 0.267206 -11.8747 solve 0.5 -11.8748 word sequence 0.741935 -11.8749 improved performance 0.869565 -11.875 ordering, 0.396396 -11.8751 in this way, 0.727273 -11.8752 linear kernel 0.5 -11.8753 object, 0.947368 -11.8755 splitting, 0.766667 -11.8755 ut 0.84 -11.8756 we conjecture 0.947368 -11.8756 creativity 0.125333 -11.8757 significantly 0.563636 -11.8757 outputs. 0.947368 -11.8758 alphabets 0.727273 -11.8762 we ignore 0.666667 -11.8762 kernels. 0.104644 -11.8762 make 0.588235 -11.8762 we explain 0.366412 -11.8764 a major 0.807692 -11.8765 thematic fit 0.642857 -11.8766 1n 0.947368 -11.8769 needed, 0.253623 -11.8769 distribution over 0.766667 -11.877 we will discuss 0.947368 -11.8771 classified, 0.947368 -11.8772 easiest 0.807692 -11.8774 ?x 0.947368 -11.8776 am-fm 0.474359 -11.8776 decisions. 0.360294 -11.8777 with high 0.947368 -11.8778 82.6 0.396396 -11.8778 not directly 0.103448 -11.8778 much 0.869565 -11.8779 super 0.346939 -11.8779 university, 0.904762 -11.878 (semantic 0.947368 -11.8783 compactly 0.947368 -11.8783 consulting 0.576923 -11.8784 (about 0.0478098 -11.8787 system 0.947368 -11.8787 skew 0.947368 -11.8787 87.0 0.947368 -11.8787 forcing 0.947368 -11.8787 dyer, 0.869565 -11.879 hyperplane 0.947368 -11.879 geolocation 0.666667 -11.8791 lambda 0.474359 -11.8793 is crucial 0.869565 -11.8794 qualified 0.21466 -11.8794 increases 0.542373 -11.8794 identifier 0.563636 -11.8794 � 0.741935 -11.8795 comunication 0.947368 -11.8796 67.0 0.947368 -11.8796 real-life 0.947368 -11.8796 forbidden 0.192067 -11.8796 chosen 0.190476 -11.8796 effect of 0.869565 -11.8797 disconnected 0.234375 -11.8798 english. 0.947368 -11.8801 yates 0.947368 -11.8801 supertag 0.947368 -11.8801 questionnaires 0.221289 -11.8801 and thus 0.947368 -11.8803 cultures 0.947368 -11.8803 upper-bound 0.947368 -11.8803 planned 0.947368 -11.8805 heads, 0.947368 -11.8805 wisdom-lmde 0.947368 -11.8805 (se 0.474359 -11.8806 differ in 0.588235 -11.8806 5% of 0.869565 -11.8807 68.8 0.947368 -11.8808 49.5 0.622222 -11.881 an expression 0.947368 -11.8812 alter 0.232416 -11.8814 nn 0.947368 -11.8815 gave us 0.947368 -11.8817 facilitated 0.904762 -11.8818 relatively easy 0.5 -11.8818 conversation. 0.576923 -11.8819 lemma, 0.947368 -11.8821 swapping 0.947368 -11.8821 bracketed 0.947368 -11.8824 classroom 0.869565 -11.8825 hindi, 0.869565 -11.8826 undergo 0.387931 -11.8826 distribution, 0.694444 -11.8827 standard phrase-based 0.576923 -11.8827 paraphrases, 0.947368 -11.8828 43.8 0.904762 -11.883 scanned 0.694444 -11.883 lecture 0.766667 -11.8831 happy 0.576923 -11.8831 0.72 0.306452 -11.8832 rates 0.947368 -11.8833 competitions 0.947368 -11.8835 self-training, 0.947368 -11.8837 hurts 0.947368 -11.8837 finite, 0.947368 -11.8838 75.1 0.947368 -11.8838 edge, 0.947368 -11.8838 positives, 0.947368 -11.8838 section 4.1, 0.947368 -11.884 78.5 0.947368 -11.884 teufel 0.134084 -11.8842 comparable 0.869565 -11.8846 range, 0.515152 -11.8846 fill 0.486486 -11.8848 small amount 0.904762 -11.885 summing over 0.666667 -11.8851 swap 0.947368 -11.8853 move, 0.807692 -11.8853 semantic templates 0.591837 -11.8855 swiss 0.869565 -11.8856 dynamics 0.947368 -11.8856 77.5 0.947368 -11.8856 automation 0.237179 -11.8857 absolute 0.298969 -11.8858 valuable 0.947368 -11.886 unwanted 0.947368 -11.886 np-hard 0.474359 -11.886 latent dirichlet allocation 0.947368 -11.8864 66.6 0.947368 -11.8865 etts 0.785714 -11.8867 most difficult 0.785714 -11.8867 p(y 0.947368 -11.8867 consumption 0.947368 -11.8867 merged, 0.52381 -11.8872 relates 0.5 -11.8872 slightly better 0.947368 -11.8873 nltk 0.869565 -11.8874 yt 0.947368 -11.8874 lengths, 0.705882 -11.8875 (lm) 0.869565 -11.8876 4 related work 0.947368 -11.8876 finitestate 0.947368 -11.8876 distributionally 0.457831 -11.8877 hebrew 0.785714 -11.8877 simple: 0.904762 -11.8879 “no 0.947368 -11.888 65.6 0.304813 -11.8881 similarity. 0.947368 -11.8882 distance-based 0.947368 -11.8883 2note that 0.947368 -11.8885 65.1 0.947368 -11.8885 54% 0.642857 -11.8885 spitkovsky 0.200461 -11.8888 (2007) 0.785714 -11.889 is, however, 0.947368 -11.8891 86% 0.357664 -11.8892 aimed 0.553571 -11.8892 constituents. 0.904762 -11.8893 rt 0.947368 -11.8894 ksummationdisplay k=1 0.947368 -11.8894 relationship, 0.947368 -11.8894 inheritance 0.947368 -11.8894 organised 0.279279 -11.8894 events. 0.576923 -11.8894 speakers. 0.443182 -11.8899 zhao 0.947368 -11.89 41% 0.947368 -11.89 hit, 0.947368 -11.89 (asr) 0.947368 -11.8903 paper) 0.398148 -11.8905 chooses 0.947368 -11.8907 committed 0.947368 -11.8909 90.0 0.947368 -11.8909 altogether 0.807692 -11.8911 infeasible 0.947368 -11.8912 34% 0.608696 -11.8912 organizers 0.947368 -11.8914 contingency 0.947368 -11.8918 suboptimal 0.947368 -11.8918 client 0.947368 -11.8919 teaching 0.807692 -11.8919 feature selection. 0.869565 -11.892 relatedness, 0.947368 -11.8921 82.8 0.245675 -11.8922 2002) 0.947368 -11.8923 events) 0.947368 -11.8923 workflow 0.947368 -11.8923 2010a) 0.947368 -11.8923 section 4.3. 0.355072 -11.8924 immediate 0.947368 -11.8925 speaks 0.807692 -11.8925 in front 0.785714 -11.8928 for this purpose 0.947368 -11.8928 slowly 0.947368 -11.8928 roget’s 0.833333 -11.8928 table 1: statistics 0.438202 -11.893 (li 0.0867127 -11.893 proposed 0.694444 -11.893 sees 0.293532 -11.8931 briefly 0.144928 -11.8931 translated 0.947368 -11.8932 webpages 0.576923 -11.8932 proxy 0.869565 -11.8934 evaluation criteria 0.666667 -11.8935 dev set 0.48 -11.8935 repeatedly 0.396396 -11.8935 integration of 0.297436 -11.8936 to incorporate 0.576923 -11.8936 sums 0.622222 -11.8937 errors made 0.869565 -11.8937 punjabi 0.387931 -11.8937 utterances. 0.869565 -11.894 unsel 0.947368 -11.8941 54.5 0.666667 -11.8943 ourselves to 0.947368 -11.8943 evoking 0.355072 -11.8943 w1 0.694444 -11.8943 base distribution 0.65 -11.8944 annotation process 0.869565 -11.8947 messaging 0.947368 -11.8947 viewpoint 0.947368 -11.8947 imp 0.947368 -11.8947 initializing 0.947368 -11.8948 inputs, 0.947368 -11.895 chosen, 0.947368 -11.895 cheap 0.947368 -11.8952 antonym 0.947368 -11.8954 deletions, 0.947368 -11.8954 seeking 0.947368 -11.8956 8.0 0.785714 -11.8958 certain types 0.947368 -11.8961 usenet 0.833333 -11.8962 hours, 0.947368 -11.8965 vectorθ 0.947368 -11.8966 open-ended 0.947368 -11.8966 recognizer, 0.833333 -11.8968 acknowledgment 0.705882 -11.897 bootstrapped 0.486111 -11.8973 french-english 0.947368 -11.8976 transformations, 0.947368 -11.8979 propositions, 0.947368 -11.8981 guessing 0.947368 -11.8983 art, 0.947368 -11.8983 expansion, 0.785714 -11.8985 cea 0.947368 -11.8986 respects 0.398148 -11.8989 surprising 0.947368 -11.899 newly generated 0.947368 -11.899 non-verbal 0.807692 -11.899 // 0.947368 -11.8992 emd 0.947368 -11.8992 bytes 0.947368 -11.8994 requirements, 0.947368 -11.8996 encompasses 0.904762 -11.8997 discounted 0.833333 -11.8999 relatively simple 0.833333 -11.9 talks 0.608696 -11.9002 fairy 0.54386 -11.9003 participating in 0.666667 -11.9003 month 0.947368 -11.9003 wsyn 0.947368 -11.9003 classifiers: 0.666667 -11.9004 tree fragments 0.947368 -11.9005 referential 0.947368 -11.9006 font 0.608696 -11.9006 by providing 0.807692 -11.9007 assessment. 0.807692 -11.9009 lu, 0.947368 -11.901 trouble 0.947368 -11.901 enormous 0.147914 -11.9011 rate 0.947368 -11.9012 decades 0.833333 -11.9013 promoting 0.785714 -11.9017 authoring 0.947368 -11.9017 drdas 0.807692 -11.9018 engagement 0.705882 -11.9018 mallet 0.278027 -11.9019 crucial 0.758621 -11.9021 lexrank 0.758621 -11.9021 ys 0.947368 -11.9021 problems: 0.947368 -11.9021 tabulation 0.105844 -11.9022 consider 0.407767 -11.9025 null. 0.947368 -11.9025 possible explanation 0.947368 -11.9025 qc 0.947368 -11.9025 168 0.807692 -11.9027 c0 0.364341 -11.9029 backward 0.947368 -11.903 projection matrix 0.947368 -11.903 tan 0.741935 -11.9031 top 20 0.833333 -11.9031 level) 0.785714 -11.9032 undesirable 0.947368 -11.9032 summarised 0.102959 -11.9033 size 0.433333 -11.9036 are highly 0.627907 -11.9038 penalized 0.947368 -11.9039 csc 0.947368 -11.9041 stylometric 0.947368 -11.9043 wins 0.758621 -11.9045 two columns 0.833333 -11.9045 sm 0.144593 -11.905 cannot 0.947368 -11.9054 book, 0.758621 -11.9055 electronics 0.238411 -11.9056 we apply 0.947368 -11.9061 types) 0.947368 -11.9061 72.4 0.16208 -11.9062 edge 0.807692 -11.9062 more flexible 0.947368 -11.9063 inspected 0.947368 -11.9063 introduced, 0.947368 -11.9065 there, 0.452381 -11.9067 a preliminary 0.71875 -11.9067 annotates 0.35 -11.9069 are available 0.863636 -11.907 graph partitioning 0.71875 -11.907 2002. 0.807692 -11.9071 2.9 0.58 -11.9071 view. 0.947368 -11.9074 320 0.675676 -11.9075 later, 0.947368 -11.9076 adoption 0.947368 -11.9076 factors: 0.947368 -11.9078 begun 0.904762 -11.9081 error analysis, 0.947368 -11.9083 bible 0.608696 -11.9083 dialects 0.947368 -11.9085 regulation, 0.553571 -11.9087 faster than 0.947368 -11.9089 gf 0.566038 -11.9089 mt05 0.705882 -11.909 when considering 0.591837 -11.909 below: 0.675676 -11.909 later. 0.947368 -11.9091 misclassification 0.904762 -11.9093 extension. 0.904762 -11.9093 thresholds. 0.947368 -11.9096 viability of 0.947368 -11.9098 labeled, 0.947368 -11.9098 (19) 0.741935 -11.91 semantic interpretation 0.947368 -11.9102 per-user 0.117363 -11.9103 nodes 0.947368 -11.9105 forecast 0.947368 -11.9105 williams 0.807692 -11.9106 associations. 0.947368 -11.9107 one would expect 0.947368 -11.9107 70.6 0.947368 -11.9107 alice 0.947368 -11.9109 rose, 0.947368 -11.9111 2.1, 0.947368 -11.9111 similarity: 0.71875 -11.9111 ca, 0.863636 -11.9112 volume37,number1 0.333333 -11.9112 significantly better 0.947368 -11.9114 11.8 0.286408 -11.9114 al., 2007) 0.863636 -11.9115 94.6 0.286408 -11.9115 optimize 0.385965 -11.9116 graphs. 0.758621 -11.9117 collocation errors 0.394495 -11.912 we construct 0.947368 -11.9122 0.1, 0.947368 -11.9122 exploitation 0.433333 -11.9122 fair 0.947368 -11.9124 boosted 0.947368 -11.9124 segmentations, 0.833333 -11.9124 collects 0.533333 -11.9125 main verb 0.308989 -11.9126 serve 0.41 -11.9126 longest 0.302703 -11.9126 assumption that 0.245552 -11.9129 sets, 0.473684 -11.913 law 0.947368 -11.9131 python 0.863636 -11.9132 tradition 0.947368 -11.9133 discards 0.595745 -11.9134 (x 0.947368 -11.9135 pressure 0.627907 -11.9136 subj 0.741935 -11.9137 exclusive 0.613636 -11.9139 just one 0.758621 -11.9139 common. 0.705882 -11.9139 denis 0.833333 -11.9139 1.9 0.758621 -11.9143 sorted by 0.947368 -11.9146 accusative 0.833333 -11.9147 cdp 0.634146 -11.9148 word classes 0.315789 -11.9148 carry 0.741935 -11.9148 (i 0.675676 -11.9148 shallow semantic 0.675676 -11.9148 emphasis on 0.295337 -11.915 a separate 0.947368 -11.9151 facilitating 0.947368 -11.9153 hmm+ref 0.685714 -11.9154 org 0.947368 -11.9155 primal 0.613636 -11.9158 chosen. 0.416667 -11.9158 wider 0.58 -11.9161 among these 0.9 -11.9162 feature template 0.833333 -11.9162 qsd 0.741935 -11.9163 frame elements 0.188679 -11.9163 our experiments 0.45679 -11.9163 factors. 0.947368 -11.9164 closeness 0.9 -11.9165 dimension, 0.52459 -11.9165 input sentences 0.461538 -11.9166 any given 0.9 -11.9167 preferences, 0.627907 -11.9167 molecular 0.414141 -11.9169 representations. 0.807692 -11.917 0.39 0.675676 -11.9171 3000 0.9 -11.9172 cm construction 0.65 -11.9173 it seems that 0.657895 -11.9173 symbol, 0.833333 -11.9173 too large 0.262295 -11.9177 interested in 0.9 -11.9178 kronecker 0.9 -11.9178 artist 0.54386 -11.9179 computational linguistics. 0.9 -11.9181 hereafter 0.515625 -11.9182 is unlikely 0.208763 -11.9184 incorporate 0.9 -11.9186 8) 0.9 -11.9186 jackson 0.627907 -11.9186 countries 0.634146 -11.9186 convolution 0.396226 -11.9189 mbr 0.777778 -11.9189 starting point for 0.295337 -11.9191 school 0.758621 -11.9192 patent 0.777778 -11.9194 svr 0.833333 -11.9196 85.5 0.613636 -11.9202 general. 1 -11.9204 deletion. 1 -11.9204 calculation. 1 -11.9204 crowdsourcing. 1 -11.9204 classifications. 1 -11.9204 statements. 1 -11.9204 benefits. 1 -11.9204 1977). 1 -11.9204 orderings. 1 -11.9204 weighting. 1 -11.9204 merged. 1 -11.9204 crucial. 1 -11.9204 specifications. 1 -11.9204 w2. 1 -11.9204 about. 1 -11.9204 communications. 1 -11.9204 paragraph. 1 -11.9204 protein. 1 -11.9204 (pp. 1 -11.9204 irony. 1 -11.9204 attachment. 1 -11.9204 content shifters 1 -11.9204 readings. 1 -11.9204 wild card 1 -11.9204 feasible. 1 -11.9204 efforts. 1 -11.9204 preferred. 1 -11.9204 uses. 1 -11.9204 tf.idf 1 -11.9204 environments. 1 -11.9204 clear. 1 -11.9204 determiners. 1 -11.9204 hansards 1 -11.9204 consistency. 1 -11.9204 duration. 1 -11.9204 desirable. 1 -11.9204 14. 1 -11.9204 connected. 1 -11.9204 missing. 1 -11.9204 models). 1 -11.9204 maps. 1 -11.9204 polysemy. 1 -11.9204 products. 1 -11.9204 uncertainty. 1 -11.9204 behaviors. 1 -11.9204 ngrams. 1 -11.9204 baseline). 1 -11.9204 studied. 1 -11.9204 determined. 1 -11.9204 dimension. 1 -11.9204 [0,1]. 1 -11.9204 operator. 1 -11.9204 effectively. 1 -11.9204 you. 0.657895 -11.9205 learning-based 0.944444 -11.9206 equivalents. 0.944444 -11.9206 2000. 0.944444 -11.9206 five-fold 0.777778 -11.9206 (%), 0.65 -11.9208 umass 0.657895 -11.9208 genres, 0.9 -11.9209 entry, 0.944444 -11.921 kendall’s τ 0.833333 -11.921 hitting 0.5 -11.921 consumer 0.944444 -11.9212 semeval 2007 0.833333 -11.9212 interlinear 0.234528 -11.9215 commonly 0.421053 -11.9215 to examine 0.705882 -11.9215 very useful 0.9 -11.9216 particularly, 0.777778 -11.9216 0.05. 0.807692 -11.9219 coefficient, 0.777778 -11.9219 ravichandran 0.516129 -11.922 a predefined 0.863636 -11.922 associative 0.71875 -11.9222 many researchers 0.9 -11.9223 discussed, 0.315476 -11.9224 3.4 0.944444 -11.9225 reachable 0.685714 -11.9226 we can now 0.758621 -11.9227 trivially 0.944444 -11.9227 graded 0.634146 -11.9227 not sufficient 0.863636 -11.9227 winner 0.944444 -11.9229 parameter estimation. 0.863636 -11.923 71.6 0.944444 -11.9231 smal 0.944444 -11.9231 clearer 0.833333 -11.9232 ate 0.25188 -11.9233 if they 0.657895 -11.9234 for estimating 0.613636 -11.9237 los 0.758621 -11.9237 losing 0.9 -11.9239 a3 0.9 -11.9239 element, 0.485714 -11.9239 in line with 0.944444 -11.924 accurately classified 0.9 -11.9241 70 75 0.258964 -11.9241 positions 0.170984 -11.9242 path 0.863636 -11.9242 prenominal 0.833333 -11.9242 cl 0.944444 -11.9244 guides 0.863636 -11.9244 partitive 0.944444 -11.9245 diverge 0.9 -11.9245 (less 0.9 -11.9245 griffiths, 0.58 -11.9245 regards 0.34507 -11.9246 (2007), 0.944444 -11.9247 publicly available. 0.685714 -11.9247 subjects were 0.9 -11.9248 literals 0.944444 -11.9249 cer 0.293814 -11.9251 assigning 0.675676 -11.9256 placing 0.758621 -11.9258 while still 0.944444 -11.926 emails. 0.9 -11.9261 three-way 0.65 -11.9261 medicine 0.545455 -11.9262 protocol 0.675676 -11.9264 phrase pairs. 0.944444 -11.9268 tested, 0.833333 -11.9268 yamada and 0.833333 -11.9268 menu 0.863636 -11.9269 78.2 0.144458 -11.9271 makes 0.4 -11.9272 apply to 0.383929 -11.9273 extracted by 0.944444 -11.9275 63% 0.944444 -11.9279 expose 0.9 -11.928 e-step 0.944444 -11.9281 32.6 0.9 -11.9282 domain selection 0.287129 -11.9283 and ney, 0.613636 -11.9284 onset 0.9 -11.9284 doctor 0.9 -11.9285 model-based 0.733333 -11.9286 stephan 0.306818 -11.9289 with no 0.944444 -11.929 second life 0.516129 -11.9292 barzilay 0.9 -11.9294 (word 0.583333 -11.9295 are indeed 0.287129 -11.9297 coefficient 0.944444 -11.9298 regulating 0.634146 -11.9298 1992) 0.685714 -11.9301 hazard 0.833333 -11.9301 binary, 0.733333 -11.9304 14: 0.944444 -11.9305 28.6 0.758621 -11.9307 maxpaths 0.516129 -11.9307 created from 0.944444 -11.9311 noisy-channel 0.445783 -11.9312 many different 0.944444 -11.9313 many-to-one 0.944444 -11.9313 syntagmatic 0.944444 -11.9314 nuanced 0.944444 -11.9314 2.1) 0.863636 -11.9317 hearst 0.153955 -11.9318 line 0.324841 -11.9318 are defined 0.634146 -11.9319 brand 0.944444 -11.932 pertinent 0.944444 -11.9324 nu-lex 0.944444 -11.9324 common, 0.8 -11.9325 confusion set 0.583333 -11.9327 formulas 0.583333 -11.9327 different languages. 0.944444 -11.9328 negligible 0.944444 -11.9328 49% 0.863636 -11.9329 sites, 0.733333 -11.933 economic 0.833333 -11.9331 know what 0.431818 -11.9332 before, 0.863636 -11.9332 romance 0.944444 -11.9333 expands 0.9 -11.9337 entities: 0.5 -11.9338 mine 0.534483 -11.9342 been studied 0.944444 -11.9343 encouraging, 0.944444 -11.9343 documents; 0.944444 -11.9343 etc) 0.863636 -11.9343 control over 0.279621 -11.9345 constant 0.71875 -11.9345 prlg 0.583333 -11.9347 input sentence. 0.944444 -11.935 (thus 0.350746 -11.935 facilitate 0.545455 -11.935 evaluations. 0.685714 -11.9355 solves 0.863636 -11.9356 [3] 0.155072 -11.9357 by using 0.9 -11.936 texts: 0.362205 -11.936 results are shown in 0.758621 -11.9363 wo 0.944444 -11.9363 robustly 0.944444 -11.9363 n-gram-based 0.777778 -11.9363 lexical entries 0.154286 -11.9364 treebank 0.758621 -11.9365 hybrid approach 0.944444 -11.9365 eventual 0.944444 -11.9367 69.7 0.0825163 -11.9367 large 0.944444 -11.9369 p5 0.944444 -11.9369 (2009), who 0.634146 -11.9369 dp 0.0621534 -11.9369 test 0.944444 -11.9371 pooling 0.944444 -11.9371 69.8 0.944444 -11.9373 meaning; 0.944444 -11.9375 segment-level 0.9 -11.9375 at test time, 0.944444 -11.9376 improved, 0.944444 -11.9378 chinese–english 0.944444 -11.9378 191 0.9 -11.938 bard 0.944444 -11.9382 57.9 0.944444 -11.9382 distrib 0.8 -11.9383 engineer 0.758621 -11.9384 deletion, 0.685714 -11.9387 des 0.195402 -11.9387 here, 0.71875 -11.9387 provided. 0.944444 -11.9388 trade-offs 0.944444 -11.9388 parent-child 0.944444 -11.9388 89.1 0.113338 -11.9388 current 0.944444 -11.939 awareness 0.461538 -11.939 target. 0.203474 -11.939 avoid 0.685714 -11.9391 two variants 0.331126 -11.9392 arg 0.8 -11.9392 ips 0.944444 -11.9393 grading 0.944444 -11.9395 persistent 0.657895 -11.9395 frames. 0.944444 -11.9397 audiovisual 0.534483 -11.9398 spectral 0.944444 -11.9399 74.5 0.944444 -11.9399 w) 0.944444 -11.9401 44.0 0.944444 -11.9401 role labelers 0.944444 -11.9401 (lopez, 0.944444 -11.9401 estimate, 0.9 -11.9402 bayes’ 0.657895 -11.9403 online learning 0.944444 -11.9405 spectral clustering 0.944444 -11.9405 63.5 0.492537 -11.9405 scene 0.944444 -11.9407 1986) 0.944444 -11.9407 alphabetic 0.944444 -11.9408 ancestors 0.944444 -11.941 novel, 0.944444 -11.941 time) 0.944444 -11.941 manifest 0.445783 -11.9411 mstparser 0.534483 -11.9412 participants. 0.634146 -11.9413 espresso 0.45 -11.9415 was used for 0.9 -11.9416 satisfied, 0.568627 -11.9416 to aid 0.863636 -11.9417 roleset 0.944444 -11.9418 38% 0.516129 -11.9418 evidences 0.863636 -11.9419 fifteen 0.944444 -11.9422 models) 0.944444 -11.9422 phonotactic 0.944444 -11.9424 interference 0.944444 -11.9425 conjunction, 0.944444 -11.9425 watermarked 0.944444 -11.9425 92.1 0.944444 -11.9427 framed 0.944444 -11.9427 systems) 0.431818 -11.9429 comprises 0.944444 -11.9429 harm 0.619048 -11.943 dataset contains 0.634146 -11.943 for example: 0.359375 -11.943 thereby 0.944444 -11.9431 translational 0.944444 -11.9431 overcomes 0.944444 -11.9433 validates 0.863636 -11.9434 efficacy of 0.9 -11.9436 protein modifications 0.237931 -11.9436 native 0.863636 -11.9436 agentive 0.641026 -11.9438 widespread 0.466667 -11.9438 10: 0.208995 -11.9438 2005) 0.45 -11.944 the following: 0.944444 -11.9441 sn 0.944444 -11.9443 region, 0.944444 -11.9443 story, 0.685714 -11.9443 0.07 0.777778 -11.9443 data sources 0.241993 -11.9443 algorithm, 0.944444 -11.9444 ancient 0.944444 -11.9444 areas, 0.944444 -11.9444 flexible, 0.944444 -11.9446 binning 0.944444 -11.9446 56.5 0.944444 -11.9446 train/test 0.944444 -11.9446 facing 0.944444 -11.9446 (steedman, 0.657895 -11.9447 lemmas, 0.8 -11.9447 concatenate 0.944444 -11.9448 arcs, 0.230519 -11.9449 extension 0.619048 -11.9451 stamatatos, 0.372881 -11.9451 example: 0.123049 -11.9451 had 0.944444 -11.9452 efforts, 0.944444 -11.9452 verses 0.944444 -11.9456 85.0 0.685714 -11.9458 semantic frame 0.641026 -11.946 (em) 0.507937 -11.946 front 0.944444 -11.9462 postpositions 0.69697 -11.9462 negative) 0.141636 -11.9462 approach to 0.944444 -11.9463 splitter 0.944444 -11.9463 configured 0.944444 -11.9463 unstable 0.478873 -11.9463 0.001 0.174397 -11.9463 figure 1 0.944444 -11.9465 relevant, 0.944444 -11.9465 authentic 0.944444 -11.9467 culotta 0.944444 -11.9469 attain 0.944444 -11.9469 backchannels, 0.365854 -11.9469 betwen 0.944444 -11.9471 (bb) 0.944444 -11.9471 slice 0.863636 -11.9472 n/a 0 0.583333 -11.9474 ibm1 0.944444 -11.9475 2008b), 0.152758 -11.9475 done 0.777778 -11.9475 append 0.362903 -11.9475 latent semantic 0.9 -11.9476 wid 0.944444 -11.9477 resolves 0.944444 -11.9477 fused 0.944444 -11.9479 6.7 0.944444 -11.9479 harmful 0.485714 -11.9479 exploration 0.944444 -11.9482 70.4 0.9 -11.9483 (labeled 0.944444 -11.9484 envision 0.535714 -11.9485 origin 0.733333 -11.9485 composed rules 0.944444 -11.9486 66% 0.944444 -11.9486 callers 0.619048 -11.9486 ∝ 0.944444 -11.9488 coincides 0.944444 -11.949 newspapers, 0.944444 -11.9492 confuse 0.944444 -11.9492 stars 0.944444 -11.9492 therapy 0.863636 -11.9493 sales 0.944444 -11.9494 (dis)agreement 0.666667 -11.9495 (finkel 0.863636 -11.95 document retrieval 0.944444 -11.9502 suitability 0.863636 -11.9502 user-generated 0.545455 -11.9502 built using 0.826087 -11.9503 executing 0.944444 -11.9503 blank 0.944444 -11.9505 quotations 0.220238 -11.9505 ter 0.516667 -11.9506 dramatically 0.944444 -11.9507 brill 0.944444 -11.9507 afterwards, 0.320755 -11.9508 viewed 0.944444 -11.9509 mature 0.944444 -11.9509 lower bound 0.944444 -11.9509 dd-admm 0.944444 -11.9509 rewarding 0.944444 -11.9511 informally, 0.944444 -11.9511 garden-path 0.8 -11.9511 team, 0.557692 -11.9511 ghkm 0.944444 -11.9515 instability 0.944444 -11.9517 shifting 0.944444 -11.9517 disparate 0.545455 -11.9517 bitext 0.408163 -11.9518 evidence that 0.944444 -11.9519 aim, 0.944444 -11.9519 fulfill 0.944444 -11.9523 68% 0.545455 -11.9523 grateful to 0.944444 -11.9524 aproaches 0.944444 -11.9524 acronyms, 0.583333 -11.9524 sort of 0.110289 -11.9524 languages 0.944444 -11.9526 subproblems 0.9 -11.9529 selectively 0.666667 -11.9529 straightforward. 0.8 -11.9529 3,000 0.826087 -11.953 lk 0.944444 -11.953 verb-object 0.15625 -11.9531 introduce 0.944444 -11.9532 implementation details 0.944444 -11.9532 collections, 0.944444 -11.9532 seg 0.6 -11.9532 similarity metric 0.826087 -11.9535 frequency threshold 0.944444 -11.9536 coreferential 0.30814 -11.9536 stack 0.944444 -11.954 react 0.944444 -11.954 87.9 0.944444 -11.954 procedures, 0.19715 -11.9541 because of 0.6 -11.9541 can be improved 0.944444 -11.9545 (cohen, 0.944444 -11.9545 94.5 0.9 -11.9545 p(w) 0.944444 -11.9547 opca 0.944444 -11.9547 reliable, 0.944444 -11.9547 ages 0.324675 -11.9548 advantages 0.733333 -11.9549 fitness 0.863636 -11.9549 header 0.419355 -11.9549 9: 0.863636 -11.9551 approximate string 0.863636 -11.9551 facts. 0.944444 -11.9553 synergy 0.9 -11.9554 $rm 0.944444 -11.9557 conjunctions, 0.944444 -11.9557 color, 0.30814 -11.9558 to help 0.826087 -11.9559 (via 0.944444 -11.9559 generalised 0.0801257 -11.9559 similarity 0.777778 -11.956 comparable documents 0.9 -11.9561 (named 0.944444 -11.9563 obstacle 0.944444 -11.9563 elsewhere 0.583333 -11.9564 0.68 0.944444 -11.9565 positions, 0.944444 -11.9565 192 0.944444 -11.9565 ti, 0.944444 -11.9567 pd 0.944444 -11.9569 competition 0.9 -11.9569 consistent improvements 0.733333 -11.957 interpersonal 0.944444 -11.9572 australian 0.944444 -11.9572 dld 0.944444 -11.9572 (xu 0.777778 -11.9572 amenable 0.944444 -11.9574 lafferty 0.460526 -11.9574 vector, 0.944444 -11.9576 (1995), 0.944444 -11.9578 failing 0.206806 -11.9578 means that 0.944444 -11.958 184 0.944444 -11.958 permitted 0.944444 -11.958 emoticons, 0.777778 -11.9583 l-lda 0.944444 -11.9584 wordbased 0.525424 -11.9587 positive examples 0.472222 -11.9587 might have 0.944444 -11.9588 b2 0.826087 -11.9589 accuracy: 0.333333 -11.9591 affected 0.944444 -11.9592 libraries 0.583333 -11.9592 connectives. 0.136054 -11.9593 few 0.944444 -11.9595 deployment 0.944444 -11.9595 305 0.619048 -11.9595 wisdom 0.777778 -11.9595 maise 0.11303 -11.9596 types of 0.641026 -11.9598 an approximate 0.944444 -11.9599 friday 0.944444 -11.9599 tn 0.944444 -11.9599 dialects, 0.777778 -11.9601 rights 0.9 -11.9602 mm 0.666667 -11.9602 human annotation 0.944444 -11.9603 btec 0.826087 -11.9604 analogously 0.944444 -11.9605 t2, 0.944444 -11.9605 soft em 0.944444 -11.9605 continuum 0.944444 -11.9607 o(n) 0.944444 -11.9607 owing to 0.944444 -11.9609 tokenize 0.944444 -11.9611 4.2, 0.535714 -11.9612 0.66 0.944444 -11.9613 deduce 0.944444 -11.9613 string-to-string 0.944444 -11.9615 trigrams, 0.944444 -11.9615 databases, 0.8 -11.9615 coref 0.69697 -11.9617 further research 0.944444 -11.9619 tagsets 0.944444 -11.9624 stressed 0.944444 -11.9626 grawltcq 0.6 -11.9626 an accurate 0.283582 -11.9629 turns 0.826087 -11.9629 2-gram 0.9 -11.9631 snyder et al. 0.944444 -11.9632 shown below. 0.944444 -11.9634 morphemes, 0.944444 -11.9634 (nelson 0.709677 -11.9634 for comparison. 0.944444 -11.9636 ft 0.586957 -11.9636 to determine if 0.944444 -11.9638 perceive 0.31875 -11.9638 retrieve 0.944444 -11.9642 entities) 0.944444 -11.9644 cooperation 0.944444 -11.9644 transcription, 0.944444 -11.9644 initiated 0.241007 -11.9644 parent 0.826087 -11.9648 quantifiers, 0.826087 -11.9651 rmd 0.944444 -11.9651 nonnegative 0.944444 -11.9653 joshi 0.857143 -11.9655 phrase-pairs 0.944444 -11.9655 evolutionary 0.944444 -11.9655 1for 0.8 -11.9656 noisy text 0.826087 -11.9658 clitic 0.465753 -11.966 we also use 0.571429 -11.966 time-consuming 0.6 -11.966 then applied 0.944444 -11.9661 “bad” 0.944444 -11.9661 son 0.944444 -11.9661 among them, 0.944444 -11.9661 evidenced 0.8 -11.9663 0.97 0.709677 -11.9663 rep 0.944444 -11.9667 linker 0.944444 -11.9667 (algorithm 0.586957 -11.9667 investigate how 0.9 -11.9668 significance. 0.8 -11.9668 eisenstein 0.944444 -11.9669 happened 0.944444 -11.9669 tell me 0.944444 -11.9669 december 0.422222 -11.9669 other types 0.40625 -11.967 development and test 0.944444 -11.9671 equivalent, 0.944444 -11.9671 mcmc 0.944444 -11.9671 delay 0.944444 -11.9671 association, 0.857143 -11.9674 (ramage et 0.944444 -11.9675 decimal 0.625 -11.9675 sc 0.69697 -11.9677 fora 0.8 -11.9679 pang et al. 0.625 -11.968 crowdsourced 0.944444 -11.9682 adaptable 0.386792 -11.9683 j=1 0.54717 -11.9683 world. 0.472222 -11.9685 dots 0.944444 -11.9688 structure: 0.944444 -11.9688 (pos 0.625 -11.9689 more closely 0.54717 -11.9696 mr 0.465753 -11.9696 event types 0.826087 -11.9697 dialogue history 0.676471 -11.9697 fitting 0.857143 -11.9701 window, 0.944444 -11.9702 deletes 0.56 -11.9705 (s 0.944444 -11.9706 73% 0.676471 -11.9706 7% 0.944444 -11.9712 w1, 0.944444 -11.9712 bins, 0.183333 -11.9713 project 0.857143 -11.9715 matcher 0.328859 -11.9715 aim to 0.944444 -11.9716 ilp-based 0.944444 -11.9716 insignificant 0.944444 -11.9717 prefixes, 0.769231 -11.9717 meetings 0.571429 -11.9717 ek 0.485294 -11.9719 1994). 0.944444 -11.9721 17: 0.944444 -11.9721 fills 0.826087 -11.9724 tri-sibling 0.516667 -11.9725 supertagger 0.944444 -11.9725 subscripts 0.944444 -11.9725 tasa 0.944444 -11.9725 subproblem 0.641026 -11.9725 also included 0.0505547 -11.9727 number of 0.709677 -11.973 specific, 0.944444 -11.9731 wsj, 0.243542 -11.9731 anonymous 0.69697 -11.9731 cost function 0.826087 -11.9733 originating 0.857143 -11.9734 us government. 0.56 -11.9734 innovation 0.944444 -11.9737 non-lexical 0.826087 -11.974 .75 0.314815 -11.974 international conference 0.944444 -11.9741 context) 0.826087 -11.9743 61.1 0.944444 -11.9743 facts, 0.648649 -11.9745 morphology. 0.944444 -11.9747 passive-aggressive 0.944444 -11.9747 formulation, 0.69697 -11.9747 counters 0.8 -11.9747 crowdflower 0.8 -11.9748 details, 0.944444 -11.9749 imagined 0.641026 -11.9749 animal 0.422222 -11.975 harmonic 0.894737 -11.975 calibration 0.944444 -11.9751 196 0.75 -11.9751 .05 0.537037 -11.9754 curran, 0.894737 -11.9754 order: 0.944444 -11.9755 (x) 0.944444 -11.9755 189 0.417582 -11.9756 hypotheses. 0.0989615 -11.9756 original 0.126494 -11.9757 cases 0.894737 -11.9758 variant, 0.857143 -11.9759 empty category 0.75 -11.9761 machine learning approaches 0.574468 -11.9763 speaker. 0.944444 -11.9764 195 0.826087 -11.9765 copies of 0.894737 -11.9765 zamparelli 0.944444 -11.9766 intuition, 0.944444 -11.9766 “it 0.894737 -11.9767 monotonicity 0.944444 -11.9768 structurally 0.826087 -11.9769 whatever 0.944444 -11.977 linguatec 0.826087 -11.9772 non-understanding 0.826087 -11.9772 nakov and 0.944444 -11.9772 mini-tacitus 0.345865 -11.9772 small set 0.944444 -11.9774 4for 0.944444 -11.9774 (21) 0.894737 -11.9774 mfs 0.526316 -11.9775 wj 0.492308 -11.9776 aforementioned 0.944444 -11.9776 does, 0.944444 -11.9776 arrow 0.894737 -11.9776 n-v 0.826087 -11.9777 subtyping 0.5 -11.9777 turn, 0.289474 -11.9778 applies 0.894737 -11.9787 tprime 0.894737 -11.9791 episodes 0.75 -11.9792 word lattice 0.676471 -11.9795 antecedent. 0.894737 -11.9795 track, 0.439024 -11.9796 adapted to 0.857143 -11.98 safe 0.351562 -11.98 fail 0.894737 -11.98 parsing: 0.894737 -11.98 below), 0.676471 -11.9802 pairs: 0.439024 -11.9802 attempted 0.208791 -11.9803 (for 0.857143 -11.9806 nom 0.8 -11.9808 ill-formed words 1 -11.981 perfect. 1 -11.981 helpful. 1 -11.981 reranking. 1 -11.981 wi. 1 -11.981 informative. 1 -11.981 crf. 1 -11.981 represented. 1 -11.981 creation. 1 -11.981 technologies. 1 -11.981 clues. 1 -11.981 max. 1 -11.981 consistent. 1 -11.981 formalism. 1 -11.981 columns. 1 -11.981 modifier. 1 -11.981 proficiency. 1 -11.981 access. 1 -11.981 identify. 1 -11.981 precision-recall curves 1 -11.981 lines. 1 -11.981 row. 1 -11.981 eqn. 1 -11.981 permutations. 1 -11.981 percentages. 1 -11.981 dredze 1 -11.981 neighbors. 1 -11.981 philipp koehn, 1 -11.981 agent. 1 -11.981 disjoint. 1 -11.981 families. 1 -11.981 board. 1 -11.981 blogs. 1 -11.981 assistive technologies, 1 -11.981 estimates. 1 -11.981 google. 1 -11.981 e2. 1 -11.981 randomly. 1 -11.981 keystroke savings 1 -11.981 5.2. 1 -11.981 conditionally independent 1 -11.981 negatives. 1 -11.981 sacrificing 1 -11.981 categorization. 1 -11.981 bnc. 1 -11.981 ranks. 0.143041 -11.9812 end 0.941176 -11.9812 orders. 0.941176 -11.9812 linked. 0.941176 -11.9812 least squares 0.941176 -11.9812 lexical, syntactic, 0.857143 -11.9814 similarity measure. 0.348837 -11.9814 interestingly, 0.857143 -11.9816 equation: 0.351562 -11.9817 probable 0.383178 -11.9818 functions. 0.857143 -11.982 kumar 0.941176 -11.982 discriminative tfidf 0.492308 -11.9824 combinations. 0.941176 -11.9824 unnatural 0.941176 -11.9824 lagrange multipliers 0.941176 -11.9826 provided, 0.894737 -11.9826 l., 0.386792 -11.9827 statement 0.826087 -11.9829 instructed to 0.676471 -11.9834 4.2.1 0.894737 -11.9834 benjamin 0.894737 -11.9836 tsp 0.941176 -11.9838 unbiased 0.894737 -11.9839 ru 0.894737 -11.9841 aditional 0.433735 -11.9843 says 0.941176 -11.9844 propbank (palmer 0.941176 -11.9844 syntax, semantics 0.345865 -11.9845 nouns. 0.941176 -11.9847 (figure 3). 0.164384 -11.9849 factor 0.941176 -11.9849 (prasad 0.941176 -11.9851 burkett 0.941176 -11.9853 monotone submodular 0.894737 -11.9854 dependency trees, 0.724138 -11.9855 wordnet-based 0.625 -11.9855 goldberg 0.894737 -11.9856 1-3 0.894737 -11.9858 prefix, 0.857143 -11.9859 75.6 0.625 -11.9859 formalize 0.459459 -11.986 vectors, 0.894737 -11.986 ps 0.425287 -11.9861 { 0.75 -11.9867 examines 0.769231 -11.9869 trajectories 0.586957 -11.9869 np. 0.769231 -11.987 dfki 0.75 -11.987 erk and 0.941176 -11.9873 formally define 0.941176 -11.9873 protein catabolism 0.316456 -11.9875 operators 0.894737 -11.9877 peng 0.941176 -11.9879 (birch 0.894737 -11.9881 linked web 0.857143 -11.9883 concordu 0.941176 -11.9883 λt 0.941176 -11.9883 replications 0.941176 -11.9883 (see section 4). 0.894737 -11.9883 1this 0.282828 -11.9885 illustrate 0.0994371 -11.9885 parse 0.941176 -11.9885 brants 0.857143 -11.9888 sigmoid 0.941176 -11.9889 stopping criterion 0.941176 -11.9891 argamon 0.894737 -11.9892 criteria: 0.941176 -11.9893 maximum-likelihood 0.941176 -11.9893 reinforce 0.941176 -11.9893 canonical forms 0.101233 -11.9894 contains 0.604651 -11.9894 an adjective 0.941176 -11.9895 26.3 0.769231 -11.9896 improves performance 0.123671 -11.9897 did 0.941176 -11.9897 abstractions 0.941176 -11.9897 valued 0.625 -11.9898 belief propagation 0.857143 -11.9899 successful, 0.857143 -11.9902 spelling, 0.941176 -11.9905 exactly, 0.941176 -11.9905 multi-document summarization. 0.941176 -11.9905 institutions 0.894737 -11.9905 filter, 0.894737 -11.9905 entrez 0.941176 -11.9909 habitat 0.941176 -11.9909 automated scoring 0.857143 -11.9911 [squaresmallsolid 0.433735 -11.9911 development data 0.648649 -11.9914 goal, 0.352 -11.9914 was also 0.941176 -11.9915 (ritter 0.604651 -11.9917 bill 0.857143 -11.9919 (two 0.204244 -11.9919 advantage of 0.448718 -11.9923 integrates 0.941176 -11.9923 watermarking 0.103309 -11.9925 result 0.609756 -11.9925 sentences, but 0.826087 -11.9926 bond 0.526316 -11.9929 cr 0.769231 -11.993 20k 0.791667 -11.993 word-pairs 0.941176 -11.993 patient’s 0.724138 -11.9931 hw 0.471429 -11.9932 collaboration 0.791667 -11.9933 wine 0.894737 -11.9933 petrov et al. 0.857143 -11.9935 an optional 0.894737 -11.9935 healthcare 0.894737 -11.9937 text summarization. 0.941176 -11.9938 85.7 0.941176 -11.9938 jones, 0.941176 -11.9938 (ratnaparkhi, 0.941176 -11.9938 weigh 0.6875 -11.994 2003. 0.769231 -11.9943 arc-standard 0.105749 -11.9943 association for computational 0.769231 -11.9945 lazy 0.769231 -11.9946 sys3 0.413043 -11.9947 c2 0.941176 -11.9948 (example 0.857143 -11.9951 refinements 0.307229 -11.9953 al., 2008), 0.941176 -11.9954 croft, 0.941176 -11.9954 yao 0.894737 -11.9954 appreciate 0.857143 -11.9955 assert 0.941176 -11.9956 28.7 0.941176 -11.9956 6.6 0.941176 -11.9956 based, 0.941176 -11.9956 low-dimensional 0.150507 -11.9956 always 0.724138 -11.9957 2d 0.791667 -11.9957 conf 0.857143 -11.9958 wikitopics 0.941176 -11.9958 gram 0.258621 -11.9959 all other 0.791667 -11.996 2007)). 0.941176 -11.996 (marcu 0.941176 -11.9962 kop 0.894737 -11.9962 traditionally, 0.894737 -11.9962 shorthand 0.648649 -11.9963 transition-based dependency 0.574468 -11.9964 dotted 0.941176 -11.9964 active, 0.941176 -11.9964 abundance 0.676471 -11.9966 induces a 0.941176 -11.9966 homonymous 0.941176 -11.9966 designer 0.471429 -11.9967 connotation 0.791667 -11.9968 no, 0.941176 -11.9968 237 0.625 -11.997 locality 0.941176 -11.9972 verbnet, 0.941176 -11.9972 pitch, 0.604651 -11.9973 twitter. 0.894737 -11.9973 j-prf 0.56 -11.9974 part. 0.941176 -11.9974 information; 0.38835 -11.9975 formulate 0.941176 -11.9976 plausibility 0.941176 -11.9976 corroborating 0.941176 -11.9976 word-alignment 0.941176 -11.9978 bipartite matching 0.941176 -11.9978 personalization 0.941176 -11.998 ldah-s 0.941176 -11.9982 coda 0.941176 -11.9988 uml 0.941176 -11.9988 expensive, 0.791667 -11.999 hier 0.941176 -11.999 unsuccessful 0.709677 -11.999 naacl 0.657143 -11.9991 mp 0.590909 -11.9991 implied 0.769231 -11.9992 implications for 0.54902 -11.9992 tools, 0.941176 -11.9992 delayed 0.941176 -11.9992 200, 0.941176 -11.9992 68.6 0.941176 -11.9992 fragment, 0.941176 -11.9994 sekine, 0.826087 -11.9996 organization. 0.857143 -11.9998 16: 0.6875 -11.9998 an auxiliary 0.941176 -11.9998 favour 0.941176 -11.9998 (pustejovsky 0.941176 -12 70.2 0.941176 -12 pasca, 0.709677 -12 predictable 0.826087 -12.0001 tutor. 0.941176 -12.0002 scholars 0.5 -12.0003 actions. 0.857143 -12.0004 bloggers 0.941176 -12.0004 decent 0.941176 -12.0004 mmr 0.709677 -12.0005 schemes, 0.22973 -12.0008 errors. 0.941176 -12.0008 placement 0.941176 -12.0008 dependency, 0.941176 -12.0008 divergent 0.609756 -12.0009 disadvantage 0.121789 -12.001 train 0.56 -12.001 are commonly 0.941176 -12.001 established, 0.769231 -12.0011 statistical parsing 0.631579 -12.0011 similarity metrics 0.941176 -12.0012 malta, 0.941176 -12.0012 carreras 0.857143 -12.0013 figurative 0.941176 -12.0014 gillenwater 0.941176 -12.0014 generalise 0.941176 -12.0018 score; 0.941176 -12.0018 samplerank training 0.894737 -12.0019 lexical items. 0.941176 -12.002 transcripts, 0.590909 -12.0022 conducted on 0.941176 -12.0022 optimised 0.941176 -12.0022 multir 0.231034 -12.0024 is often 0.791667 -12.0024 categorial grammar 0.941176 -12.0024 tarau, 0.941176 -12.0024 qazvinian 0.676471 -12.0025 track. 0.192399 -12.0025 ways 0.197007 -12.0026 it was 0.941176 -12.0026 forecasting 0.941176 -12.0026 margin, 0.941176 -12.0026 paraphrasing, 0.857143 -12.0027 attardi 0.941176 -12.003 duan 0.941176 -12.0032 institute, 0.941176 -12.0032 sch¨utze, 0.857143 -12.0033 u1 0.941176 -12.0034 automate 0.941176 -12.0034 unmatched 0.941176 -12.0034 nombank 0.894737 -12.0034 multistream 0.941176 -12.0036 turk, 0.941176 -12.0036 drda 0.941176 -12.0038 predicateargument 0.791667 -12.0039 distribute 0.941176 -12.004 54.3 0.894737 -12.004 relation: 0.1975 -12.0041 way to 0.941176 -12.0042 indices, 0.941176 -12.0042 25.5 0.894737 -12.0042 ap, 0.941176 -12.0044 experiences 0.941176 -12.0044 adhere to 0.267281 -12.0046 combines 0.941176 -12.0046 gain, 0.791667 -12.0047 dependency parser, 0.235714 -12.0047 we show that 0.857143 -12.0048 fj 0.941176 -12.0048 archaic 0.857143 -12.0049 while, 0.453333 -12.005 (this 0.894737 -12.0051 this area. 0.941176 -12.0052 status, 0.941176 -12.0052 rasp 0.894737 -12.0053 indian language 0.894737 -12.0055 ci, 0.631579 -12.0055 four types 0.676471 -12.0057 will always 0.538462 -12.0057 = 0. 0.537037 -12.0058 every sentence 0.941176 -12.006 65.7 0.941176 -12.006 55.1 0.590909 -12.0061 samplerank 0.201044 -12.0061 tool 0.941176 -12.0062 non-convex 0.941176 -12.0064 above) 0.537037 -12.0065 contributed 0.657143 -12.0066 canada 0.941176 -12.0066 scatter 0.941176 -12.0066 2010a), 0.484848 -12.0067 author’s 0.941176 -12.0068 one) 0.941176 -12.007 “x” 0.941176 -12.007 multilayer 0.527273 -12.0073 noise. 0.517241 -12.0073 mix 0.941176 -12.0074 unk 0.464789 -12.0077 we showed that 0.361345 -12.0077 variance 0.338235 -12.0078 aware 0.941176 -12.008 qa-sys 0.941176 -12.008 hubs 0.857143 -12.0081 simple heuristic 0.941176 -12.0084 (4 0.894737 -12.0084 computing, 0.941176 -12.0086 disjunction 0.941176 -12.0086 1300 0.941176 -12.0086  0.791667 -12.0088 lowering 0.941176 -12.0088 agglutinative 0.857143 -12.009 logical structure 0.941176 -12.0091 wide-coverage 0.941176 -12.0091 0.3, 0.941176 -12.0091 occurs, 0.857143 -12.0093 zone 0.941176 -12.0093 82.5 0.195062 -12.0093 2.1 0.10515 -12.0094 c 0.740741 -12.0094 gpe 0.941176 -12.0095 solvers 0.432099 -12.0096 located 0.941176 -12.0097 iconic 0.941176 -12.0097 78.3 0.791667 -12.0098 ptm 0.941176 -12.0099 35.0 0.941176 -12.0099 alignment-based 0.941176 -12.0099 178 0.941176 -12.0099 depth, 0.941176 -12.0101 departure 0.657143 -12.0103 rd 0.769231 -12.0103 west 0.941176 -12.0103 172 0.5625 -12.0104 significantly improve 0.941176 -12.0107 17% 0.941176 -12.0109 designs 0.941176 -12.0109 radius 0.941176 -12.0109 71.2 0.941176 -12.0109 45.5 0.894737 -12.011 shift-reduce parsing 0.941176 -12.0111 interannotator agreement 0.941176 -12.0111 paragraphs, 0.857143 -12.0112 commit 0.941176 -12.0113 samt-style 0.941176 -12.0113 taxonomy, 0.941176 -12.0115 soil 0.423529 -12.0115 categories: 0.508475 -12.0115 predicted by 0.941176 -12.0117 sentence-based 0.857143 -12.0119 encounters 0.4 -12.0119 drops 0.941176 -12.0119 justification 0.941176 -12.0119 ecu 0.941176 -12.0121 diachronic 0.941176 -12.0121 distributive 0.857143 -12.0125 spelled 0.791667 -12.0125 jonathan 0.941176 -12.0125 phrase) 0.330935 -12.0125 to decide 0.857143 -12.0127 :) 0.818182 -12.0127 plane 0.590909 -12.0127 many nlp 0.791667 -12.0127 f(y) 0.941176 -12.0127 87% 0.941176 -12.0127 circumstances 0.609756 -12.0129 wtm 0.941176 -12.0129 non-anaphoric 0.857143 -12.013 training examples, 0.941176 -12.0131 (2b) 0.818182 -12.0132 [2] 0.577778 -12.0133 bigrams, 0.941176 -12.0133 pointers 0.941176 -12.0135 ids 0.941176 -12.0135 termination 0.330935 -12.0135 0.3 0.941176 -12.0137 scus 0.941176 -12.0137 institutional 0.941176 -12.0137 52.2 0.355372 -12.0139 sources of 0.724138 -12.014 event structures 0.39 -12.014 this experiment 0.941176 -12.0141 afford 0.941176 -12.0141 complementing 0.941176 -12.0141 treebanks, 0.857143 -12.0143 sociolinguistic 0.857143 -12.0145 dependency arcs 0.941176 -12.0145 modification, 0.941176 -12.0145 clients 0.941176 -12.0145 (right 0.941176 -12.0145 complexities 0.470588 -12.0147 f, 0.941176 -12.0147 throughput 0.941176 -12.0147 labeledlda 0.941176 -12.0149 participant’s 0.941176 -12.0149 400k 0.894737 -12.0149 documents: 0.173913 -12.015 spoken 0.769231 -12.0151 freq 0.941176 -12.0151 252 0.894737 -12.0151 mind. 0.527273 -12.0151 note, 0.941176 -12.0153 request, 0.941176 -12.0153 lg 0.857143 -12.0154 separable 0.6875 -12.0154 code, 0.769231 -12.0155 nan 0.724138 -12.0155 context matching 0.941176 -12.0155 intersecting 0.6875 -12.0156 an element 0.941176 -12.0157 layout 0.941176 -12.0157 rigid 0.941176 -12.0157 wizard-of-oz 0.941176 -12.0157 multi-output 0.538462 -12.0157 syntactic parsing 0.517857 -12.0159 engine. 0.941176 -12.0161 occurrences, 0.470588 -12.0164 similarity scores 0.857143 -12.0167 aspectual 0.941176 -12.0167 openccg 0.791667 -12.0168 62.4 0.470588 -12.0169 threshold, 0.941176 -12.0169 regmt 0.941176 -12.0169 y(i) = 0.941176 -12.0169 verbally 0.769231 -12.0171 pseudocode 0.941176 -12.0171 completed, 0.941176 -12.0174 grown 0.857143 -12.0175 0). 0.857143 -12.0175 discovery. 0.941176 -12.0176 crf-based 0.941176 -12.0176 sketched 0.55102 -12.0178 cues. 0.941176 -12.0178 temporal, 0.631579 -12.018 intervening 0.194581 -12.0183 similarity between 0.941176 -12.0184 equal, 0.941176 -12.0184 valuable resource 0.941176 -12.0184 culture, 0.941176 -12.0184 can, 0.818182 -12.0185 sp 0.941176 -12.0186 extension, 0.941176 -12.0186 simulation-based 0.657143 -12.0187 input sentence, 0.214502 -12.0187 keep 0.941176 -12.0188 arrived 0.941176 -12.0188 0.800 0.941176 -12.0188 for brevity, 0.0519116 -12.0191 sentence 0.740741 -12.0191 repeated until 0.492063 -12.0191 intuition that 0.941176 -12.0192 verifying 0.894737 -12.0193 fair comparison 0.538462 -12.0193 german. 0.6875 -12.0194 1− 0.941176 -12.0194 compilation 0.261261 -12.0194 entities. 0.894737 -12.0195 (here 0.818182 -12.0196 (hearst, 0.195 -12.0197 tagged 0.740741 -12.0197 1990), 0.941176 -12.0198 nn, 0.941176 -12.0198 schematic 0.5625 -12.0199 (except 0.941176 -12.02 himself 0.941176 -12.02 87.8 0.6875 -12.0202 distributional models 0.740741 -12.0202 0.47 0.769231 -12.0203 _ 0.609756 -12.0204 we followed 0.941176 -12.0204 archive 0.14324 -12.0204 token 0.470588 -12.0206 tied 0.941176 -12.0206 208 0.577778 -12.0207 0.57 0.941176 -12.0212 not surprisingly, 0.55102 -12.0214 (zhao 0.609756 -12.0214 subjects. 0.0560786 -12.0214 models 0.941176 -12.0214 aren’t 0.818182 -12.0215 af 0.941176 -12.0216 rel task 0.5625 -12.0218 syntactic, 0.4375 -12.0218 nlu 0.941176 -12.0218 features), 0.941176 -12.022 (forbes-riley 0.941176 -12.022 formatting 0.240458 -12.0222 namely 0.941176 -12.0224 exemplifies 0.941176 -12.0227 hiberno-english 0.941176 -12.0227 fold, 0.941176 -12.0227 verse 0.941176 -12.0229 equipped 0.941176 -12.0229 borrow 0.941176 -12.0229 bell 0.6875 -12.023 university. 0.894737 -12.023 (henceforth, 0.470588 -12.0231 removal 0.492063 -12.0231 each instance 0.538462 -12.0232 to maintain 0.941176 -12.0233 czech-english 0.941176 -12.0237 sides, 0.941176 -12.0237 pop 0.418605 -12.0237 broadcast 0.186788 -12.0238 equal 0.941176 -12.0239 360 0.941176 -12.0239 bootcat 0.470588 -12.024 centre 0.170132 -12.024 discuss 0.941176 -12.0241 328 0.941176 -12.0241 encompass 0.818182 -12.0242 dependency parses 0.341085 -12.0243 (from 0.657143 -12.0244 ccgbank 0.941176 -12.0245 voices 0.538462 -12.0245 model; 0.941176 -12.0247 adequacy, 0.740741 -12.0249 (column 0.941176 -12.0249 3.1.2 0.941176 -12.0249 deviate 0.7 -12.0251 attribute-value 0.222222 -12.0251 dependent 0.941176 -12.0251 stick 0.638889 -12.0252 that occurred 0.595238 -12.0254 j) 0.7 -12.0255 we saw 0.740741 -12.0255 ka 0.941176 -12.0255 nfl 0.941176 -12.0255 pending 0.941176 -12.0255 caches 0.791667 -12.0256 covariance matrices 0.941176 -12.0257 predication 0.818182 -12.0258 somasundaran 0.941176 -12.0259 8the 0.894737 -12.0259 similarity measures. 0.941176 -12.0261 88% 0.818182 -12.0262 errors: 0.941176 -12.0263 (nn) 0.107332 -12.0264 main 0.159396 -12.0265 to make 0.941176 -12.0265 xk 0.538462 -12.0267 a large corpus 0.85 -12.0272 ill-formed word 0.5625 -12.0274 table 1). 0.941176 -12.0274 catalog 0.941176 -12.0274 (null) 0.941176 -12.0274 insertions, 0.941176 -12.0274 annotation: 0.941176 -12.0274 unpredictable 0.55102 -12.0276 leverages 0.321918 -12.0277 lee, 0.724138 -12.0279 detailed description of 0.941176 -12.028 llda 0.941176 -12.028 copestake 0.941176 -12.0282 8.1 0.724138 -12.0283 persuasion 0.269231 -12.0284 space, 0.941176 -12.0286 95.8 0.941176 -12.0286 ua 0.941176 -12.029 uncertain+correct 0.818182 -12.0292 ntcir 0.941176 -12.0292 3-way 0.0946502 -12.0292 sense 0.6875 -12.0294 to prune 0.941176 -12.0294 nonliteral 0.631579 -12.0294 tuning. 0.406593 -12.0295 we develop 0.290503 -12.0297 worth 0.941176 -12.0298 (1990) 0.85 -12.0302 user simulation 0.941176 -12.0305 joel 0.724138 -12.0306 training data size 0.7 -12.0306 minute 0.941176 -12.0307 9) 0.941176 -12.0307 demanding 0.941176 -12.0309 -ing 0.313725 -12.0309 likelihood of 0.615385 -12.031 topic models. 0.484375 -12.0311 devtest 0.941176 -12.0311 21% 0.941176 -12.0311 pinc 0.508772 -12.0313 for comparison, 0.149123 -12.0313 performance. 0.146515 -12.0313 a small 0.565217 -12.0314 download 0.791667 -12.0316 xiong et 0.791667 -12.0316 1989) 0.894737 -12.0316 1985). 0.120773 -12.0317 prior 0.6875 -12.0321 reading. 0.941176 -12.0321 cursor 0.5625 -12.0323 simplicity 0.273632 -12.0325 depending on the 0.55102 -12.0325 an application 0.941176 -12.0325 restrictions, 0.941176 -12.0325 348 0.941176 -12.0325 truncation 0.941176 -12.0325 medication 0.538462 -12.0325 graphics 0.491803 -12.0326 lateen 0.941176 -12.0327 reviewers, 0.476923 -12.0329 grows 0.258929 -12.0331 this task 0.6875 -12.0332 plant 0.791667 -12.0332 nv 0.638889 -12.0332 0.53 0.24498 -12.0334 so, 0.85 -12.0343 (das 0.666667 -12.0344 0.24 0.7 -12.0344 will call 0.615385 -12.0346 integrated with 0.941176 -12.0346 ttm 0.76 -12.0346 adv 0.395833 -12.0348 sampler 0.791667 -12.0348 2010b; 0.941176 -12.0348 sending 0.941176 -12.0348 gordon 0.941176 -12.035 poesio 0.476923 -12.0352 0.04 0.941176 -12.0352 pzd 0.615385 -12.0353 entrainment 0.528302 -12.0356 works on 0.382353 -12.0356 reduction in 0.941176 -12.0356 spanned 0.818182 -12.0358 journal, 0.85 -12.036 62.1 0.85 -12.036 labeled. 0.458333 -12.0362 omit 0.941176 -12.0362 model), 0.657143 -12.0363 east 0.517857 -12.0366 entity type 0.216981 -12.0366 closely 0.595238 -12.0373 multiplicative 0.406593 -12.0376 model parameters 0.413793 -12.0378 gale 0.0405405 -12.0381 two 0.740741 -12.0382 slower than 0.941176 -12.0383 1987; 0.888889 -12.0384 cooccurrences 0.402174 -12.0385 language resources 0.818182 -12.0385 hyponyms 0.538462 -12.0385 defense advanced 0.7 -12.0386 letting 0.888889 -12.0386 divergences 0.888889 -12.0386 twiter 0.484375 -12.0386 we present a novel 0.888889 -12.0388 most, 0.85 -12.0388 saliency 0.888889 -12.039 defined 0.888889 -12.039 words” 0.615385 -12.039 stacking 0.85 -12.039 pseudo-documents 0.235294 -12.039 quality. 0.818182 -12.0391 hao 0.941176 -12.0391 finaly, 0.941176 -12.0391 lapata’s 0.888889 -12.0392 bllip 0.888889 -12.0392 charge 0.941176 -12.0393 (maximum 0.941176 -12.0393 sleep 0.888889 -12.0394 system; 0.241245 -12.0394 (i) 0.615385 -12.0395 decoding time 0.76 -12.0395 arise from 0.888889 -12.0396 group-lasso 0.941176 -12.0396 humanities 0.941176 -12.0396 adition 0.301205 -12.0398 will not 0.409091 -12.0398 1997; 0.818182 -12.04 58.5 0.615385 -12.04 in principle 0.888889 -12.0402 paradigms, 0.941176 -12.0402 verbs: 0.666667 -12.0403 73.1 0.85 -12.0403 when calculating 0.266667 -12.0404 to explore 0.382353 -12.0405 value, 0.85 -12.0407 76.2 0.349593 -12.0408 processing. 0.941176 -12.0408 2in 0.818182 -12.0409 thoroughly 0.888889 -12.041 mention detector 0.888889 -12.041 w2, 0.941176 -12.041 lower, 0.382353 -12.0414 ). 0.888889 -12.0416 book” 0.888889 -12.0417 stratified 0.581395 -12.0417 nonparametric 0.666667 -12.0418 to validate 0.888889 -12.0421 2b 0.888889 -12.0423 9 conclusion 0.0775723 -12.0424 standard 0.791667 -12.0424 category: 0.791667 -12.0424 passed to 0.888889 -12.0425 korean, 0.888889 -12.0427 organism 0.76 -12.0427 72.1 0.791667 -12.0428 mappings. 0.888889 -12.0429 giza++, 0.528302 -12.0429 duc 0.85 -12.0429 ettm 0.888889 -12.0431 translation probabilities. 0.888889 -12.0435 prunes 0.565217 -12.0435 (liang et 0.25431 -12.0436 topics. 0.666667 -12.0437 loss. 0.888889 -12.0441 § 0.5 -12.0441 details on 0.565217 -12.0442 den 0.402174 -12.0446 distributions. 0.818182 -12.0451 erg 0.740741 -12.0452 our model’s 0.85 -12.0452 familiarity 0.85 -12.0452 upspeak 0.259091 -12.0453 pair. 0.888889 -12.0453 aspell 0.888889 -12.0455 docs 1 -12.0456 parsed. 1 -12.0456 stories. 1 -12.0456 red. 1 -12.0456 styles. 1 -12.0456 bad. 1 -12.0456 classified. 1 -12.0456 exception. 1 -12.0456 months. 1 -12.0456 seen. 1 -12.0456 sparse. 1 -12.0456 irrelevant. 1 -12.0456 projects. 1 -12.0456 connections. 1 -12.0456 re-ranking. 1 -12.0456 occurred. 1 -12.0456 wsd. 1 -12.0456 singapore. 1 -12.0456 .83 1 -12.0456 examples). 1 -12.0456 engines. 1 -12.0456 performances. 1 -12.0456 100.00 1 -12.0456 personalized pagerank 1 -12.0456 nominals. 1 -12.0456 languages). 1 -12.0456 naïve bayes 1 -12.0456 antecedent, 1 -12.0456 marked. 1 -12.0456 appear. 1 -12.0456 letter. 1 -12.0456 pruned. 1 -12.0456 nullnull. 1 -12.0456 distances. 1 -12.0456 patients. 1 -12.0456 changed. 1 -12.0456 total). 1 -12.0456 poorly. 1 -12.0456 acl. 1 -12.0456 approx. 1 -12.0456 fast. 1 -12.0456 country. 1 -12.0456 (a). 1 -12.0456 four. 1 -12.0456 ney. 1 -12.0456 addressed. 1 -12.0456 mistake. 1 -12.0456 incrementally. 1 -12.0456 triggers. 1 -12.0456 understand. 1 -12.0456 individuals. 1 -12.0456 fit. 1 -12.0456 standards. 1 -12.0456 me. 1 -12.0456 in-context 1 -12.0456 language). 1 -12.0456 2007b). 1 -12.0456 mintz 1 -12.0456 ms. 1 -12.0456 brackets. 0.85 -12.0456 convex, 0.9375 -12.0457 retained. 0.9375 -12.0457 detector. 0.9375 -12.0457 context). 0.9375 -12.0457 ram. 0.9375 -12.0457 does. 0.9375 -12.0457 priority queue 0.818182 -12.0458 descriptors 0.435897 -12.0458 hope to 0.445946 -12.0459 evaluator 0.85 -12.0461 (mcdonald, 0.7 -12.0462 3.9 0.508772 -12.0463 connected by 0.76 -12.0464 portuguese, 0.888889 -12.0465 ii: 0.9375 -12.0466 scaling factors 0.888889 -12.0467 pos tagger, 0.261468 -12.0467 fig. 0.206304 -12.0472 moses 0.76 -12.0472 conj 0.9375 -12.0472 feature potency 0.476923 -12.0473 board 0.6 -12.0473 bd 0.9375 -12.0474 ful 0.202778 -12.0474 connected 0.9375 -12.0476 etc 0.888889 -12.0477 (there 0.0957643 -12.0477 are not 0.85 -12.0478 74.0 0.463768 -12.0479 love 0.888889 -12.0479 krey`ol 0.55102 -12.048 (2004). 0.264151 -12.048 parser, 0.184758 -12.0481 details 0.7 -12.0485 compositionality. 0.278947 -12.0485 covered 0.9375 -12.0485 projected onto 0.9375 -12.0487 conll 2008 0.714286 -12.0489 ice 0.888889 -12.0489 influential 0.9375 -12.0489 branching factor 0.5 -12.0489 permits 0.888889 -12.0491 240 0.9375 -12.0491 teacher’s 0.888889 -12.0495 it: 0.43038 -12.0496 smith, 0.85 -12.0496 participants who 0.714286 -12.0497 yprime 0.888889 -12.0497 performance: 0.888889 -12.0498 limit, 0.43038 -12.0499 outcome 0.9375 -12.0499 hints 0.9375 -12.0499 (sun 0.9375 -12.0501 rimell 0.226027 -12.0503 al., 2010), 0.888889 -12.0504 security 0.9375 -12.0506 naseem 0.818182 -12.0507 tree-to-string translation 0.55102 -12.0508 acknowledgments the authors 0.888889 -12.051 derivatives 0.9375 -12.051 modifications, 0.888889 -12.0512 7the 0.469697 -12.0513 bipartite 0.714286 -12.0513 across languages. 0.818182 -12.0514 terminates 0.581395 -12.0514 (an 0.621622 -12.0515 costly 0.76 -12.0515 joshi, 0.7 -12.0518 loglikelihood 0.595238 -12.0518 pronouns. 0.888889 -12.0522 meta-data 0.888889 -12.0522 about, 0.76 -12.0522 (u,v) 0.9375 -12.0522 91.4 0.356522 -12.0524 w, 0.818182 -12.0524 a 3-gram 0.9375 -12.0524 well-suited 0.9375 -12.0524 facial expresions 0.9375 -12.0526 (liu, 0.76 -12.0528 0.91 0.666667 -12.053 will give 0.17773 -12.053 corresponding to 0.9375 -12.0531 representatives 0.409091 -12.0533 this step 0.9375 -12.0533 0.2% 0.311258 -12.0535 indeed, 0.76 -12.0535 processor 0.9375 -12.0535 (see fig. 0.222973 -12.0538 similarly, 0.888889 -12.054 750 0.409091 -12.0541 (one 0.85 -12.0541 manli 0.85 -12.0541 ma, 0.483871 -12.0541 their method 0.9375 -12.0541 improvements, 0.9375 -12.0541 voorhees, 0.565217 -12.0541 syntactic category 0.9375 -12.0543 incur 0.740741 -12.0545 bansal 0.9375 -12.0545 favorably 0.9375 -12.0547 schler 0.647059 -12.0549 wilson 0.85 -12.0549 maxsim 0.9375 -12.055 koller, 0.85 -12.0551 twitter messages 0.76 -12.0551 evaluation methodology 0.9375 -12.0552 transitional 0.9375 -12.0554 dotted lines 0.9375 -12.0554 test) 0.9375 -12.0554 pre-specified 0.782609 -12.0555 relaxing 0.76 -12.0556 negative effect 0.818182 -12.0558 λ, 0.740741 -12.0558 heads. 0.9375 -12.0558 (section 6). 0.9375 -12.0558 asr, 0.782609 -12.0559 approached 0.666667 -12.056 4the 0.888889 -12.056 2.2.2 0.6 -12.0561 tr 0.37037 -12.0562 w/o 0.581395 -12.0563 each entry 0.508772 -12.0564 entailment, 0.740741 -12.0565 crucially, 0.9375 -12.0566 overcoming 0.818182 -12.0569 google, 0.7 -12.057 were chosen 0.9375 -12.0571 noord, 0.9375 -12.0571 fleeting 0.9375 -12.0571 best-link 0.666667 -12.0574 gw 0.363636 -12.0574 better performance 0.818182 -12.0575 terms: 0.529412 -12.0575 document classification 0.9375 -12.0575 renaming acts 0.85 -12.0576 following sections, 0.740741 -12.0577 goal. 0.9375 -12.0577 switches 0.9375 -12.0577 innovations 0.888889 -12.0578 large datasets 0.888889 -12.058 ł 0.469697 -12.0581 glue 0.9375 -12.0581 (bengio 0.782609 -12.0583 stages: 0.9375 -12.0583 pursuing 0.638889 -12.0585 10) 0.9375 -12.0585 figure 1 presents 0.818182 -12.0586 scenario where 0.888889 -12.0586 scott 0.9375 -12.0588 clustering-based 0.9375 -12.0588 halteren 0.9375 -12.0588 decision, 0.9375 -12.0588 graduate school 0.85 -12.0589 lewis 0.9375 -12.059 multiple, 0.621622 -12.0593 (johansson and 0.615385 -12.0594 ki 0.9375 -12.0594 carpuat 0.85 -12.0598 black, 0.85 -12.0598 rebuttals 0.161634 -12.0598 that, 0.677419 -12.06 mismatches 0.9375 -12.06 3.5% 0.714286 -12.0602 row shows 0.888889 -12.0602 designate 0.782609 -12.0603 pas 0.76 -12.0604 ptop 0.9375 -12.0604 reserve 0.9375 -12.0604 gold, 0.888889 -12.0606 triangulated 0.782609 -12.0606 moreover 0.621622 -12.0607 rouge-2 0.9375 -12.0607 interconnected 0.9375 -12.0607 quantificational 0.76 -12.0609 pronunciations 0.9375 -12.0609 geography 0.647059 -12.061 vast majority 0.9375 -12.0611 36% 0.9375 -12.0613 61.3 0.782609 -12.0615 expansion. 0.9375 -12.0615 schwenk 0.85 -12.0616 abc 0.43038 -12.0618 inclusion of 0.9375 -12.0621 stages, 0.9375 -12.0621 respectively), 0.509091 -12.0622 displayed in 0.581395 -12.0625 communication. 0.9375 -12.0626 banerjee 0.9375 -12.0628 redundancy, 0.9375 -12.0628 poesio, 0.9375 -12.0628 favorable 0.9375 -12.0628 (raux 0.9375 -12.063 26.5 0.85 -12.0631 bunt 0.9375 -12.0632 57.0 0.9375 -12.0632 semisupervised learning 0.9375 -12.0632 (un)certainty 0.818182 -12.0636 computationallinguistics 0.568182 -12.0637 nguyen 0.581395 -12.0637 accidents 0.568182 -12.064 mohammad 0.9375 -12.064 mk 0.366972 -12.0642 multi-document 0.9375 -12.0643 (ganchev 0.366972 -12.0646 t = 0.888889 -12.0646 157 0.85 -12.0647 cause, 0.9375 -12.0649 parseval 0.9375 -12.0649 subtracted 0.7 -12.065 trains 0.7 -12.0652 word co-occurrence 0.9375 -12.0653 delivers 0.9375 -12.0653 evolving 0.9375 -12.0655 sgn 0.9375 -12.0655 .48 0.9375 -12.0657 test items. 0.9375 -12.0657 duplication 0.888889 -12.0658 nonparametric bayesian 0.888889 -12.066 “an 0.111018 -12.0662 considered 0.9375 -12.0664 slav 0.9375 -12.0664 iaa 0.9375 -12.0666 singular, 0.85 -12.067 interruption 0.85 -12.067 absolutely 0.9375 -12.067 bagof-words 0.9375 -12.067 81.3 0.295181 -12.0671 union 0.9375 -12.0672 pca 0.85 -12.0673 30,000 0.85 -12.0673 experiencer 0.9375 -12.0674 56.7 0.9375 -12.0674 pos tag, 0.888889 -12.0679 operator, 0.888889 -12.0679 (usually 0.9375 -12.0679 proposition, 0.9375 -12.0679 wording 0.9375 -12.0681 collocation extraction 0.85 -12.0683 manually evaluated 0.9375 -12.0683 .19 0.9375 -12.0683 distinguishes between 0.9375 -12.0685 262 0.9375 -12.0685 feeds 0.9375 -12.0685 agenda-based 0.509091 -12.0687 we removed 0.9375 -12.0687 42.1 0.647059 -12.0689 0.89 0.196809 -12.0689 2.2 0.258065 -12.0689 implement 0.9375 -12.0691 parse forest 0.888889 -12.0693 t=1 0.9375 -12.0694 untranslated 0.9375 -12.0694 quiz 0.85 -12.0695 scales, 0.76 -12.0695 15, 0.647059 -12.0697 distractor 0.9375 -12.0698 taggers, 0.9375 -12.07 accommodating 0.9375 -12.07 density, 0.347107 -12.0701 secondly, 0.9375 -12.0702 17.2 0.5 -12.0703 allowing for 0.5 -12.0703 word tokens 0.9375 -12.0704 training) 0.9375 -12.0704 cooccur 0.36036 -12.0705 to integrate 0.9375 -12.0706 66.8 0.9375 -12.0706 syntax-only 0.888889 -12.0707 employees 0.325926 -12.0707 correlate 0.85 -12.0708 harmony 0.9375 -12.0709 remedy 0.9375 -12.0711 read–it 0.529412 -12.0713 attach 0.9375 -12.0713 (gim´enez 0.9375 -12.0713 91.6 0.9375 -12.0713 out-ofvocabulary 0.888889 -12.0715 memory usage 0.9375 -12.0715 54.6 0.647059 -12.0716 svm. 0.0660096 -12.0716 any 0.647059 -12.0717 automatic extraction 0.9375 -12.0717 incompatible 0.9375 -12.0717 dundee 0.888889 -12.0719 post-test 0.9375 -12.0719 template-based 0.335938 -12.072 intended to 0.9375 -12.0721 constituent, 0.85 -12.0722 r∈r 0.782609 -12.0725 matters 0.123431 -12.0725 apply 0.9375 -12.0726 one-to-many 0.9375 -12.0726 54.8 0.9375 -12.0728 curation 0.9375 -12.0728 keyphrases, 0.9375 -12.073 violates 0.553191 -12.0731 action, 0.9375 -12.0732 2500 0.9375 -12.0734 veterinary 0.9375 -12.0734 entitled 0.9375 -12.0734 3.2) 0.9375 -12.0734 cm sketch 0.9375 -12.0734 pageviews 0.181406 -12.0736 suggest 0.9375 -12.0736 duplicated 0.9375 -12.0739 chiang’s 0.9375 -12.0739 18.4 0.9375 -12.0741 lewis, 0.9375 -12.0741 selection: 0.76 -12.0742 spech 0.677419 -12.0743 merit 0.85 -12.0743 notations 0.9375 -12.0743 .63 0.647059 -12.0744 aggregating 0.450704 -12.0745 callison-burch, 0.714286 -12.0746 tu 0.888889 -12.0746 in sum, 0.9375 -12.0747 45m 0.9375 -12.0747 modifier ordering 0.714286 -12.0749 input string 0.76 -12.0749 rationality 0.9375 -12.0749 rare, 0.9375 -12.0749 86.8 0.782609 -12.075 macro 0.9375 -12.0751 reflection 0.483871 -12.0752 ‘the 0.888889 -12.0754 turning to 0.782609 -12.0754 filling 0.9375 -12.0754 spotting 0.9375 -12.0754 (1997), 0.9375 -12.0756 certain+correct 0.518519 -12.0758 1993; 0.888889 -12.0758 training material 0.730769 -12.0759 ↔ 0.621622 -12.076 which indicates that 0.9375 -12.076 entities; 0.9375 -12.076 younger 0.9375 -12.0762 table 3 presents 0.450704 -12.0763 y, 0.529412 -12.0764 must have 0.809524 -12.0764 f-measure, 0.9375 -12.0764 quantized 0.85 -12.0765 .31 0.888889 -12.0766 case: 0.9375 -12.0766 jordan, 0.444444 -12.0769 electronic 0.9375 -12.0769 bionlp’11 shared 0.9375 -12.0769 anotations 0.9375 -12.0769 decreases, 0.553191 -12.077 by choosing 0.275132 -12.0772 0.2 0.9375 -12.0773 additions 0.85 -12.0774 reality 0.428571 -12.0775 integer linear 0.5 -12.0775 [ ] 0.85 -12.0776 atb 0.888889 -12.0778 better suited 0.229091 -12.0779 result, 0.888889 -12.078 proceeding 0.605263 -12.0781 iso 0.809524 -12.0781  0.782609 -12.0781 illinois 0.9375 -12.0781 4 0 . 0.9375 -12.0781 regularly 0.85 -12.0784 pprime 0.9375 -12.0784 attention, 0.509091 -12.0785 relation type 0.76 -12.0785 (surdeanu et al., 0.9375 -12.0786 .15 0.9375 -12.0786 ppis 0.647059 -12.0788 smt system. 0.9375 -12.079 k-th 0.6 -12.0792 shortcomings 0.621622 -12.0793 domingos, 0.9375 -12.0794 (yang 0.9375 -12.0794 surface, 0.730769 -12.0796 va 0.585366 -12.0796 ⊆ 0.9375 -12.0797 asr-based 0.9375 -12.0797 recalibration 0.9375 -12.0797 quantifies 0.9375 -12.0797 fox 0.9375 -12.0797 (tromble 0.9375 -12.0799 chair 0.9375 -12.0799 wordform 0.9375 -12.0799 hint 0.85 -12.08 phoenix 0.621622 -12.0801 these issues 0.9375 -12.0801 wizard’s 0.9375 -12.0801 88.6 0.888889 -12.0803 primary goal 0.9375 -12.0803 navigate 0.278075 -12.0804 expressed in 0.0948814 -12.0805 label 0.9375 -12.0807 mimic 0.9375 -12.0807 mei 0.9375 -12.0807 proposal, 0.4 -12.0809 word pair 0.9375 -12.0809 trees) 0.9375 -12.0809 69.3 0.85 -12.081 simple. 0.9375 -12.0812 stemming, 0.888889 -12.0813 bulk of 0.9375 -12.0814 more elaborate 0.9375 -12.0816 revise 0.0762712 -12.0817 analysis 0.76 -12.0817 performance gains 0.9375 -12.0818 weasel 0.9375 -12.0818 stag 0.9375 -12.082 crawling 0.176856 -12.0821 we evaluate 0.455882 -12.0821 impose 0.9375 -12.0825 constituted 0.9375 -12.0827 28.0 0.9375 -12.0827 exercise, 0.9375 -12.0827 62.2 0.305195 -12.0829 a special 0.76 -12.0829 p(c) 0.9375 -12.0829 non-deterministic 0.9375 -12.0831 traces 0.9375 -12.0831 (ne) 0.9375 -12.0833 intersection, 0.9375 -12.0833 superficial 0.9375 -12.0835 365 0.9375 -12.0835 dtk 0.9375 -12.0835 word-lattice-based 0.76 -12.0836 restrictions. 0.65625 -12.0839 an obvious 0.621622 -12.0839 noticeable 0.730769 -12.0842 validation set 0.9375 -12.0842 collaborations 0.9375 -12.0844 2 0 . 0.689655 -12.0845 blitzer 0.605263 -12.0845 levels, 0.9375 -12.0846 facilities 0.9375 -12.0846 71.1 0.428571 -12.0848 specifications 0.9375 -12.0848 experiment: 0.9375 -12.0848 1985; 0.9375 -12.0848 unlimited 0.628571 -12.0849 ganchev 0.809524 -12.085 genitive 0.555556 -12.0853 minus 0.9375 -12.0853 criticized 0.555556 -12.0854 a huge 0.9375 -12.0855 backspace 0.9375 -12.0857 uncertainty about 0.689655 -12.0858 when combining 0.605263 -12.0859 bigrams. 0.9375 -12.0859 .45 0.689655 -12.0861 past work 0.647059 -12.0861 0.10 0.364486 -12.0862 proposed approach 0.628571 -12.0862 nonlinear 0.585366 -12.0862 connects 0.260664 -12.0863 we extract 0.9375 -12.0864 330 0.9375 -12.0866 connected, 0.9375 -12.0866 silence 0.689655 -12.0868 main contribution of 0.809524 -12.0869 mo 0.809524 -12.0869 (turney, 0.434211 -12.0869 hovy 0.888889 -12.087 confidence measure 0.9375 -12.087 lexical-semantic 0.9375 -12.087 abilities 0.372549 -12.0871 understood 0.9375 -12.0872 companies, 0.9375 -12.0872 gerber 0.9375 -12.0879 printed 0.9375 -12.0879 requests, 0.9375 -12.0881 mandatory 0.5 -12.0881 different methods 0.47619 -12.0883 regression model 0.9375 -12.0883 (defined 0.395604 -12.0883 table 7 0.301282 -12.0884 that contain 0.541667 -12.0885 there is some 0.9375 -12.0885 176 0.555556 -12.0886 modified kneser-ney 0.0723949 -12.0886 context 0.9375 -12.0894 ratios, 0.9375 -12.0894 on-the-fly 0.888889 -12.0895 abductive 0.529412 -12.0896 in conjunction 0.509091 -12.0896 that combines 0.541667 -12.0898 christopher 0.9375 -12.0898 cache, 0.9375 -12.0898 robot’s 0.714286 -12.0899 per-word 0.888889 -12.0899 portable 0.9375 -12.0901 259 0.9375 -12.0901 communicating 0.809524 -12.0902 sb 0.9375 -12.0903 cellular 0.65625 -12.0905 for computational linguistics. 0.9375 -12.0905 assertion 0.689655 -12.0906 consonants 0.782609 -12.0909 pron 0.9375 -12.0909 non-experts 0.414634 -12.091 corpus contains 0.9375 -12.0911 ge, 0.9375 -12.0911 mediocre 0.541667 -12.0912 grained 0.647059 -12.0912 further analysis 0.341667 -12.0912 builds 0.042848 -12.0914 results 0.264706 -12.0915 it also 0.9375 -12.0916 maximum-entropy 0.9375 -12.0916 diference 0.621622 -12.0918 micro-averaged 0.9375 -12.0918 bcub 0.714286 -12.0921 again. 0.9375 -12.0922 (n) 0.176087 -12.0924 close 0.809524 -12.0925 efl 0.568182 -12.0925 uttered 0.377551 -12.0926 abstract we present 0.621622 -12.093 simultaneously. 0.65625 -12.0931 their frequency 0.357798 -12.0933 we developed 0.555556 -12.0933 l0 0.9375 -12.0935 nsubj 0.714286 -12.0937 interpolated with 0.65625 -12.0938 an example. 0.9375 -12.0938 (johnson, 0.9375 -12.0938 presently 0.9375 -12.094 levin’s 0.809524 -12.0942 double-dop 0.134347 -12.0942 is also 0.9375 -12.0942 creole-english 0.9375 -12.0944 n0 0.541667 -12.0946 weather 0.9375 -12.0946 θi 0.677419 -12.0948 0-1 0.9375 -12.0949 dit++ 0.888889 -12.0951 proper name 0.571429 -12.0953 the first author 0.9375 -12.0953 (mitchell 0.541667 -12.0954 parse, 0.782609 -12.0955 five times 0.842105 -12.0957 album 0.9375 -12.096 regularization, 0.677419 -12.0962 challenge, 0.782609 -12.0962 2001 0.172269 -12.0965 likelihood 0.428571 -12.0966 simplicity, 0.782609 -12.0966 preserved. 0.0526521 -12.0967 first 0.483333 -12.0967 jump 0.842105 -12.0968 (dunning, 0.9375 -12.0968 210 0.9375 -12.0968 experimental design 0.9375 -12.0968 0.01) 0.211356 -12.0969 stage 0.9375 -12.0971 definitely 0.0852041 -12.0977 structure 0.9375 -12.0977 exemplar-based 0.842105 -12.0978 noun-phrase 0.730769 -12.0978 to reconstruct 0.9375 -12.0979 without requiring 0.9375 -12.0979 section 2 discusses 0.219931 -12.0979 looking 0.689655 -12.098 traum, 0.677419 -12.098 introduce a new 0.75 -12.0981 contentious 0.555556 -12.0981 experimental results on 0.568182 -12.0981 an algorithm for 0.9375 -12.0981 patron 0.782609 -12.0982 livejournal 0.888889 -12.0984 unknown words. 0.9375 -12.0984 english), 0.730769 -12.0985 post-hoc 0.9375 -12.0986 seeding 0.75 -12.0988 ncbi 0.888889 -12.0988 .70 0.139696 -12.0988 produced 0.782609 -12.0988 220 0.782609 -12.0988 stating 0.138587 -12.0989 identified 0.809524 -12.0989 89.2 0.782609 -12.099 main focus 0.9375 -12.099 would, 0.193634 -12.0991 know 0.782609 -12.0991 ends with 0.9375 -12.0992 hierarchical dirichlet 0.9375 -12.0992 fdn 0.9375 -12.0992 intervention 0.9375 -12.0995 terminological 0.9375 -12.0995 copula 0.842105 -12.0996 coders 0.414634 -12.0996 prep 0.9375 -12.0997 n-foil 0.809524 -12.0999 accomplished by 0.9375 -12.1003 florida 0.9375 -12.1006 (most 0.689655 -12.1007 inference algorithm 0.9375 -12.1008 garbage 0.75 -12.1009 alex 0.842105 -12.101 77.2 0.809524 -12.101 hiersum 0.9375 -12.101 minister 0.428571 -12.1013 not explicitly 0.809524 -12.1014 unidirectional 0.730769 -12.1015 factual 0.75 -12.1015 x,y 0.782609 -12.1017 beautiful 0.9375 -12.1017 posting 0.809524 -12.1018 fubl 0.9375 -12.1019 dyadic 0.9375 -12.1019 nonterminals, 0.9375 -12.1019 yj 0.9375 -12.1021 weblog 0.9375 -12.1023 68.9 0.9375 -12.1023 beginnings 0.9375 -12.1023 ckk 0.9375 -12.1025 evolved 0.9375 -12.103 280 0.9375 -12.1032 value) 0.409639 -12.1034 better. 0.9375 -12.1037 hierarchically 0.409639 -12.1038 to address this 0.782609 -12.1039 filtering. 0.9375 -12.1039 cnn 0.730769 -12.1041 tiger 0.75 -12.1041 examine how 0.628571 -12.1041 beam-width 0.21843 -12.1042 2006), 0.0756167 -12.1042 computational 0.9375 -12.1043 western 0.9375 -12.1043 estimations 0.519231 -12.1045 objects. 0.782609 -12.1045 raghavan 0.809524 -12.1048 for download 0.9375 -12.105 mcfg 0.483333 -12.105 each category 0.404762 -12.1054 for each sentence 0.9375 -12.1054 blog, 0.9375 -12.1054 vein, 0.809524 -12.1056 une 0.9375 -12.1056 (g) 0.9375 -12.1056 tone 0.5 -12.1058 hedges 0.9375 -12.1059 resources: 0.782609 -12.1061 criterion, 0.9375 -12.1061 mt03 0.555556 -12.1063 0.08 0.302632 -12.1064 well, 0.9375 -12.1065 researches 0.9375 -12.1065 vj 0.9375 -12.1065 keystrokes 0.75 -12.1068 (charniak, 0.9375 -12.107 likelihood, 0.882353 -12.1071 estimates, 0.730769 -12.1072 processed, 0.9375 -12.1072 string-rewriting 0.882353 -12.1073 assumption: 0.882353 -12.1073 ten-fold 0.730769 -12.1074 existing systems 0.842105 -12.1075 verb-noun 0.842105 -12.1075 λ-hybrid 0.402299 -12.1075 union of 0.882353 -12.1075 53% 0.882353 -12.1075 annealing 0.882353 -12.1075 unsup 0.882353 -12.1075 what’s 0.882353 -12.1077 statements, 0.882353 -12.1077 obama, 0.782609 -12.1078 ding 0.882353 -12.1081 ground-truth 0.65625 -12.1082 the latter, 0.9375 -12.1083 layer, 0.338843 -12.1086 emails 0.461538 -12.1087 sequence, 0.432432 -12.1087 (eq. 0.628571 -12.109 editor 0.842105 -12.1091 72.2 0.605263 -12.1091 baroni and 0.9375 -12.1092   • 0.454545 -12.1094 maximum number 0.543478 -12.1094 decoding algorithm 0.605263 -12.1094 to simulate 0.882353 -12.1096 convenience 0.9375 -12.1096 multi-lingual 0.9375 -12.1099 (6), 0.882353 -12.11 dairy 0.543478 -12.11 biggest 0.809524 -12.11 pp-attachment 0.628571 -12.1102 data structures 0.882353 -12.1104 (ji 0.882353 -12.1104 chief 0.365385 -12.1105 we cannot 0.65625 -12.1105 we can achieve 0.119145 -12.1107 one of the 0.882353 -12.1108 doubling 0.809524 -12.1108 guesses 0.585366 -12.1109 (3), 0.882353 -12.1113 marginalizing 0.882353 -12.1113 0; 0.730769 -12.1115 corrects 0.882353 -12.1115 (mostly 0.882353 -12.1115 window around 0.387097 -12.1115 conditions. 0.809524 -12.1115 free text 0.75 -12.1116 finance 0.842105 -12.1117 74.4 0.882353 -12.1117 (indicated 0.65625 -12.1117 explosion 0.842105 -12.1121 text) 0.882353 -12.1121 anual 0.882353 -12.1125 omission 0.555556 -12.1125 a finite set 0.809524 -12.1127 many-to-many 0.571429 -12.1128 area. 0.882353 -12.1131 204 0.703704 -12.1133 dialogue-style 0.809524 -12.1136 30.2 0.589744 -12.114 experiments: 0.882353 -12.1142 syn 0.75 -12.1142 dialogic 0.809524 -12.1142 cell, 0.842105 -12.1143 hash table 0.882353 -12.1144 intra-sentential 0.882353 -12.1144 quarter 0.397727 -12.1145 (d) 0.882353 -12.1146 society, 1 -12.1146 instructions. 1 -12.1146 constituent. 1 -12.1146 practical. 1 -12.1146 a). 1 -12.1146 (ed.) 1 -12.1146 rules). 1 -12.1146 completion. 1 -12.1146 analyzed. 1 -12.1146 dualist 1 -12.1146 drastic 1 -12.1146 less. 1 -12.1146 correctness. 1 -12.1146 inventories. 1 -12.1146 cell. 1 -12.1146 management. 1 -12.1146 dmv. 1 -12.1146 microblogs. 1 -12.1146 curve. 1 -12.1146 unlearned 1 -12.1146 nonpar. 1 -12.1146 last year’s 1 -12.1146 haiti. 1 -12.1146 simulation. 1 -12.1146 smaller. 1 -12.1146 abstraction. 1 -12.1146 50.00 1 -12.1146 expertise. 1 -12.1146 job. 1 -12.1146   1 -12.1146 successful. 1 -12.1146 solved. 1 -12.1146 interests. 1 -12.1146 (c). 1 -12.1146 intersection. 1 -12.1146 exercise. 1 -12.1146 contains. 1 -12.1146 si. 1 -12.1146 cycles. 1 -12.1146 completed. 1 -12.1146 .. 1 -12.1146 chat. 1 -12.1146 philips, 1 -12.1146 phenotypes. 1 -12.1146 publication. 1 -12.1146 hard. 1 -12.1146 preference. 1 -12.1146 know. 1 -12.1146 decipherment. 1 -12.1146 maximum. 1 -12.1146 (209). 1 -12.1146 concluding remarks 1 -12.1146 schema. 1 -12.1146 2008b). 1 -12.1146 address. 1 -12.1146 turkish. 1 -12.1146 poor. 1 -12.1146 exercises. 1 -12.1146 intuition. 1 -12.1146 observation. 1 -12.1146 platform. 1 -12.1146 percentage. 1 -12.1146 treelets. 1 -12.1146 specific. 1 -12.1146 forests. 1 -12.1146 medicine. 1 -12.1146 answering. 1 -12.1146 short. 1 -12.1146 0.001). 1 -12.1146 follow. 1 -12.1146 c2. 1 -12.1146 beliefs. 0.882353 -12.1148 fst 0.316547 -12.1148 1998). 0.933333 -12.1148 realization. 0.933333 -12.1148 1987). 0.933333 -12.1148 affinity propagation 0.933333 -12.1148 spelling correction. 0.933333 -12.1148 conceptresolver’s 0.933333 -12.1148 coordination. 0.933333 -12.1148 yates, 0.611111 -12.115 media, 0.571429 -12.115 a big 0.842105 -12.1151 vithea 0.387097 -12.1151 directly from 0.842105 -12.1153 67.1 0.75 -12.1153 gazetteers 0.330709 -12.1155 points. 0.882353 -12.1157 2003, 0.703704 -12.1158 police 0.882353 -12.1159 weakness of 0.882353 -12.1159 residual 0.933333 -12.1159 lcfrss 0.933333 -12.1159 part-ofspeech tags 0.842105 -12.1161 52.5 0.933333 -12.1161 strong, 0.933333 -12.1161 genia tagger 0.605263 -12.1162 in spite 0.628571 -12.1162 correlation. 0.605263 -12.1165 (15) 0.933333 -12.1166 jaist 0.882353 -12.1167 analogously, 0.933333 -12.1168 acl hlt 0.933333 -12.1168 coding scheme 0.611111 -12.117 alexander 0.842105 -12.1171 topically 0.882353 -12.1171 tree decomposition 0.882353 -12.1171 explicit connectives 0.4125 -12.1173 1997). 0.225926 -12.1174 vp 0.809524 -12.1175 shop 0.842105 -12.1177 randomization 0.689655 -12.1177 instantiation of 0.933333 -12.1177 wherever 0.933333 -12.1177 addressees 0.882353 -12.118 favourability 0.882353 -12.118 67% 0.571429 -12.118 moschitti, 0.809524 -12.1181 wonder 0.730769 -12.1185 b@ 0.882353 -12.1186 paper: 0.65625 -12.1186 riloff 0.373737 -12.1187 productdisplay 0.263959 -12.1188 infer 0.123094 -12.1189 first, 0.842105 -12.1191 israeli 0.882353 -12.1192 acceptability 0.589744 -12.1192 combined. 0.933333 -12.1195 partition, 0.842105 -12.1197 edit operations 0.842105 -12.1199 73.7 0.933333 -12.1199 semantic drift 0.432432 -12.1199 the usual 0.842105 -12.1201 stymne 0.882353 -12.1201 pn 0.882353 -12.1203 cohen et al. 0.404762 -12.1207 thanks to 0.842105 -12.1208 andrea 0.809524 -12.1208 baseline) 0.882353 -12.1209 suffices to 0.454545 -12.121 business 0.0454073 -12.121 training 0.75 -12.1212 triage 0.809524 -12.1212 excessive 0.933333 -12.1213 circumvent 0.933333 -12.1213 37.1 0.882353 -12.1214 verb) 0.882353 -12.1214 algorithms: 0.628571 -12.1214 nullnullnullnullnullnull 0.809524 -12.1214 returning to 0.842105 -12.1216 literally 0.842105 -12.1216 (johnson 0.75 -12.1216 retrained 0.809524 -12.1216 unanswered 0.772727 -12.1216 sec 0.509434 -12.122 consequence, 0.933333 -12.1222 bayesian networks 0.933333 -12.1222 scores) 0.809524 -12.1222 anova 0.244541 -12.1223 al., 2009), 0.417722 -12.1225 also include 0.636364 -12.1226 compression. 0.19774 -12.1226 we need 0.933333 -12.1228 gurevych, 0.274725 -12.1229 submitted 0.809524 -12.1229 first letter 0.882353 -12.123 increased, 0.933333 -12.1231 74.2 0.933333 -12.1233 rejects 0.933333 -12.1233 disparity 0.882353 -12.1235 writing, 0.611111 -12.1235 0.69 0.55814 -12.1235 these resources 0.933333 -12.1237 dagan 0.933333 -12.1237 pmi, 0.530612 -12.1239 shares 0.666667 -12.124 grammaticality, 0.182927 -12.124 ≤ 0.933333 -12.124 .) 0.772727 -12.124 .87 0.882353 -12.1241 events: 0.933333 -12.1242 latent variables, 0.52 -12.1242 this leads to 0.882353 -12.1243 245 0.933333 -12.1244 (robaldo, 0.933333 -12.1244 (see, 0.772727 -12.1244 full range 0.636364 -12.1246 another important 0.933333 -12.1246 29.3 0.703704 -12.1246 0.005 0.809524 -12.1247 hypothesis: 0.809524 -12.1249 stationary 0.102273 -12.125 although 0.882353 -12.1254 (2a) 0.543478 -12.1254 fr-en 0.933333 -12.1255 downsampling 0.933333 -12.1255 stories, 0.842105 -12.1258 (very 0.933333 -12.1258 freebase, 0.75 -12.1259 is-a 0.842105 -12.126 irstlm 0.933333 -12.126 64.6 0.689655 -12.1262 gold mentions 0.933333 -12.1262 accommodation 0.882353 -12.1264 endpoints 0.4 -12.1264 syntactic features 0.933333 -12.1264 user simulations 0.636364 -12.1265 ud 0.933333 -12.1266 (riezler 0.933333 -12.1266 59.9 0.344828 -12.127 is trained on 0.666667 -12.1271 event detection 0.933333 -12.1273 word-sense 0.141813 -12.1273 included 0.842105 -12.1274 feature functions. 0.882353 -12.1275 (lexical 0.933333 -12.1275 zaidan 0.933333 -12.1278 33.0 0.933333 -12.1278 qiu 0.482759 -12.1278 validity of 0.689655 -12.128 organize 0.933333 -12.128 inits 0.703704 -12.128 statistical approaches 0.5 -12.1281 slower 0.842105 -12.1284 priors, 0.933333 -12.1285 fine-tuned 0.933333 -12.1285 mai 0.842105 -12.1287 3-5 0.589744 -12.1287 arguably 0.933333 -12.1287 fuzzification 0.933333 -12.1287 ‖ (pronounced 0.882353 -12.1288 sentence length, 0.395349 -12.1289 network. 0.933333 -12.1289 0.1% 0.882353 -12.129 (what 0.666667 -12.129 moment 0.703704 -12.1292 more accurately 0.666667 -12.1293 durme 0.933333 -12.1294 stir 0.933333 -12.1294 55.8 0.689655 -12.1295 limited. 0.636364 -12.1295 (2009) proposed 0.43662 -12.1295 tried to 0.571429 -12.1296 limited number 0.772727 -12.1296 evoked by 0.75 -12.1297 initial experiments 0.842105 -12.1299 ___ 0.933333 -12.13 especially, 0.842105 -12.1303 austronesian 0.666667 -12.1303 w3 0.933333 -12.1303 78.0 0.882353 -12.1305 207 0.933333 -12.1307 2008); 0.933333 -12.1307 frame identification 0.75 -12.1308 s) 0.37 -12.1309 array 0.933333 -12.1309 vectorial 0.933333 -12.1309 characterizes 0.809524 -12.1309 unigrams. 0.312057 -12.1312 then used 0.933333 -12.1312 thematic role 0.933333 -12.1312 weight vector w 0.137174 -12.1313 < 0.772727 -12.1315 discharge 0.882353 -12.1316 ologreg 0.933333 -12.1316 krahmer 0.933333 -12.1316 2008b) 0.933333 -12.1316 82.0 0.842105 -12.1317 fluency. 0.882353 -12.1318 (sometimes 0.933333 -12.1318 resource-poor 0.933333 -12.1318 scores; 0.933333 -12.1318 predicate’s 0.43662 -12.1319 much better 0.933333 -12.1321 major challenge 0.842105 -12.1325 pervasive 0.933333 -12.1325 26.2 0.55814 -12.1326 kyoto 0.183623 -12.1327 look 0.933333 -12.1327 translation) 0.933333 -12.1327 “and” 0.666667 -12.1329 disjunctive 0.933333 -12.133 post-editing 0.933333 -12.133 pinyin input 0.933333 -12.133 46.7 0.933333 -12.1332 73.4 0.430556 -12.1333 familiar 0.933333 -12.1334 72.7 0.933333 -12.1334 agree, 0.933333 -12.1334 routine 0.933333 -12.1336 wordlevel 0.933333 -12.1336 high-confidence 0.933333 -12.1336 90.1 0.611111 -12.1339 wt 0.933333 -12.1339 homonymy 0.933333 -12.1339 tac-kbp 0.0496492 -12.1341 information 0.933333 -12.1341 fourteen 0.933333 -12.1341 discourage 0.454545 -12.1342 development. 0.358491 -12.1343 found by 0.378947 -12.1343 ensures 0.933333 -12.1343 up-to-date 0.933333 -12.1343 attend 0.842105 -12.135 complement, 0.882353 -12.135 (> 0.933333 -12.135 27.3 0.175281 -12.1351 rules. 0.933333 -12.1352 unsupervised manner. 0.933333 -12.1352 alm 0.933333 -12.1352 heterogenous 0.375 -12.1352 word w 0.933333 -12.1355 combat 0.933333 -12.1355 6.9 0.933333 -12.1355 walker, 0.842105 -12.1356 hdt 0.933333 -12.1357 frenchenglish 0.772727 -12.1358 function tags 0.933333 -12.1359 non-transliteration 0.933333 -12.1359 informative, 0.933333 -12.1361 formant 0.636364 -12.1362 text summarization 0.842105 -12.1364 lattice parsing 0.842105 -12.1368 .24 0.933333 -12.1368 copied 0.933333 -12.1368 46.9 0.933333 -12.1368 2009b), 0.842105 -12.137 unsmoothed 0.933333 -12.137 intuitions about 0.933333 -12.137 misunderstandings 0.703704 -12.1372 70.1 0.933333 -12.1375 uiuc 0.933333 -12.1375 msz 0.933333 -12.1377 domain) 0.154116 -12.1379 taken 0.75 -12.1379 6), 0.933333 -12.138 pbmt 0.933333 -12.138 contrasted 0.72 -12.138 450 0.666667 -12.1381 empirical evaluation 0.882353 -12.1382 product review 0.882353 -12.1382 more effective than 0.72 -12.1384 lrec 0.933333 -12.1386 simultaneously, 0.933333 -12.1386 mer 0.467742 -12.1388 instantiations 0.933333 -12.1389 reformulation 0.594595 -12.139 training dataset 0.543478 -12.1391 an image 0.933333 -12.1391 circle 0.589744 -12.1392 possible values 0.933333 -12.1393 supertags 0.842105 -12.1395 multi 0.342105 -12.1395 category, 0.933333 -12.1396 arbitrary, 0.323077 -12.1398 passive 0.55814 -12.1398 observations, 0.842105 -12.1399 surpasses 0.842105 -12.1399 ns responses 0.933333 -12.14 pulman 0.882353 -12.1401 fp 0.460317 -12.1404 as noted 0.169492 -12.1405 we compare 0.882353 -12.1406 highly effective 0.0618325 -12.1406 target 0.589744 -12.1407 faculty 0.933333 -12.1407 region. 0.933333 -12.1407 ccg, 0.933333 -12.1407 (bottom 0.338983 -12.1407 was performed 0.933333 -12.1409 exemplify 0.882353 -12.141 confusion between 0.842105 -12.1411 instructions, 0.933333 -12.1411 delexicalized 0.933333 -12.1411 (resnik, 0.842105 -12.1413 nth 0.882353 -12.1414 van de 0.933333 -12.1414 56.0 0.933333 -12.1414 80.7 0.842105 -12.1415 appositive 0.933333 -12.1416 mb 0.933333 -12.1416 energy, 0.35514 -12.1417 which allows 0.842105 -12.1418 low-quality 0.933333 -12.1418 interior nodes 0.933333 -12.1418 reality, 0.882353 -12.1421 at hand 0.933333 -12.1421 online, 0.933333 -12.1421 dialogs. 0.933333 -12.1421 c4.5 0.933333 -12.1421 77% 0.933333 -12.1423 sentence), 0.933333 -12.1425 8.9 0.933333 -12.1425 emphasizing 0.842105 -12.1426 geotagged 0.380435 -12.1426 a substantial 0.933333 -12.1427 hms 0.933333 -12.1427 all-event 0.933333 -12.1427 ecological 0.933333 -12.1427 he/she 0.509804 -12.1427 imported 0.272222 -12.1428 to allow 0.882353 -12.1429 sequential information 0.933333 -12.143 ii, 0.933333 -12.143 focus, 0.933333 -12.143 face-to-face 0.509804 -12.143 exercise 0.5 -12.143 besides, 0.543478 -12.1431 opportunity to 0.138418 -12.1431 (the 0.842105 -12.1432 parliamentary 0.882353 -12.1432 spearman’s correlation 0.75 -12.1432 n−1 0.933333 -12.1432 agarwal 0.933333 -12.1432 notational 0.703704 -12.1432 phrase-table 0.772727 -12.1435 battery 0.296053 -12.1436 3). 0.933333 -12.1437 94.9 0.933333 -12.1437 jensen-shannon 0.933333 -12.1437 algebraic 0.933333 -12.1437 and ralph 0.933333 -12.1437 doubt 0.8 -12.1437 walking 0.842105 -12.1438 dialogue acts. 0.594595 -12.1438 party 0.933333 -12.1439 contributions: 0.933333 -12.1439 translators, 0.421053 -12.144 by selecting 0.375 -12.144 insight 0.164659 -12.1441 utterance 0.933333 -12.1441 25.4 0.933333 -12.1441 #1 0.842105 -12.1442 λi 0.933333 -12.1443 evolve 0.933333 -12.1443 human-machine 0.460317 -12.1444 we also show 0.933333 -12.1446 human judgment. 0.447761 -12.1447 (12) 0.933333 -12.1448 71.5 0.933333 -12.1448 decomposable 0.703704 -12.1449 pattern based 0.75 -12.145 grishman 0.933333 -12.145 stat-xfer 0.772727 -12.145 morphologically complex 0.842105 -12.1451 sys1 0.882353 -12.1453 82.4 0.453125 -12.1453 we collect 0.933333 -12.1453 meta-evaluation 0.666667 -12.1454 nd 0.933333 -12.1455 fourgram 0.390805 -12.1456 for improving 0.594595 -12.1457 y) 0.933333 -12.1457 gfs 0.933333 -12.1457 wordspace 0.933333 -12.1457 .82 0.882353 -12.1458 variables: 0.666667 -12.1459 features, including 0.678571 -12.1459 recipient 0.430556 -12.146 position, 0.933333 -12.146 allowed, 0.933333 -12.1462 (pitler 0.933333 -12.1462 lower-cased 0.933333 -12.1462 2009b) 0.933333 -12.1462 veale 0.933333 -12.1464 consistently outperform 0.842105 -12.1465 minimum-cut 0.666667 -12.1465 spain 0.678571 -12.1469 t., 0.933333 -12.1469 7.7 0.933333 -12.1469 concentration parameter 0.8 -12.1469 wordsim 0.8 -12.147 153 0.933333 -12.1471 filters, 0.933333 -12.1471 nm 0.8 -12.1472 providing us 0.14845 -12.1473 entire 0.441176 -12.1474 to explain 0.933333 -12.1478 longer, 0.933333 -12.1478 subjective, 0.320611 -12.1479 nearest 0.75 -12.148 system demonstrations, 0.933333 -12.148 months, 0.933333 -12.148 workers’ 0.103834 -12.1481 features. 0.933333 -12.1482 dificult 0.933333 -12.1482 9.3 0.5 -12.1483 al., 2001) 0.75 -12.1484 completely different 0.933333 -12.1485 214 0.933333 -12.1485 hypernymy 0.933333 -12.1485 8.3 0.882353 -12.1486 processing: 0.882353 -12.1486 data points, 0.933333 -12.1487 t-tests 0.8 -12.1488 sell 0.933333 -12.1489 dna methylation 0.441176 -12.1489 frequency. 0.441176 -12.1491 decoder. 0.933333 -12.1492 context-based 0.772727 -12.1492 72.0 0.933333 -12.1494 unique-event 0.474576 -12.1494 standard, 0.220217 -12.1494 translating 0.933333 -12.1496 subsume 0.933333 -12.1496 45.3 0.882353 -12.1497  the 0.678571 -12.1497 qi 0.933333 -12.1499 folds, 0.933333 -12.1499 fired 0.933333 -12.1499 75.2 0.882353 -12.1501 (words) 0.933333 -12.1501 lord 0.933333 -12.1501 manualy 0.933333 -12.1503 (salton 0.842105 -12.1504 different lengths 0.453125 -12.1505 und 0.441176 -12.1505 lexico-syntactic 0.430556 -12.1506 hovy, 0.75 -12.1506 conceived 0.490909 -12.1507 reciprocal 0.933333 -12.1508 mixes 0.933333 -12.151 allowable 0.192308 -12.1511 message 0.933333 -12.1512 (be 0.933333 -12.1515 73.9 0.933333 -12.1515 optima 0.39759 -12.1515 exchange 0.882353 -12.1516 appeal 0.520833 -12.1518 variable. 0.933333 -12.1519 recipients 0.0963271 -12.1521 up 0.933333 -12.1524 mturk, 0.933333 -12.1524 figure 1(a) 0.933333 -12.1524 flags 0.0930834 -12.1524 available 0.882353 -12.1525 ‘a 0.933333 -12.1526 me-seq 0.509804 -12.1526 peter 0.125 -12.1526 aligned 0.531915 -12.1528 null2 0.933333 -12.1531 80.3 0.933333 -12.1531 58% 0.053277 -12.1534 sentences 0.933333 -12.1535 work; 0.72 -12.1536 real users 0.636364 -12.154 too. 0.933333 -12.154 1999 0.351852 -12.1541 third, 0.933333 -12.1542 intends 0.933333 -12.1542 moderately 0.933333 -12.1542 modalities 0.772727 -12.1542 ok 0.320312 -12.1545 with multiple 0.933333 -12.1545 dibels 0.933333 -12.1545 syntax-semantics 0.666667 -12.1546 lexicon-based 0.703704 -12.1546 ..., 0.882353 -12.1547 p(c 0.177305 -12.1547 are shown in 0.933333 -12.1547 restrictor 0.933333 -12.1549 headden 0.72 -12.1549 a thorough 0.933333 -12.1552 pre-existing 0.933333 -12.1552 hand-picked 0.933333 -12.1552 mark, 0.645161 -12.1554 corpus statistics 0.933333 -12.1554 29.1 0.933333 -12.1554 48.9 0.933333 -12.1554 hyper-parameter 0.933333 -12.1559 training; 0.594595 -12.156 hmeant 0.678571 -12.1561 deposition 0.933333 -12.1561 days, 0.933333 -12.1561 49.0 0.933333 -12.1563 nouns) 0.933333 -12.1563 participant, 0.933333 -12.1563 moves, 0.933333 -12.1563 confined 0.678571 -12.1564 histories 0.367347 -12.1564 in: 0.0875286 -12.1565 natural 0.933333 -12.1566 drink 0.933333 -12.1566 “one 0.636364 -12.1568 30% of 0.933333 -12.1568 aachen 0.933333 -12.1568 scripts, 0.330579 -12.1568 we had 0.482143 -12.1569 dna 0.933333 -12.157 243 0.474576 -12.157 data; 0.933333 -12.1573 applicable, 0.933333 -12.1573 judgements, 0.8 -12.1573 96% 0.645161 -12.1575 metaphor 0.882353 -12.1575 large corpora, 0.72 -12.1575 shiny 0.594595 -12.1576 breakdown 0.281437 -12.1576 judged 0.933333 -12.1577 4-way 0.933333 -12.1577 tradeoffs 0.594595 -12.1579 each utterance 0.933333 -12.1579 hedging 0.933333 -12.1579 embodied 0.772727 -12.158 instance weighting 0.645161 -12.1583 training examples. 0.933333 -12.1584 achievable 0.0736671 -12.1585 better 0.933333 -12.1586 nobody 0.933333 -12.1586 disclosure 0.678571 -12.1588 instantiations of 0.933333 -12.1589 inefficient 0.933333 -12.1589 it; 0.933333 -12.1589 importance, 0.157407 -12.1589 we describe 0.8 -12.1589 and sebastiani, 0.8 -12.1589 binding events 0.666667 -12.1591 year. 0.933333 -12.1591 fw 0.933333 -12.1591 above: 0.933333 -12.1591 91% 0.933333 -12.1593 mappings, 0.933333 -12.1593 83.0 0.703704 -12.1594 vbz 0.289308 -12.1596 institute of 0.933333 -12.1596 open-class 0.128911 -12.1598 created 0.933333 -12.16 preconditions 0.933333 -12.16 significance, 0.933333 -12.16 macro-averaged 0.933333 -12.16 temporary 0.222222 -12.1602 element 0.933333 -12.1603 portability 0.933333 -12.1605 emissions 0.933333 -12.1607 operated 0.232653 -12.1607 arc 0.376344 -12.1609 sequences. 0.933333 -12.161 logged 0.933333 -12.1612 aggregates 0.933333 -12.1612 multi-way 0.933333 -12.1614 allen 0.933333 -12.1614 (total 0.575 -12.1616 expresions 0.933333 -12.1617 occupy 0.933333 -12.1617 an in-depth 0.159004 -12.1618 choice 0.460317 -12.1619 we restrict 0.933333 -12.1624 signal, 0.933333 -12.1624 comparative analysis 0.933333 -12.1624 yeah 0.39759 -12.1625 tag, 0.933333 -12.1626 5.6 0.933333 -12.1626 lives 0.72 -12.1627 jst model 0.8 -12.1627 by utilizing 0.933333 -12.1628 42% 0.933333 -12.1628 register 0.520833 -12.1628 turney 0.666667 -12.1629 extensible 0.560976 -12.1629 not enough 0.933333 -12.1635 duc-05 0.72 -12.1638 mcc 0.933333 -12.164 non-adaptive 0.933333 -12.1642 bj 0.933333 -12.1642 ⊂ 0.933333 -12.1642 iht 0.281437 -12.1642 n/a 0.367347 -12.1643 driven 0.933333 -12.1645 corpora; 0.933333 -12.1645 attitude holder 0.575 -12.1646 manager 0.772727 -12.1646 confluence 0.933333 -12.1647 split, 0.933333 -12.1647 53.5 0.933333 -12.1647 83.4 0.933333 -12.1654 grass 0.134328 -12.1655 whose 0.933333 -12.1656 baldridge, 0.933333 -12.1656 (10 0.73913 -12.1659 data mining 0.933333 -12.1661 opposition 0.8 -12.1661 released by 0.578947 -12.1661 h., 0.376344 -12.1662 figure 5 0.933333 -12.1663 defeasible 0.72 -12.1663 less common 0.382022 -12.1664 wm 0.72 -12.1665 logic. 0.933333 -12.1668 (subject 0.933333 -12.1668 sessions, 0.0890616 -12.1669 important 0.772727 -12.1671 tree) 0.72 -12.1672 ra 0.459016 -12.1673 our implementation 0.933333 -12.1673 sharp 0.317829 -12.1673 search. 0.8 -12.1673 dj 0.72 -12.1674 selector 0.933333 -12.1675 taxon 0.933333 -12.1677 phonetically 0.097897 -12.1678 future 0.933333 -12.168 4this 0.933333 -12.168 1992), 0.73913 -12.1683 review, 0.933333 -12.1684 nb, 0.933333 -12.1684 differentiated 0.434783 -12.1684 lavie, 0.308824 -12.1686 a threshold 0.933333 -12.1687 chains, 0.933333 -12.1687 concept-based 0.560976 -12.1689 nullnullnullnullnull 0.933333 -12.1689 attributes: 0.933333 -12.1689 admm 0.882353 -12.169 hotels 0.933333 -12.1692 rho 0.283951 -12.1693 queries. 0.933333 -12.1696 full, 0.434783 -12.1697 men 0.933333 -12.1701 (mintz et 0.933333 -12.1703 confidently 0.933333 -12.1703 depfix 0.578947 -12.1703 ˆy 0.8 -12.1705 ba 0.73913 -12.1706 num 0.313433 -12.1707 ways to 0.882353 -12.1708 away. 0.882353 -12.1708 kwiatkowski 0.73913 -12.1708 (2009b) 0.933333 -12.1708 (measured 0.933333 -12.1708 reranking, 0.285714 -12.1709 lot of 0.8 -12.1711 bag-of-words model 0.490566 -12.1712 pattern, 0.933333 -12.1717 need, 0.560976 -12.172 have become 0.933333 -12.172 44% 0.933333 -12.172 6.8 0.8 -12.172 beetle 0.933333 -12.1722 7.0 0.933333 -12.1722 posit 0.466667 -12.1722 decay 0.8 -12.1722 coast 0.126061 -12.1726 select 0.0519853 -12.1726 its 0.8 -12.1728 (blitzer 0.933333 -12.1729 217 0.833333 -12.173 (1): 0.482143 -12.1732 estimation. 0.933333 -12.1734 “i” 0.933333 -12.1734 for convenience, 0.933333 -12.1736 quickly, 0.933333 -12.1736 sizable 0.73913 -12.1741 further improvements 0.242152 -12.1743 defines 0.933333 -12.1743 auxiliary+verb 0.933333 -12.1743 basketball 0.933333 -12.1743 intonation 0.282209 -12.1743 are still 0.6 -12.1743 visiting 0.36 -12.1745 entries. 0.933333 -12.1746 consultation 0.933333 -12.1746 trainable 0.0798258 -12.1746 function 0.73913 -12.1747 target word, 0.201863 -12.1748 usa 0.933333 -12.175 question recommendation 0.120575 -12.1751 times 0.933333 -12.1753 after, 0.203175 -12.1754 to measure 0.933333 -12.1755 multi-level 0.933333 -12.1757 transparency 0.560976 -12.176 dependency trees. 0.8 -12.176 countries, 0.833333 -12.1761 baseline; 0.933333 -12.1762 mwe, 0.933333 -12.1762 automating 0.520833 -12.1763 summation 0.578947 -12.1763 would allow 0.933333 -12.1765 paraphrase, 0.933333 -12.1765 tt 0.933333 -12.1765 regulations 0.933333 -12.1767 weekend 0.933333 -12.1769 q3 0.72 -12.1769 8 conclusions 0.349057 -12.1769 microsoft 0.933333 -12.1772 architectures 0.251208 -12.1773 arguments. 0.933333 -12.1776 caller 0.692308 -12.1776 operational 0.8 -12.1776 to date, 0.8 -12.1776 embed 0.692308 -12.1778 (iv) 0.678571 -12.1779 is beneficial 0.933333 -12.1779 described below, 0.692308 -12.1779 processors 0.933333 -12.1783 corelation 0.342342 -12.1783 node, 0.194767 -12.1783 direction 0.8 -12.1784 inferred. 0.933333 -12.1786 flagged 0.428571 -12.1786 one could 0.833333 -12.1787 78.9 0.833333 -12.1787 deeply 0.833333 -12.1787 82% 0.933333 -12.1788 watch 0.833333 -12.1789 47.2 0.933333 -12.1791 268 0.645161 -12.1792 patern 0.678571 -12.1793 conference. 0.933333 -12.1793 head) 0.875 -12.1793 ϕ 0.875 -12.1793 100m 0.875 -12.1793 integers 0.143987 -12.1794 idea 0.73913 -12.1795 radev 0.875 -12.1795 (milne 0.875 -12.1795 disaster 0.8 -12.1796 issues related 0.208754 -12.1797 chose 0.286624 -12.1798 best results 0.933333 -12.1798 3.1), 0.875 -12.18 disagreement, 0.32 -12.18 option 0.833333 -12.1801 depvsm 0.490566 -12.1804 3-gram 0.933333 -12.1807 foundations 0.933333 -12.1807 interactively 0.5 -12.1807 most of these 0.933333 -12.1812 survey, 0.692308 -12.1813 imageability 0.875 -12.1813 cp-nets 0.833333 -12.1814 distributed over 0.73913 -12.1816 shifted 0.875 -12.1817 kybots 0.8 -12.1817 b., 0.547619 -12.1819 we don’t 0.933333 -12.1819 executes 0.8 -12.1819 77.4 0.73913 -12.182 89.8 0.875 -12.182 cyc 0.933333 -12.1821 conciliation 0.933333 -12.1821 subsets, 0.6 -12.1821 translations into 0.560976 -12.1823 interior 0.933333 -12.1824 emotion classification 0.933333 -12.1824 feature-level 0.933333 -12.1824 allen, 0.533333 -12.1825 ≡ 0.833333 -12.1825 186 0.8 -12.1825 1): 0.933333 -12.1826 pregroup 0.72 -12.1827 forth 0.617647 -12.1828 unlabeled attachment 0.833333 -12.1829 (eisner, 0.875 -12.1829 hired 0.6 -12.183 paradise 0.73913 -12.1831 more stable 0.933333 -12.1831 (some 0.933333 -12.1831 causative 0.933333 -12.1831 hill 0.345794 -12.1832 placed 0.933333 -12.1833 0.655172 -12.1836 protein transport 0.4 -12.1836 in, 0.933333 -12.1836 5this 0.875 -12.1838 administrative 0.413333 -12.1841 rooted 0.8 -12.1841 reusing 0.625 -12.1842 sem 0.678571 -12.1843 europarl, 0.72 -12.1844 factoids 0.349057 -12.1845 r1 0.8 -12.1845 nnf 0.72 -12.1846 orange 0.875 -12.1846 353 0.833333 -12.1848 68.3 0.73913 -12.185 vanderwende 0.0947441 -12.185 (e.g., 0.298611 -12.1853 variants of 0.875 -12.1853 227 0.73913 -12.1854 role induction 0.6 -12.1854 presented. 0.655172 -12.1855 constant. 0.833333 -12.1857 hidden state 0.875 -12.1858 robotic 0.560976 -12.1859 similarity function 0.833333 -12.1859 .03 0.428571 -12.186 (1999) 0.8 -12.186 skipped 0.692308 -12.1861 diagram of 0.875 -12.1862 reputation 0.875 -12.1864 1600 0.678571 -12.1865 implication 0.473684 -12.1866 sagae 0.833333 -12.1867 1984 0.72 -12.1868 possible reason 0.875 -12.1869 (201) 0.521739 -12.187 poon 0.875 -12.1873 hypergraph. 0.875 -12.1873 many, 0.144231 -12.1874 2009) 0.364583 -12.1875 probabilities, 0.105489 -12.1882 which are 0.875 -12.1882 pre-processed 0.875 -12.1882 n., 0.692308 -12.1885 further improved 1 -12.1887 terminology. 1 -12.1887 (1996). 1 -12.1887 patient. 1 -12.1887 mit press. 1 -12.1887 maltparser. 1 -12.1887 label). 1 -12.1887 t). 1 -12.1887 english–french 1 -12.1887 correlations. 1 -12.1887 oracle. 1 -12.1887 (no. 1 -12.1887 incomplete. 1 -12.1887 stimuli. 1 -12.1887 submodular. 1 -12.1887 proof. 1 -12.1887 formulation. 1 -12.1887 participation. 1 -12.1887 .62 1 -12.1887 n-gram. 1 -12.1887 1982). 1 -12.1887 intent. 1 -12.1887 ex. 1 -12.1887 flow. 1 -12.1887 manual. 1 -12.1887 co-occurrences. 1 -12.1887 (6). 1 -12.1887 pronunciation. 1 -12.1887 theirs. 1 -12.1887 polarities. 1 -12.1887 ∅. 1 -12.1887 revision. 1 -12.1887 forums. 1 -12.1887 (d). 1 -12.1887 [graça 1 -12.1887 divergence. 1 -12.1887 pipelines. 1 -12.1887 online. 1 -12.1887 student. 1 -12.1887 decoders. 1 -12.1887 frequent. 1 -12.1887 intervals. 1 -12.1887 correlated. 1 -12.1887 .73 1 -12.1887 affix stripping 1 -12.1887 minimal. 1 -12.1887 writer. 1 -12.1887 3.4. 1 -12.1887 allowed. 1 -12.1887 10). 1 -12.1887 slovene. 1 -12.1887 decades. 1 -12.1887 findings. 1 -12.1887 distinction. 1 -12.1887 contents. 1 -12.1887 urls. 1 -12.1887 dates. 1 -12.1887 documents). 1 -12.1887 june. 1 -12.1887 taggers. 1 -12.1887 judge. 1 -12.1887 phonemes. 1 -12.1887 reasonable. 1 -12.1887 gsc. 1 -12.1887 shown). 1 -12.1887 impairment. 1 -12.1887 means. 1 -12.1887 instances). 1 -12.1887 setups. 1 -12.1887 quotes. 1 -12.1887 nc-only strict 1 -12.1887 equally. 1 -12.1887 derived. 1 -12.1887 intervention. 1 -12.1887 acceptable. 1 -12.1887 learn. 1 -12.1887 synthesis. 1 -12.1887 education. 1 -12.1887 food. 1 -12.1887 tokenization. 1 -12.1887 interpolation. 1 -12.1887 votes. 1 -12.1887 co-training. 1 -12.1887 das. 1 -12.1887 i.e, 1 -12.1887 brevity. 0.24 -12.1888 they have 0.928571 -12.1889 li. 0.928571 -12.1889 restriction. 0.928571 -12.1889 slightly. 0.928571 -12.1889 quantifiers. 0.928571 -12.1889 breast cancer 0.875 -12.1889 mbr-sc 0.352941 -12.189 consequence 0.928571 -12.1891 v.s. 0.0723232 -12.1892 possible 0.875 -12.1896 human-robot 0.875 -12.1896 (2009a) 0.928571 -12.1899 verbs, adjectives, 0.8 -12.1899 chang, 0.875 -12.19 sources: 0.928571 -12.1901 punctuation variety 0.928571 -12.1901 new york: 0.875 -12.1903 pe 0.73913 -12.1907 november 0.928571 -12.1908 mirror 0.928571 -12.1908 37,number 1 0.246445 -12.1908 reasons 0.4 -12.191 ill-formed 0.439394 -12.1911 (mcdonald et 0.928571 -12.1915 emotional states 0.8 -12.1915 kybot 0.875 -12.1916 two-level 0.875 -12.1923 supplied by 0.875 -12.1923 washington, 0.928571 -12.1925 learner’s utterance 0.928571 -12.1925 non-compositional phrases 0.833333 -12.1925 liberal 0.875 -12.1925 eats 0.625 -12.1927 (landauer 0.875 -12.1927 cognitive science 0.761905 -12.1927 component-based 0.101026 -12.1928 note that 0.171625 -12.193 this approach 0.875 -12.1932 existential 0.761905 -12.1935 an scfg 0.451613 -12.1936 trade-off 0.928571 -12.1939 enhancements 0.439394 -12.1941 emphasis 0.928571 -12.1942 nakagawa 0.928571 -12.1942 hadoop 0.73913 -12.1945 mw 0.875 -12.1945 7.4 0.4 -12.1946 marker 0.73913 -12.1947 directive 0.521739 -12.1947 continue to 0.451613 -12.1948 links. 0.875 -12.1948 twelve 0.833333 -12.1951 biology, 0.875 -12.1952 past. 0.875 -12.1952 civilization 0.465517 -12.1955 networks. 0.0840659 -12.1955 case 0.928571 -12.1956 admit 0.395062 -12.1957 del 0.875 -12.1957 measurable 0.465517 -12.1958 we took 0.451613 -12.1958 benefits of 0.8 -12.1958 tokenisation 0.287582 -12.1958 determines 0.833333 -12.1959 multiply 0.8 -12.196 pattern: 0.928571 -12.1961 (madnani 0.833333 -12.1962 linux 0.692308 -12.1962 [0,1] 0.928571 -12.1963 (leacock 0.928571 -12.1963 (stage 0.928571 -12.1963 23.5 0.928571 -12.1966 ahead 0.140216 -12.1968 occur 0.4 -12.1968 parenrightbig 0.928571 -12.1968 1-to-1 0.655172 -12.1969 function f 0.8 -12.1969 pustejovsky, 0.928571 -12.197 jelinek 0.833333 -12.197 biological events 0.8 -12.1971 (all) 0.875 -12.1972 enjoy 0.928571 -12.1973 similarities, 0.928571 -12.1973 wordaligned 0.928571 -12.1973 small-scale 0.928571 -12.1975 chineseenglish 0.928571 -12.1975 graph-based ranking 0.833333 -12.1975 domains: 0.928571 -12.1978 largescale 0.928571 -12.1978 inspect 0.928571 -12.1978 trie, 0.0517529 -12.1978 to be 0.692308 -12.1978 argument instances 0.73913 -12.1979 11% 0.73913 -12.1979 null) 0.875 -12.1979 (1986) 0.928571 -12.1982 78.4 0.875 -12.1984 à 0.928571 -12.1985 event-aspect 0.928571 -12.1987 tiedemann 0.0919921 -12.1993 improve 0.928571 -12.1994 inspire 0.928571 -12.1994 haghighi et al. 0.6 -12.1996 mwen 0.8 -12.1996 szpektor 0.928571 -12.1997 33.2 0.6 -12.2003 hear 0.330435 -12.2003 obtained using 0.928571 -12.2004 transfers 0.928571 -12.2004 bind 0.388235 -12.2004 details. 0.928571 -12.2006 unprecedented 0.928571 -12.2006 listened 0.8 -12.2006 2009a; 0.8 -12.2008 we leave this 0.875 -12.2009 crm 0.928571 -12.2011 incrementally, 0.928571 -12.2011 degenerate 0.928571 -12.2011 systran 0.928571 -12.2011 contribution, 0.928571 -12.2011 38.9 0.692308 -12.2014 bannard and 0.578947 -12.2014 concentration 0.928571 -12.2016 abstractness ratings 0.928571 -12.2016 ?p 0.140649 -12.2017 the highest 0.294521 -12.2017 proposition 0.875 -12.2018 bird 0.294521 -12.202 0.0 0.564103 -12.202 hlt 0.928571 -12.2021 56.6 0.510204 -12.2021 (joachims, 0.833333 -12.2022 st. 0.928571 -12.2023 66.3 0.16777 -12.2026 2007) 0.928571 -12.2026 to summarize, 0.833333 -12.2026 transcribe 0.928571 -12.2028 deceptive opinions 0.928571 -12.2028 datadriven 0.928571 -12.2028 grandchild 0.928571 -12.2031 (sagae 0.928571 -12.2033 48.6 0.928571 -12.2033 pseudo-parallel 0.928571 -12.2033 22% 0.625 -12.2035 sie 0.928571 -12.2038 ,..., 0.875 -12.2038 word boundaries, 0.928571 -12.204 critically 0.8 -12.2041 player. 0.8 -12.2041 .88 0.928571 -12.2043 domain’s 0.928571 -12.2043 marcus 0.928571 -12.2045 almuhareb 0.928571 -12.2045 cognitively 0.928571 -12.2045 time; 0.928571 -12.2048 64-bit 0.928571 -12.2048 circles 0.928571 -12.2048 81.8 0.928571 -12.2048 39.9 0.928571 -12.205 reflective 0.928571 -12.2052 c-feel-it 0.928571 -12.2052 66.2 0.928571 -12.2052 featuring 0.928571 -12.2052 dirichlet priors 0.928571 -12.2052 0.3% 0.73913 -12.2054 afp 0.833333 -12.2055 motor 0.833333 -12.2055 wikipedia articles. 0.0482633 -12.2055 when 0.294521 -12.2057 verb, 0.928571 -12.2057 language-dependent 0.928571 -12.2057 .54 0.928571 -12.2057 non-syntactic 0.465517 -12.2059 an f-score 0.360825 -12.206 scheme, 0.289474 -12.206 measures. 0.928571 -12.206 archeology 0.928571 -12.2062 29.2 0.928571 -12.2062 human-computer 0.875 -12.2063 both directions, 0.928571 -12.2067 74.9 0.928571 -12.2069 realised 0.383721 -12.2069 recognition. 0.22093 -12.2071 seem 0.73913 -12.2073 different sources 0.875 -12.2075 cross language 0.6 -12.2076 affix 0.655172 -12.2077 paragraphs. 0.390244 -12.2077 correction. 0.928571 -12.2077 case study, 0.928571 -12.2077 strive 0.875 -12.2077 rock 0.875 -12.2077 monte-carlo 0.928571 -12.2079 descriptor 0.928571 -12.2079 67.5 0.564103 -12.2079 conclusions. 0.132231 -12.2081 left 0.928571 -12.2081 span-1 0.357143 -12.2084 δ 0.833333 -12.2085 first time, 0.928571 -12.2086 32.0 0.247573 -12.2088 sensitive 0.390244 -12.2088 swsd 0.156371 -12.2088 equation 0.578947 -12.2088 entire set 0.761905 -12.2089 overlaps with 0.422535 -12.2091 hyperparameters 0.0696784 -12.2091 so 0.928571 -12.2091 arguments) 0.928571 -12.2091 1988), 0.833333 -12.2092 professor 0.875 -12.2093 tourism 0.314961 -12.2096 freely 0.928571 -12.2096 data-text 0.928571 -12.2096 informally 0.928571 -12.2096 (copestake 0.928571 -12.2096 simulation, 0.6 -12.2096 discuss how 0.928571 -12.2098 parenthetical 0.761905 -12.2099 steadily 0.833333 -12.21 expert m 0.833333 -12.21 sheffield 0.432836 -12.2101 indication of 0.928571 -12.2103 whitespace 0.833333 -12.2103 companion 0.875 -12.2105 hans 0.655172 -12.2106 ead 0.928571 -12.2106 310 0.928571 -12.2106 95.5 0.73913 -12.2108 (task 0.385542 -12.2108 u. 0.928571 -12.2108 79.3 0.928571 -12.2108 (last 0.833333 -12.2109 (section 3), 0.833333 -12.2111 engaged in 0.833333 -12.2111 latency 0.833333 -12.2111 eq 0.472727 -12.2111 by setting 0.457627 -12.2113 5th 0.625 -12.2114 living 0.928571 -12.2115 (titov 0.928571 -12.2115 ruledef 0.833333 -12.2116 multinomials 0.10625 -12.2118 performed 0.761905 -12.2119 reweighting 0.833333 -12.212 b’s 0.928571 -12.2123 footprint 0.928571 -12.2125 uncertainty, 0.928571 -12.2125 citing 0.928571 -12.2125 73.8 0.833333 -12.2127 null0 0.761905 -12.2127 pearson’s correlation 0.928571 -12.2128 term/sentence 0.928571 -12.2128 proportionally 0.6 -12.2129 required, 0.928571 -12.213 lemmatisation 0.928571 -12.213 schoenmackers et al. 0.136499 -12.2131 expected 0.928571 -12.2133 arab 0.928571 -12.2133 unrealistic 0.928571 -12.2133 klose 0.353535 -12.2134 parsers. 0.928571 -12.2135 44.6 0.928571 -12.2135 bohus 0.761905 -12.2135 automatically learned 0.357143 -12.2138 confirms 0.102226 -12.2138 rather 0.465517 -12.2139 conversation, 0.761905 -12.2139 raters. 0.457627 -12.214 opi 0.928571 -12.214 5.9 0.928571 -12.214 aspect-oriented 0.928571 -12.214 actors 0.833333 -12.214 svm classifier. 0.268571 -12.214 others. 0.928571 -12.2145 batches 0.928571 -12.2145 84.4 0.928571 -12.2145 63.7 0.928571 -12.2147 outcomes, 0.928571 -12.2147 ukwac corpus 0.928571 -12.2147 equivalences 0.928571 -12.2147 transcriber 0.928571 -12.2147 cheng 0.928571 -12.2147 .35 0.625 -12.2148 sounds 0.283871 -12.2148 demonstrates 0.666667 -12.2149 evaluation results. 0.692308 -12.215 newstest2010 0.833333 -12.2153 fiction 0.188202 -12.2153 types. 0.928571 -12.2155 (gale 0.928571 -12.2155 barely 0.928571 -12.2157 62.3 0.928571 -12.2157 geared 0.928571 -12.2157 54.7 0.928571 -12.2157 11.1 0.928571 -12.2157 (second 0.761905 -12.2157 belgium 0.928571 -12.2159 autoencoder 0.833333 -12.2159 (test 0.708333 -12.2159 coherence, 0.606061 -12.216 leacock 0.928571 -12.2162 helpfulness, 0.928571 -12.2162 (approximately 0.928571 -12.2162 wbow 0.928571 -12.2164 wong, 0.633333 -12.2165 [the 0.833333 -12.2166 89.0 0.928571 -12.2167 formalizing 0.416667 -12.2168 cited 0.625 -12.2168 upward 0.928571 -12.2169 φ-edge 0.928571 -12.2169 subsystem 0.928571 -12.2169 differentiable 0.875 -12.2171 model’s performance 0.246377 -12.2172 mentioned in 0.375 -12.2173 extracted. 0.761905 -12.2177 simpleew 0.55 -12.2178 nsf grant 0.327586 -12.2179 we address 0.928571 -12.2179 permissible 0.928571 -12.2179 keywords, 0.928571 -12.2179 d2w 0.276074 -12.218 adapt 0.103596 -12.218 might 0.928571 -12.2181 human-authored 0.928571 -12.2181 62.7 0.928571 -12.2181 (lewis 0.928571 -12.2181 males 0.928571 -12.2181 6 0 . 0.330435 -12.2181 agreement. 0.564103 -12.2183 this stage 0.928571 -12.2184 55.9 0.928571 -12.2186 61.6 0.928571 -12.2186 represented, 0.410959 -12.2187 stanford parser 0.547619 -12.2187 (chang 0.666667 -12.2189 second order 0.928571 -12.2189 synchronization 0.928571 -12.2189 (nenkova 0.928571 -12.2189 punishment 0.928571 -12.2189 disregarding 0.833333 -12.2192 s4 0.789474 -12.2193 volumes of 0.666667 -12.2194 possible combinations 0.928571 -12.2194 transparent 0.789474 -12.2195 fun 0.789474 -12.2195 kuhn 0.510638 -12.2196 (2005). 0.564103 -12.2197 morning 0.833333 -12.2199 ncls 0.346154 -12.22 attached 0.583333 -12.2201 such cases 0.928571 -12.2201 passonneau, 0.928571 -12.2201 75.8 0.928571 -12.2201 fewest 0.928571 -12.2201 detections 0.42029 -12.2201 air 0.833333 -12.2203 grafting 0.928571 -12.2204 ambiguities, 0.209059 -12.2204 back 0.789474 -12.2206 specia 0.928571 -12.2206 confusable 0.928571 -12.2206 235 0.928571 -12.2206 normalisation, 0.928571 -12.2206 thin 0.875 -12.2206 matrix-space model 0.875 -12.2206 earthquakes 0.928571 -12.2208 .6 0.928571 -12.2208 cities 0.928571 -12.2208 expressivity 0.875 -12.2208 prosodic cues 0.761905 -12.2208 60.1 0.37931 -12.2209 dan 0.928571 -12.2211 entropy-based 0.928571 -12.2211 positivity 0.928571 -12.2211 manual annotation, 0.928571 -12.2211 compositionality, 0.521739 -12.2212 you have 0.928571 -12.2213 30. 0.928571 -12.2213 0.9, 0.42029 -12.2213 an experiment 0.708333 -12.2213 only considers 0.789474 -12.2216 disorder 0.4375 -12.2218 nr 0.928571 -12.2218 co-occurred 0.928571 -12.2218 frameset 0.875 -12.2218 island 0.521739 -12.2219 “x 0.833333 -12.2219 ent 0.655172 -12.2221 gender. 0.73913 -12.2221 instructors 0.928571 -12.2221 speeding 0.928571 -12.2221 88.0 0.928571 -12.2221 (1985) 0.928571 -12.2221 opinion analysis. 0.761905 -12.2222 (quirk 0.402597 -12.2225 ranking. 0.666667 -12.2226 tuples. 0.928571 -12.2226 contextualized 0.928571 -12.2226 rapp, 0.928571 -12.2226 violations 0.928571 -12.2226 rapport 0.789474 -12.2227 .86 0.761905 -12.2228 and baldridge 0.928571 -12.2231 short-term 0.55 -12.2232 award 0.928571 -12.2233 reranked 0.928571 -12.2233 domainindependent 0.708333 -12.2234 parsing accuracy. 0.928571 -12.2236 example) 0.928571 -12.2236 hearer 0.692308 -12.2236 encouraged 0.875 -12.2236 traversed 0.55 -12.2236 significant, 0.928571 -12.2238 uncertainty-adaptive 0.625 -12.2241 protein coreference 0.928571 -12.2241 contradictory 0.761905 -12.2242 part-of-speech, 0.928571 -12.2243 polysemy, 0.928571 -12.2243 unmarked 0.928571 -12.2245 71.3 0.928571 -12.2245 33.5 0.928571 -12.2245 m¨uller 0.666667 -12.2246 (2010), which 0.875 -12.2248 opted for 0.481481 -12.225 inner 0.928571 -12.225 ˆpmle 0.319672 -12.2252 spanning 0.875 -12.2252 rises 0.928571 -12.2253 locating 0.928571 -12.2255 helpful, 0.928571 -12.2255 singer 0.928571 -12.2255 checkitout 0.928571 -12.2255 s-v 0.928571 -12.2255 1996 0.928571 -12.2255 twitter’s 0.625 -12.2256 by calculating 0.875 -12.2257 (berger 0.875 -12.2257 topics: 0.928571 -12.2258 sima’an, 0.928571 -12.226 lowered 0.875 -12.2262 you, 0.666667 -12.2263 complex. 0.928571 -12.2263 coercions 0.928571 -12.2263 narrowing 0.928571 -12.2263 section 2 presents 0.928571 -12.2265 total) 0.606061 -12.2267 ieee 0.928571 -12.2268 (d 0.928571 -12.2268 three-step 0.928571 -12.2268 slides 0.928571 -12.2268 93.0 0.142857 -12.2268 predict 0.875 -12.2269 knew 0.357143 -12.227 gi 0.534884 -12.2272 æ 0.457627 -12.2273 which provides 0.928571 -12.2273 consistency, 0.928571 -12.2273 lvcsr 0.472727 -12.2273 closest to 0.789474 -12.2277 299 0.3125 -12.2278 we examine 0.666667 -12.2278 ranked according 0.583333 -12.2278 document level 0.928571 -12.2278 talukdar 0.875 -12.2278 〈i,j〉 0.875 -12.2278 part-of-speech tagger 0.833333 -12.2279 increase. 0.128137 -12.2281 achieved 0.606061 -12.2283 user utterances 0.928571 -12.2283 slavic 0.928571 -12.2283 q2 0.928571 -12.2285 originate 0.875 -12.2285 < 0.01). 0.327586 -12.2286 test data. 0.928571 -12.229 vadas 0.875 -12.229 name strings 0.310078 -12.2293 fall 0.426471 -12.2294 were found 0.928571 -12.2295 degradations 0.928571 -12.2295 statdef 0.928571 -12.2295 constituting 0.928571 -12.2295 “s” 0.5 -12.2296 which requires 0.928571 -12.2297 cross-validation, 0.928571 -12.2297 separator 0.928571 -12.2297 appending 0.928571 -12.2297 constraint-based 0.875 -12.2297 (often 0.761905 -12.2297 these problems, 0.633333 -12.2298 constructional 0.928571 -12.23 requesting 0.928571 -12.23 regulate 0.928571 -12.23 uniqueness 0.928571 -12.23 fhmm 0.928571 -12.2305 parsing) 0.928571 -12.2307 lefthand 0.457627 -12.2309 all four 0.392405 -12.2311 we collected 0.928571 -12.2312 types; 0.928571 -12.2312 form-based 0.928571 -12.2315 indexing, 0.303704 -12.2316 end, 0.789474 -12.2317 hand-written 0.928571 -12.232 kneser-ney smoothing. 0.928571 -12.232 7), 0.928571 -12.232 differently, 0.928571 -12.232 copa 0.430769 -12.2322 sources, 0.928571 -12.2322 s-factors 0.928571 -12.2322 18: 0.928571 -12.2322 59.5 0.875 -12.2322 role labeling, 0.261111 -12.2323 sequences of 0.172249 -12.2323 or more 0.928571 -12.2325 beating 0.583333 -12.2327 improves over 0.928571 -12.2327 train, 0.928571 -12.2327 symmetry 0.928571 -12.2332 1000-best 0.928571 -12.2332 197 0.928571 -12.2332 those, 0.875 -12.2332 weighted finite-state 0.875 -12.2334 ca. 0.928571 -12.2335 autoencoders 0.928571 -12.2335 deliberately 0.928571 -12.2335 cross-linguistic 0.928571 -12.2335 simfusion 0.567568 -12.2337 other languages, 0.928571 -12.2337 smartpg 0.928571 -12.2337 substructures 0.928571 -12.2337 nouveau-rouge 0.928571 -12.2337 specification, 0.928571 -12.2337 power-law 0.875 -12.2337 349 0.55 -12.2339 msd 0.789474 -12.234 to mitigate 0.522727 -12.234 children, 0.928571 -12.234 functional trees 0.282895 -12.2341 an accuracy 0.472727 -12.2341 tuned on 0.708333 -12.2341 completions 0.633333 -12.2342 in favor 0.928571 -12.2342 wrapper 0.271084 -12.2343 exception 0.606061 -12.2345 factorial 0.928571 -12.2345 occuring 0.928571 -12.2345 differentiation 0.666667 -12.2347 this confirms 0.928571 -12.2347 healthy 0.928571 -12.2347 axes 0.928571 -12.2347 digitized 0.928571 -12.2347 motivational 0.928571 -12.2347 denkowski, 0.426471 -12.2347 semantic distance 0.875 -12.2348 (schwenk, 0.761905 -12.2348 200k 0.928571 -12.235 minimum, 0.761905 -12.235 hatzivassiloglou 0.928571 -12.2352 264 0.928571 -12.2352 tnt 0.928571 -12.2352 off-line 0.928571 -12.2352 physically 0.522727 -12.2354 derivations. 0.522727 -12.2354 to quantify 0.583333 -12.2356 ready 0.928571 -12.2357 logarithmic 0.205479 -12.2358 itself 0.4375 -12.2359 term, 0.928571 -12.236 [end] 0.928571 -12.236 clubs 0.928571 -12.236 logical form. 0.928571 -12.236 neutrality 0.928571 -12.2362 stricter 0.928571 -12.2362 kamps 0.928571 -12.2362 minimise 0.928571 -12.2365 established authors 0.336449 -12.2365 generation. 0.375 -12.2366 t2 0.708333 -12.2366 0.11 0.928571 -12.2367 bird, 0.534884 -12.2369 (cohen 0.633333 -12.2371 grammatical error 0.928571 -12.2375 documentation, 0.875 -12.2376 cssc 0.928571 -12.238 column represents 0.928571 -12.238 5.1.1 0.761905 -12.238 yesterday 0.327434 -12.2382 floor 0.928571 -12.2382 semi-supervised learning. 0.928571 -12.2382 1986), 0.928571 -12.2382 2008): 0.928571 -12.2385 arizona 0.928571 -12.2387 linear-chain 0.789474 -12.2389 2, 3, 0.480769 -12.239 gather 0.875 -12.2391 association lexicon 0.928571 -12.2392 64.1 0.727273 -12.2393 developmental 0.633333 -12.2393 self-trained 0.583333 -12.2394 to address these 0.928571 -12.2395 enriches 0.928571 -12.2395 stand-off 0.606061 -12.2397 can potentially 0.928571 -12.2397 (source 0.928571 -12.2397 33.8 0.727273 -12.2399 violated 0.281046 -12.2399 hard to 0.633333 -12.24 mapping. 0.928571 -12.24 default, 0.928571 -12.24 sentence compression, 0.666667 -12.2401 faster. 0.928571 -12.2402 (nguyen et 0.928571 -12.2402 leaders 0.928571 -12.2405 reconcileconll 0.928571 -12.2405 parenthesis 0.761905 -12.2405 (pantel 0.928571 -12.2407 hash functions 0.588235 -12.2408 scisumm 0.534884 -12.2411 have seen 0.928571 -12.2412 floor, 0.789474 -12.2415 re-rank 0.928571 -12.2415 n) 0.928571 -12.2415 unvocalized 0.928571 -12.2415 2(b) 0.789474 -12.2417 four main 0.928571 -12.2417 1.3% 0.875 -12.2417 “catholic 0.823529 -12.242 university press. 0.928571 -12.242 adjustments 0.789474 -12.2421 qj 0.708333 -12.2421 cu-bojar 0.928571 -12.2422 xle 0.928571 -12.2422 luke 0.928571 -12.2425 marks, 0.928571 -12.2425 ir, 0.928571 -12.2427 overload 0.45 -12.2428 exponentially 0.928571 -12.243 “enjoy 0.928571 -12.243 aquaint 0.823529 -12.2431 216 0.534884 -12.2431 semantic head 0.928571 -12.2432 dash 0.928571 -12.2432 particles, 0.928571 -12.2435 therapeutic 0.928571 -12.2435 fictional 0.928571 -12.2435 toolkit, 0.708333 -12.2435 uninformative 0.708333 -12.2435 building, 0.115092 -12.2437 larger 0.392405 -12.2438 in combination with 0.567568 -12.2438 exception is 0.319328 -12.2438 written in 0.22 -12.2439 score, 0.928571 -12.244 human judgement 0.789474 -12.2443 1977) 0.305344 -12.2444 marks 0.464286 -12.2445 0.62 0.928571 -12.2445 wmt09 0.708333 -12.2445 crfsgd 0.928571 -12.2447 posteriori 0.761905 -12.2448 f score 0.210145 -12.245 = summationdisplay 0.928571 -12.245 parser; 0.789474 -12.2451 extreme, 0.928571 -12.2452 2002, 0.708333 -12.2452 fifty 0.823529 -12.2453 dts 0.0979624 -12.2453 every 0.633333 -12.2455 meets 0.789474 -12.2455 strongly associated 0.510638 -12.2456 teachers 0.928571 -12.2457 lp” 0.633333 -12.2458 en-de 0.789474 -12.2458 interesting observation 0.875 -12.2459 =summationtext 0.761905 -12.2459 f3 0.928571 -12.246 loops 0.606061 -12.2461 mistakenly 0.761905 -12.2461 comparisons, 0.442623 -12.2462 depicts 0.928571 -12.2465 knowledge sources, 0.928571 -12.2468 slot, 0.928571 -12.2468 consulted 0.928571 -12.2468 right-most 0.727273 -12.2472 simon 0.606061 -12.2474 this yields 0.567568 -12.2478 unit. 0.928571 -12.2478 209 0.928571 -12.2478 phrases: 0.928571 -12.2478 que, 0.928571 -12.248 89.5 0.928571 -12.248 post-level 0.55 -12.2483 e., 0.204082 -12.2484 these features 0.928571 -12.2485 knowledge: 0.928571 -12.2485 295 0.252632 -12.2487 illustrated 0.350515 -12.2489 we hope 0.823529 -12.2489 hermann 0.928571 -12.249 (c), 0.928571 -12.249 base-phrase 0.408451 -12.2491 problems with 0.928571 -12.2493 accumulate 0.928571 -12.2493 fixes 0.928571 -12.2493 ideas, 0.928571 -12.2493 bag-ofwords 0.727273 -12.2494 is required. 0.928571 -12.2495 sighan 0.789474 -12.2498 incremental syntactic 0.823529 -12.2503 “who 0.350515 -12.2504 dimensionality 0.727273 -12.2504 equation 2 0.761905 -12.2504 ˇ@ 0.4 -12.2505 to encode 0.928571 -12.2505 start, 0.928571 -12.2505 meaning: 0.761905 -12.2506 ter-plus 0.761905 -12.2506 provide evidence 0.68 -12.2508 matsumoto 0.928571 -12.2508 type-level 0.789474 -12.2509 likert 0.928571 -12.2511 afect 0.928571 -12.2511 morpheme segmentation 0.928571 -12.2511 up until 0.928571 -12.2511 water, 0.511111 -12.2512 are used, 0.928571 -12.2513 efects 0.928571 -12.2513 9the 0.928571 -12.2513 “the” 0.489796 -12.2516 labeler 0.789474 -12.2516 lexical choice 0.928571 -12.2516 lcs 0.45614 -12.2518 it requires 0.928571 -12.2518 type) 0.442623 -12.2518 (barzilay 0.789474 -12.252 marc 0.522727 -12.252 constructions, 0.789474 -12.2522 zipfian 0.928571 -12.2523 o., 0.928571 -12.2523 employee 0.633333 -12.2526 first order 0.211896 -12.2526 program 0.928571 -12.2526 sheer 0.489796 -12.2527 see how 0.761905 -12.2527 bio 0.727273 -12.2529 neighbour 0.761905 -12.2529 fold. 0.612903 -12.2529 rising 0.823529 -12.253 49.6 0.823529 -12.253 su, 0.666667 -12.2531 for testing, 0.928571 -12.2531 .06 0.928571 -12.2531 submitting 0.708333 -12.2533 br 0.174242 -12.2535 detect 0.928571 -12.2536 (< 0.928571 -12.2536 fourth, 0.928571 -12.2536 emotinet 0.928571 -12.2538 limitation, 0.5 -12.2538 scale, 0.928571 -12.2541 ): 0.392405 -12.2542 is used. 0.261364 -12.2543 distinction 0.928571 -12.2544 reformulated 0.928571 -12.2544 non-comparative 0.928571 -12.2546 evaluative 0.666667 -12.2548 comes with 0.928571 -12.2549 barcelona, 0.928571 -12.2549 (22) 0.612903 -12.2549 place, 0.25 -12.2552 the largest 0.928571 -12.2554 mgiza++ 0.928571 -12.2554 summer 0.928571 -12.2554 surrounded 0.382716 -12.2555 steps: 0.68 -12.2556 better performance. 0.928571 -12.2556 decreasing order 0.928571 -12.2556 programming, 0.928571 -12.2556 480 0.928571 -12.2556 cased 0.928571 -12.2556 problematic, 0.727273 -12.2557 cutwidth 0.928571 -12.2559 suc 0.823529 -12.2561 95.9 0.928571 -12.2561 nonempty 0.928571 -12.2561 x” 0.45614 -12.2562 can lead 0.928571 -12.2564 (# 0.928571 -12.2564 observe, 0.928571 -12.2564 similar, but 0.928571 -12.2564 meteor, 0.928571 -12.2569 contextual information. 0.22 -12.2571 baseline, 0.612903 -12.2571 (l) 0.928571 -12.2572 238 0.642857 -12.2573 international workshop 0.866667 -12.2579 autonomous 0.866667 -12.2581 coarser 0.333333 -12.2581 (liu 0.435484 -12.2582 ag 0.928571 -12.2582 (their 0.866667 -12.2584 promises 0.866667 -12.2584 guevara 0.928571 -12.2584 αj 0.928571 -12.2584 learned, 0.928571 -12.2584 club 0.68 -12.2586 alg. 0.612903 -12.2586 perspective. 0.928571 -12.2587 (s,t) 0.68 -12.2592 consideration, 0.928571 -12.2592 hrl 0.588235 -12.2593 smoothing. 0.928571 -12.2594 feature decay 0.789474 -12.2596 47% 0.928571 -12.2597 regularization parameter 0.928571 -12.2597 concrete, 0.866667 -12.2598 qg rules 0.727273 -12.2599 serial 0.612903 -12.2601 score: 0.866667 -12.2603 wordlm 0.866667 -12.2603 296 0.128415 -12.2605 subject 0.928571 -12.2605 soldier 0.350515 -12.2606 training corpus. 0.480769 -12.2606 ourselves 0.435484 -12.2607 labeling. 0.928571 -12.2607 interview 0.928571 -12.2607 generation: 0.928571 -12.2607 google’s 0.866667 -12.2608 datasets: 0.612903 -12.2608 attention. 0.243781 -12.2609 0.6 0.866667 -12.261 virtue 0.789474 -12.2611 update. 0.789474 -12.2611 their frequencies 0.866667 -12.2612 88.2 0.928571 -12.2612 kid 0.928571 -12.2612 bikes 0.928571 -12.2612 final 0.928571 -12.2615 simple yet 0.928571 -12.2615 œ˝@ 0.567568 -12.2616 some examples of 0.727273 -12.2617 loc 0.866667 -12.2617 accessor 0.928571 -12.2618 logics 0.928571 -12.2618 bruno 0.866667 -12.262 papineni, 0.928571 -12.262 coarsegrained 0.489796 -12.2621 (lines 0.823529 -12.2622 rule applications 0.68 -12.2622 kingdom 0.430769 -12.2622 per sentence 0.866667 -12.2622 p(t 0.727273 -12.2623 a1, 0.928571 -12.2623 weighted, 0.866667 -12.2624 discussed above, 0.68 -12.2626 catalysis 0.823529 -12.2627 transitions between 0.866667 -12.2627 precomputed 0.666667 -12.2628 sentiment expressions 0.536585 -12.2628 enju 0.139837 -12.2629 presents 0.346939 -12.2629 maximizes 0.666667 -12.263 misses 0.68 -12.2631 were used. 0.708333 -12.2633 denominator 0.823529 -12.2634 70.3 0.866667 -12.2634 orientation, 0.866667 -12.2636 successively 0.866667 -12.2636 to ascertain 0.866667 -12.2636 cz 0.866667 -12.2639 « 0.866667 -12.2639 (according 0.727273 -12.2643 prevention 0.173367 -12.2643 then, 0.866667 -12.2644 ibm1 scores 0.114917 -12.2644 distance 0.866667 -12.2646 washing 0.177249 -12.2646 methods. 0.372093 -12.2647 prediction. 0.708333 -12.265 overall quality 0.866667 -12.2651 predicted, 0.866667 -12.2651 forest-to-string 0.866667 -12.2656 aggressively 0.240385 -12.2658 below. 0.204225 -12.2661 classification. 0.727273 -12.2663 y∗ 0.823529 -12.2663 turkey 0.823529 -12.2663 0.0% 0.552632 -12.2663 speed. 0.823529 -12.2666 16. 0.823529 -12.267 in-vocabulary 0.866667 -12.267 zhu et al. 0.38961 -12.2676 alternatively, 0.866667 -12.2677 word: 0.789474 -12.2679 bleu scores. 0.247423 -12.2681 t, 0.727273 -12.2682 placed on 0.866667 -12.2682 279 0.552632 -12.2683 source language. 0.171149 -12.2683 measured 0.866667 -12.2684 inherit 0.866667 -12.2684 significantly reduces 0.612903 -12.2684 mert (och, 0.823529 -12.2686 hashtags, 0.072305 -12.2686 previous 1 -12.2687 66.67 1 -12.2687 report. 1 -12.2687 kb. 1 -12.2687 disfluencies. 1 -12.2687 it? 1 -12.2687 redundant. 1 -12.2687 specificity. 1 -12.2687 stretchy paterns 1 -12.2687 courses. 1 -12.2687 arguments). 1 -12.2687 core. 1 -12.2687 (7). 1 -12.2687 j). 1 -12.2687 dialog. 1 -12.2687 document-based 1 -12.2687 regularization. 1 -12.2687 pred. 1 -12.2687 get rid 1 -12.2687 0.3. 1 -12.2687 pass. 1 -12.2687 innovation. 1 -12.2687 verbnet. 1 -12.2687 identifier. 1 -12.2687 average). 1 -12.2687 transfer. 1 -12.2687 theories. 1 -12.2687 modification. 1 -12.2687 h2. 1 -12.2687 slow. 1 -12.2687 logistic regression, 1 -12.2687 disputants. 1 -12.2687 inefficient. 1 -12.2687 corrected. 1 -12.2687 longer. 1 -12.2687 treelet. 1 -12.2687 c). 1 -12.2687 subjective. 1 -12.2687 10%. 1 -12.2687 partition. 1 -12.2687 vice-versa. 1 -12.2687 specification. 1 -12.2687 over. 1 -12.2687 (2008a). 1 -12.2687 properly. 1 -12.2687 removal. 1 -12.2687 name-based 1 -12.2687 premise. 1 -12.2687 nn. 1 -12.2687 taken. 1 -12.2687 pi. 1 -12.2687 bigram. 1 -12.2687 causality. 1 -12.2687 array. 1 -12.2687   1 -12.2687 freebase. 1 -12.2687 teus 1 -12.2687 human. 1 -12.2687 wj. 1 -12.2687 optimum. 1 -12.2687 metadata. 1 -12.2687 1000. 1 -12.2687 ans. 1 -12.2687 conventions. 1 -12.2687 processor. 1 -12.2687 overlapped 1 -12.2687 spain. 1 -12.2687 improves. 1 -12.2687 repeated. 1 -12.2687 most. 1 -12.2687 highlighted. 1 -12.2687 scanning. 1 -12.2687 active. 1 -12.2687 competitive. 1 -12.2687 15. 1 -12.2687 beginning. 1 -12.2687 modeled. 1 -12.2687 realizations. 1 -12.2687 systems). 1 -12.2687 zens 1 -12.2687 former. 1 -12.2687 relations). 1 -12.2687 seattle, wa 1 -12.2687 depth. 1 -12.2687 capture. 1 -12.2687 axiom. 1 -12.2687 take. 1 -12.2687 theme. 1 -12.2687 distinct. 1 -12.2687 dialects. 1 -12.2687 motivation. 1 -12.2687 counterpart. 1 -12.2687 modified. 0.866667 -12.2689 parametrized 0.923077 -12.2689 ties. 0.923077 -12.2689 anaphora. 0.923077 -12.2689 social networking 0.923077 -12.2689 idea. 0.923077 -12.2689 pronominal resolution 0.923077 -12.2689 easily. 0.923077 -12.2689 city. 0.923077 -12.2689 objectives. 0.823529 -12.2691 53.9 0.727273 -12.2692 this quantity 0.923077 -12.2692 synchronized 0.923077 -12.2692 substitution ciphers. 0.552632 -12.2694 matches any 0.866667 -12.2694 particularly important 0.823529 -12.2695 terminate 0.789474 -12.2698 more recently 0.923077 -12.27 .” 0.923077 -12.27 colour associations 0.823529 -12.2702 appos 0.923077 -12.2702 27.8 0.789474 -12.2703 link. 0.294118 -12.2705  0.789474 -12.2705 guarantees that 0.923077 -12.2705 cube-pruning 0.240196 -12.2706 respective 0.866667 -12.2709 alternatively 0.536585 -12.2709 sri 0.789474 -12.2711 rm 0.823529 -12.2712 he, 0.52381 -12.2713 user state 0.536585 -12.2717 results demonstrate 0.866667 -12.2718 site, 0.866667 -12.2721 transformation, 0.260116 -12.2721 replace 0.75 -12.2723 readymade 0.866667 -12.2723 measure: 0.866667 -12.2725 section 3.3, 0.588235 -12.2725 giga 0.923077 -12.273 succinct 0.923077 -12.273 33.1 0.923077 -12.273 schroeder, 0.923077 -12.273 part-of-spech 0.923077 -12.273 weaken 0.789474 -12.2731 hull 0.923077 -12.2733 parsed, 0.923077 -12.2736 deserve 0.923077 -12.2738 34.6 0.75 -12.274 proofs 0.306452 -12.274 dictionary, 0.823529 -12.2741 top 15 0.823529 -12.2741 ution 0.727273 -12.2742 lse 0.923077 -12.2743 rationales 0.923077 -12.2743 petrov, 0.121287 -12.2745 report 0.789474 -12.2747 aplied 0.5 -12.2747 an incorrect 0.823529 -12.2748 vehicle 0.923077 -12.2749 expert, 0.923077 -12.2749 granularity, 0.923077 -12.2749 score). 0.216867 -12.275 phrases. 0.823529 -12.2751 hand-curated 0.789474 -12.2751 (see also 0.642857 -12.2751 popescu 0.923077 -12.2751 (habash, 0.823529 -12.2753 bombing 0.789474 -12.2756 play, 0.923077 -12.2756 (negative) 0.923077 -12.2759 (farkas 0.923077 -12.2761 marsi, 2006; 0.923077 -12.2767 opponent-based 0.642857 -12.2768 encoded as 0.394737 -12.2771 (callison-burch et 0.866667 -12.2774 summation over 0.923077 -12.2774 versions, 0.866667 -12.2776 peers 0.866667 -12.2776 67.2 0.866667 -12.2776 ig 0.378049 -12.2777 role. 0.311475 -12.2779 is called 0.789474 -12.278 sieves 0.642857 -12.278 textrank 0.923077 -12.278 lowers 0.789474 -12.2782 noted, 0.823529 -12.2783 n-ary 0.258621 -12.2784 framework, 0.866667 -12.2784 autism 0.923077 -12.2785 short) 0.68 -12.279 (miller, 0.75 -12.2792 c1, 0.923077 -12.2792 (2007a) 0.866667 -12.2793 vectorw 0.75 -12.2794 25.0 0.923077 -12.2795 .55 0.923077 -12.2795 33.7 0.789474 -12.28 c-lda-a 0.263158 -12.28 this, we 0.923077 -12.28 text-bound 0.923077 -12.28 heading 0.923077 -12.28 lms, 0.923077 -12.28 modules: 0.866667 -12.2803 punc 0.923077 -12.2803 (training 0.923077 -12.2803 wordnets 0.923077 -12.2803 sub-trees 0.190769 -12.2804 to compare 0.866667 -12.2805 human effort 0.823529 -12.2806 order) 0.923077 -12.2808 kudo 0.324324 -12.2809 this approach is 0.923077 -12.2811 fx2 0.59375 -12.2812 information content 0.823529 -12.2813 entity’s 0.923077 -12.2813 gale, 0.263158 -12.2814 correlation with 0.75 -12.2815 78% 0.144643 -12.2815 explore 0.789474 -12.2815 stringent 0.612903 -12.2815 by summing 0.923077 -12.2816 (rabiner, 0.923077 -12.2816 aac-like 0.68 -12.2817 safety 0.866667 -12.2818 346 0.866667 -12.2818 conceptually, 0.923077 -12.2818 supertagger, 0.923077 -12.2818 cristianini, 0.923077 -12.2821 49.3 0.923077 -12.2821 colloquial 0.923077 -12.2821 (roark 0.29771 -12.2823 bag-of-words 0.642857 -12.2824 hearst, 0.923077 -12.2824 sensed 0.923077 -12.2824 expansions 0.923077 -12.2824 recalls 0.411765 -12.2826 we think 0.923077 -12.2826 43.1 0.923077 -12.2826 snow et al. 0.923077 -12.2826 surface string 0.214008 -12.2826 errors in 0.823529 -12.2827 fβ=1 0.68 -12.2827 nist mt 0.34 -12.2828 figure 2, 0.552632 -12.2829 samples from 0.923077 -12.2829 speed-up 0.923077 -12.2829 b+ 0.866667 -12.283 table 1 reports 0.823529 -12.2831 74% 0.923077 -12.2831 languages; 0.75 -12.2834 theory gi 0.923077 -12.2834 37.0 0.923077 -12.2834 86.1 0.411765 -12.2837 guaranteed to 0.923077 -12.2837 (jiampojamarn 0.923077 -12.2837 optimising 0.923077 -12.2837 remarkably, 0.923077 -12.2837 90.3 0.923077 -12.2837 noisier 0.923077 -12.2837 15.6 0.823529 -12.2838 announced 0.39726 -12.2839 , where 0.923077 -12.2839 nonnative 0.923077 -12.2839 expository 0.923077 -12.2839 initialization, 0.612903 -12.2841 granularity. 0.923077 -12.2842 softmax-margin 0.75 -12.2843 f1 measure 0.462963 -12.2843 more details 0.727273 -12.2844 existing methods 0.923077 -12.2844 (e.g 0.923077 -12.2844 boost-micro 0.923077 -12.2844 over-fitting 0.923077 -12.2847 cycles, 0.160793 -12.2849 contribution 0.612903 -12.2849 prior, 0.75 -12.2853 s∗ 0.612903 -12.2853 discussion, 0.866667 -12.2854 candidatearcs 0.823529 -12.2855 schunn, 0.59375 -12.2855 10 times 0.923077 -12.2855 50.7 0.923077 -12.2855 occur, 0.866667 -12.2857 198 0.588235 -12.2857 ads 0.923077 -12.2857 (ng, 0.923077 -12.286 proceeded 0.823529 -12.2862 making them 0.823529 -12.2862 (soon et 0.207407 -12.2863 area 0.695652 -12.2863 in short, 0.923077 -12.2863 50.2 0.923077 -12.2863 terms) 0.347368 -12.2864 a clear 0.866667 -12.2864 domain expert 0.28777 -12.2865 distinguishing 0.923077 -12.2865 eigenvectors 0.923077 -12.2865 86.6 0.923077 -12.2865 yih, 0.823529 -12.2866 sadness 0.923077 -12.2868 31.9 0.823529 -12.2869 ⇒ 0.642857 -12.2871 (mihalcea 0.923077 -12.2871 disregard 0.923077 -12.2871 52.0 0.923077 -12.2871 (?) 0.923077 -12.2873 3.3.2 0.923077 -12.2873 notes, 0.923077 -12.2873 exact-tree 0.923077 -12.2876 untrusted 0.612903 -12.2877 snapshot 0.402778 -12.2878 lexical semantic 0.41791 -12.2879 grammars. 0.68 -12.288 lig 0.75 -12.2881 (best 0.923077 -12.2881 discourse structure, 0.923077 -12.2881 ^ 0.161435 -12.2882 statistically 0.923077 -12.2884 topology 0.571429 -12.2886 likewise 0.923077 -12.2886 77.1 0.923077 -12.2886 76.8 0.612903 -12.2886 eduard 0.923077 -12.2889 86.4 0.923077 -12.2889 lags 0.923077 -12.2889 costs, 0.923077 -12.2889 manipulating 0.823529 -12.289 colour association 0.823529 -12.289 celct 0.823529 -12.289 piecewise 0.727273 -12.2892 no additional 0.235577 -12.2894 converted 0.923077 -12.2894 vocabularies, 0.923077 -12.2894 wavelets 0.653846 -12.2896 event narrative 0.310924 -12.2896 may also 0.612903 -12.2896 colleagues 0.823529 -12.2897 269 0.923077 -12.2897 253 0.923077 -12.2897 41.6 0.59375 -12.2898 we wanted 0.923077 -12.2899 67.7 0.0436122 -12.2902 translation 0.48 -12.2902 during training 0.923077 -12.2902 feeding 0.923077 -12.2902 49.1 0.923077 -12.2902 four, 0.923077 -12.2902 productions, 0.923077 -12.2905 paragraph, 0.923077 -12.2907 (positive, 0.923077 -12.2907 authorial 0.59375 -12.2908 matter of 0.727273 -12.291 easier. 0.727273 -12.291 cheaper 0.923077 -12.291 approach; 0.923077 -12.291 elsner 0.823529 -12.2911 kate 0.727273 -12.2912 stringnet 0.923077 -12.2912 most notably 0.866667 -12.2913 automaticaly 0.727273 -12.2914 suffix. 0.923077 -12.2915 cky-style 0.923077 -12.2915 watanabe 0.923077 -12.2915 diminishes 0.923077 -12.2915 (c 0.866667 -12.2916 al-masdar 0.866667 -12.2918 2007); 0.866667 -12.2918 monday 0.642857 -12.2918 that integrates 0.923077 -12.292 9.6 0.923077 -12.292 palladian 0.923077 -12.292 subtrees, 0.653846 -12.2922 gate 0.75 -12.2923 common-sense 0.923077 -12.2923 pairs), 0.923077 -12.2923 kernel-based 0.612903 -12.2924 decide which 0.823529 -12.2925 demonstrate how 0.52381 -12.2925 limitation of 0.653846 -12.2926 thai 0.866667 -12.2926 335 0.923077 -12.2926 grec 0.923077 -12.2928 sekine 0.923077 -12.2928 systems; 0.68 -12.2929 directly related 0.923077 -12.2931 52.8 0.923077 -12.2931 lossy 0.923077 -12.2931 24.8 0.823529 -12.2932 syntactic structure, 0.923077 -12.2933 weighing 0.923077 -12.2933 (wer) 0.923077 -12.2933 76.7 0.923077 -12.2936 lines) 0.923077 -12.2936 84% 0.923077 -12.2936 9.8 0.923077 -12.2936 iob 0.923077 -12.2936 prohibitive 0.612903 -12.2936 positives. 0.59375 -12.2938 (agirre 0.923077 -12.2939 establishes 0.653846 -12.2941 (eds.), 0.923077 -12.2941 high-precision 0.923077 -12.2941 368 0.923077 -12.2941 radically 0.440678 -12.2944 taken as 0.923077 -12.2944 manner: 0.923077 -12.2944 vm 0.923077 -12.2944 65.0 0.866667 -12.2945 chunk, 0.923077 -12.2947 pragmatically 0.923077 -12.2947 rigorous 0.923077 -12.2947 1.1% 0.923077 -12.2947 implemented, 0.923077 -12.2949 55.7 0.923077 -12.2949 infant 0.48 -12.2951 it is easy 0.923077 -12.2952 28.3 0.923077 -12.2952 announcer 0.923077 -12.2952 talbot 0.923077 -12.2952 teach 0.923077 -12.2952 491 0.202128 -12.2953 finite 0.923077 -12.2954 cooccurring 0.923077 -12.2957 (joshi 0.923077 -12.2957 (svms) 0.923077 -12.2957 labels) 0.653846 -12.2958 feminine 0.823529 -12.296 poetry 0.923077 -12.296 (bos 0.923077 -12.296 94.1 0.923077 -12.2962 backing 0.923077 -12.2962 misspellings 0.923077 -12.2962 numerically 0.923077 -12.2965 25.3 0.379747 -12.2967 equivalents 0.148008 -12.2969 sequence of 0.52381 -12.297 tier 0.923077 -12.297 bikel 0.923077 -12.297 blocked 0.923077 -12.297 pdt 0.695652 -12.2971 tom 0.612903 -12.2971 world model 0.923077 -12.2973 4k 0.923077 -12.2973 re-ranked 0.923077 -12.2976 ideological 0.923077 -12.2976 84.0 0.59375 -12.2977 situations. 0.923077 -12.2978 stop-words 0.68 -12.2979 peripheral 0.923077 -12.2981 mdl 0.923077 -12.2981 biotopes 0.75 -12.2983 tree-rewriting 0.379747 -12.2984 translation model. 0.923077 -12.2986 manandhar, 0.923077 -12.2986 subsentential 0.923077 -12.2986 smoothing, 0.695652 -12.2987 matthew 0.552632 -12.2989 gtrec 0.923077 -12.2989 lim 0.923077 -12.2989 4in 0.923077 -12.2989 buying 0.923077 -12.2989 domain-relevant 0.923077 -12.2991 rank, 0.923077 -12.2991 collocation, 0.866667 -12.2992 automatically generating 0.571429 -12.2992 tuning, 0.923077 -12.2994 e: 0.923077 -12.2994 coincide 0.923077 -12.2994 nasukawa 0.923077 -12.2994 typologically 0.923077 -12.2994 81.1 0.823529 -12.2995 selftraining 0.104962 -12.2996 relative 0.379747 -12.2997 we describe our 0.923077 -12.2997 i+1 0.923077 -12.2997 ported 0.923077 -12.2997 coalesced with 0.68 -12.2999 some important 0.68 -12.2999 variable-length 0.923077 -12.2999 analyst 0.923077 -12.2999 82.2 0.923077 -12.2999 word(s) 0.823529 -12.3 accuracy, but 0.411765 -12.3001 p, 0.695652 -12.3001 two consecutive 0.923077 -12.3002 wordnet) 0.923077 -12.3002 crossdocument 0.923077 -12.3002 visiting probability 0.923077 -12.3002 (mei 0.347368 -12.3003 magnitude 0.642857 -12.3003 trbm 0.777778 -12.3004 o2 0.75 -12.3005 turn-initial 0.866667 -12.3005 complex predicate 0.923077 -12.3005 75.7 0.923077 -12.3005 57.3 0.923077 -12.3005 32.3 0.823529 -12.3007 p(a) 0.923077 -12.3007 kprelint 0.923077 -12.3007 pseudo function 0.923077 -12.3007 unfamiliar 0.923077 -12.3007 (2010)), 0.59375 -12.3008 low recall 0.777778 -12.3009 settled 0.923077 -12.301 comic 0.923077 -12.301 outscoping 0.923077 -12.301 engines, 0.923077 -12.301 1988; 0.777778 -12.3011 >, 0.923077 -12.3013 pairings 0.923077 -12.3013 copenhagen 0.823529 -12.3014 would probably 0.923077 -12.3015 arm 0.75 -12.3016 y∈y 0.823529 -12.3017 i love 0.777778 -12.3018 .22 0.923077 -12.3018 83.8 0.923077 -12.3018 15.7 0.923077 -12.302 luc 0.75 -12.3022 worst-case 0.75 -12.3022 the penn treebank, 0.923077 -12.3023 tuples, 0.923077 -12.3023 95.2 0.923077 -12.3023 permitting 0.777778 -12.3024 smart 0.923077 -12.3026 (taskar 0.866667 -12.3027 outscopes 0.369048 -12.3027 depicted in 0.923077 -12.3028 religion 0.653846 -12.3029 benefited 0.923077 -12.3031 62.0 0.923077 -12.3031 men, 0.923077 -12.3031 rejecting 0.923077 -12.3031 correspondingly 0.68 -12.3032 gene expression 0.511628 -12.3034 hyperedge 0.75 -12.3035 93.2 0.777778 -12.3036 beta 0.923077 -12.3036 nonzero 0.923077 -12.3036 60s 0.255682 -12.3038 terms. 0.923077 -12.3039 alphabet, 0.923077 -12.3039 did, 0.823529 -12.304 obtained via 0.75 -12.3041 ritter 0.777778 -12.3042 (23) 0.923077 -12.3042 ~w 0.923077 -12.3042 (lsa) 0.68 -12.3044 discourse role 0.923077 -12.3044 93.5 0.923077 -12.3044 individuals, 0.923077 -12.3044 channels 0.48 -12.3047 mining. 0.923077 -12.3047 accommodates 0.923077 -12.3047 meaningless 0.923077 -12.3047 sentiment-bearing 0.384615 -12.3049 an optimal 0.823529 -12.305 bow, 0.923077 -12.305 core, 0.923077 -12.305 relative-salience 0.923077 -12.305 issues: 0.538462 -12.3052 efficiency. 0.923077 -12.3052 substance 0.923077 -12.3052 60.8 0.923077 -12.3052 commander 0.489362 -12.3055 a fair 0.923077 -12.3055 grid, 0.923077 -12.3055 debt 0.149805 -12.3057 conclusions 0.200704 -12.3058 understand 0.923077 -12.3058 svms, 0.923077 -12.3058 23.2 0.923077 -12.3058 +0.01 0.923077 -12.3058 wf 0.923077 -12.306 72.3 0.62069 -12.3063 this corresponds to 0.923077 -12.3063 ahn 0.114548 -12.3065 version 0.923077 -12.3066 weighting, 0.923077 -12.3066 entity), 0.923077 -12.3066 high-order 0.923077 -12.3066 banea 0.923077 -12.3066 junk 0.59375 -12.3067 colorado 0.538462 -12.3067 amazon mechanical 0.923077 -12.3068 floating 0.923077 -12.3068 sets) 0.923077 -12.3068 rte-2 0.923077 -12.3068 37.5 0.777778 -12.3069 ˆw 0.5 -12.307 are derived from 0.923077 -12.3071 manually, 0.923077 -12.3071 11.6 0.75 -12.3072 60.5 0.923077 -12.3074 eaten 0.923077 -12.3074 bleu) 0.923077 -12.3074 rca 0.866667 -12.3075 vectors: 0.777778 -12.3076 lesk 0.923077 -12.3076 tsur 0.923077 -12.3076 output) 0.695652 -12.3081 61.5 0.330097 -12.3082 science foundation 0.923077 -12.3082 50.4 0.923077 -12.3082 4a 0.823529 -12.3083 coling 0.923077 -12.3084 singh 0.923077 -12.3084 extraction) 0.347368 -12.3085 al., 2005), 0.555556 -12.3086 most commonly 0.923077 -12.3087 lexicalisation 0.923077 -12.3087 conjoin 0.923077 -12.3087 sparql 0.923077 -12.3087 occurred, 0.823529 -12.3088 faster, 0.555556 -12.3089 (kumar 0.62069 -12.309 answer, 0.923077 -12.309 53.3 0.923077 -12.309 rte-5 0.538462 -12.3091 neighbours 0.538462 -12.3092 entity mention 0.923077 -12.3092 confusion, 0.923077 -12.3092 formalism, 0.216327 -12.3094 we observe 0.923077 -12.3095 dictation 0.923077 -12.3095 (α 0.777778 -12.3098 jan 0.923077 -12.3098 reduction, 0.923077 -12.3098 other) 0.923077 -12.3103 stadium 0.923077 -12.3103 57.6 0.923077 -12.3103 randomly, 0.575758 -12.3105 linear interpolation 0.777778 -12.3105 type: 0.923077 -12.3106 1987), 0.923077 -12.3106 elra 0.923077 -12.3106 colon 0.433333 -12.3107 disambiguation, 0.923077 -12.3108 discretization 0.923077 -12.3108 reformulate 0.866667 -12.311 164 0.241026 -12.3113 dataset, 0.923077 -12.3114 oversampling 0.086785 -12.3115 small 0.923077 -12.3116 disregarded 0.777778 -12.3117 45.9 0.923077 -12.3119 .04 0.923077 -12.3119 engaging 0.923077 -12.3119 dialog, 0.866667 -12.312 86.7 0.777778 -12.3121 53.4 0.866667 -12.3122 1 iff 0.923077 -12.3122 goldman 0.923077 -12.3122 proximal 0.923077 -12.3124 brandeis university 0.923077 -12.3127 organisms 0.923077 -12.3127 collaboratively 0.263804 -12.3129 summarized 0.866667 -12.313 vectors representing 0.777778 -12.313 o, 0.777778 -12.313 discriminant 0.923077 -12.313 (weighted) 0.923077 -12.313 considered: 0.923077 -12.313 raw text. 0.446429 -12.3132 processing (nlp) 0.415385 -12.3134 separately. 0.695652 -12.3134 reports. 0.777778 -12.3135 administration 0.923077 -12.3135 rte-6 0.923077 -12.3135 17, 0.695652 -12.3138 our observations 0.923077 -12.3138 multiscale 0.923077 -12.3138 gillick 0.923077 -12.3138 micro-average 0.237624 -12.3141 adds 0.923077 -12.3141 θr 0.923077 -12.3141 60.4 0.923077 -12.3141 wherein 0.653846 -12.3142 statistically significant 0.479167 -12.3143 other languages. 0.923077 -12.3143 23rd international 0.923077 -12.3143 dimensions: 0.777778 -12.3146 mead 0.923077 -12.3146 polarity-bearing 0.236453 -12.3148 2004), 0.315789 -12.3149 5 conclusion 0.923077 -12.3149 radial 0.923077 -12.3149 37.7 0.470588 -12.315 ideally, 0.923077 -12.3154 syllables, 0.923077 -12.3154 daelemans 0.866667 -12.3155 ter, 0.571429 -12.3157 bracketrightbig 0.923077 -12.3157 normalized-cut 0.923077 -12.3157 than, 0.923077 -12.3159 needs, 0.923077 -12.3159 joy, 0.923077 -12.3159 a2, 0.866667 -12.316 what kind 0.866667 -12.316 then: 0.923077 -12.3162 backend 0.923077 -12.3162 artifacts 0.866667 -12.3163 collapsed into 0.479167 -12.3164 obviously, 0.369048 -12.3164 are summarized 0.923077 -12.3165 treatment, 0.923077 -12.3165 endangered 0.923077 -12.3165 1980) 0.37037 -12.3166 confident 0.62069 -12.3167 examples include 0.923077 -12.3167 bangalore, 0.777778 -12.3171 largely due 0.923077 -12.3175 optimizations 0.923077 -12.3175 70.7 0.355556 -12.3178 to denote 0.923077 -12.3178 272 0.923077 -12.3178 three, 0.923077 -12.3178 classes) 0.923077 -12.3181 french-english, 0.923077 -12.3181 augments 0.653846 -12.3184 brown, 0.923077 -12.3184 collected, 0.923077 -12.3184 psmt 0.226244 -12.3185 bound 0.923077 -12.3186 natural-language 0.294574 -12.3187 mass 0.923077 -12.3189 7.5 0.923077 -12.3189 auto/auto 0.777778 -12.3191 wh 0.75 -12.3192 g0 0.923077 -12.3192 pre-process 0.777778 -12.3193 1:1 0.695652 -12.3194 holders 0.62069 -12.3194 reward function 0.923077 -12.3194 disadvantages 0.923077 -12.3194 intentional 0.866667 -12.3195 lexicon: 0.409091 -12.3196 examination 0.360465 -12.3197 substitute 0.1875 -12.3197 child 0.923077 -12.3197 (ab) 0.653846 -12.3198 miller 0.386667 -12.3198 texas 0.866667 -12.3198 gems 0.62069 -12.32 an exhaustive 0.923077 -12.32 re-implemented 0.923077 -12.32 ver 0.923077 -12.32 232 0.236453 -12.3201 prime 0.923077 -12.3202 profiling 0.923077 -12.3202 (= 0.37037 -12.3204 evaluated by 0.714286 -12.3205 .80 0.714286 -12.3205 english speakers 0.923077 -12.3205 mwetoolkit 0.923077 -12.3205 distinguishable 0.653846 -12.3207 a tendency 0.6 -12.3207 x3 0.653846 -12.3209 2-21 0.163121 -12.3211 that can be 0.923077 -12.3211 7-gram 0.75 -12.3213 nlp systems 0.923077 -12.3213 57.8 0.288889 -12.3215 entity. 0.923077 -12.3216 ah 0.714286 -12.3217 criticism 0.777778 -12.3219 bob 0.923077 -12.3219 alarm 0.777778 -12.3221 viewpoints 0.62069 -12.3222 three methods 0.653846 -12.3224 parallel, 0.381579 -12.3224 we leave 0.923077 -12.3224 feature-rich 0.479167 -12.3227 singapore 0.923077 -12.3227 93.3 0.923077 -12.3227 ethnic 0.18541 -12.3228 approaches to 0.866667 -12.3228 (f 0.438596 -12.3229 which takes 0.923077 -12.3229 semi-structured 0.923077 -12.3229 annotations: 0.6 -12.323 child language 0.59375 -12.3231 we hope that 0.59375 -12.3231 …, 0.923077 -12.3232 causation 0.923077 -12.3232 magazine 0.923077 -12.3232 scfg-based 0.162353 -12.3234 necessary 0.75 -12.3235 66.1 0.923077 -12.3235 rescored 0.174863 -12.3236 a significant 0.866667 -12.3236 wordnet.pt 0.695652 -12.3237 customers 0.575758 -12.3238 mle 0.923077 -12.3238 26.8 0.923077 -12.3238 relaxcor 0.923077 -12.324 sequencing 0.923077 -12.324 (used 0.262195 -12.3243 will have 0.923077 -12.3243 frustration 0.695652 -12.3245 scorers 0.923077 -12.3246 concave 0.62069 -12.3247 lengths. 0.923077 -12.3248 (ml) 0.923077 -12.3248 ¯m) 0.923077 -12.3248 jointly, 0.866667 -12.3249 p(x,y) 0.415385 -12.325 let’s 0.8125 -12.325 “we 0.386667 -12.3251 firstly, 0.923077 -12.3251 technical report 0.75 -12.3253 cooperative 0.866667 -12.3254 provided us 0.923077 -12.3254 reparandum 0.479167 -12.3255 significant differences 0.0456722 -12.3257 into 0.923077 -12.3257 arc, 0.0498135 -12.3261 most 0.923077 -12.3262 forward, 0.923077 -12.3265 dwds 0.923077 -12.3265 297 0.356322 -12.3266 describe how 0.923077 -12.3267 error: 0.923077 -12.3267 populating 0.923077 -12.3267 theories, 0.25731 -12.3269 (2009), 0.714286 -12.327 under some 0.923077 -12.327 lexicalize 0.923077 -12.327 pertain 0.923077 -12.327 behaves 0.923077 -12.327 isomorphism 0.923077 -12.327 flat, 0.8125 -12.3272 average) 0.923077 -12.3273 permission 0.923077 -12.3273 severity 0.923077 -12.3273 shortening 0.695652 -12.3274 realized as 0.923077 -12.3276 rich, 0.923077 -12.3276 qualia 0.555556 -12.3277 learned. 0.866667 -12.3277 about 10% 0.923077 -12.3278 79% 0.28169 -12.3278 tweets. 0.0851064 -12.3281 1. 0.923077 -12.3281 1976; 0.923077 -12.3281 20th 0.923077 -12.3281 corpus), 0.923077 -12.3281 notoriously 0.489362 -12.3282 these techniques 0.575758 -12.3283 agg 0.923077 -12.3284 (compare 0.409091 -12.3286 our main 0.923077 -12.3286 4.1) 0.923077 -12.3286 recommender 0.59375 -12.3289 parsing algorithms 0.218884 -12.3289 wmt 0.777778 -12.3289 wilcoxon 0.235294 -12.3289 constraints. 0.62069 -12.3289 to meet 0.923077 -12.3289 “other” 0.0930769 -12.329 way 0.282609 -12.3291 communicative 0.75 -12.3293 log-probability 0.433333 -12.3293 0.35 0.13799 -12.3294 aspects 0.62963 -12.3294 interpolate 0.75 -12.3297 manuscript 0.923077 -12.3297 multi-token 0.62963 -12.3298 emotions. 0.45283 -12.3299 identified. 0.923077 -12.33 (person, 0.923077 -12.33 louis 0.923077 -12.33 liberman 0.307692 -12.3301 room 0.138462 -12.3301 information. 0.695652 -12.3303 odds 0.777778 -12.3303 js 0.923077 -12.3303 look-up 0.653846 -12.3305 x) 0.777778 -12.3305 grandparent 0.923077 -12.3306 left: 0.923077 -12.3306 semantic primitives 0.923077 -12.3306 asr-08-train 0.923077 -12.3306 linear transformation 0.171958 -12.3307 2003) 0.45283 -12.3308 aligner 0.866667 -12.331 excel 0.923077 -12.3311 compounding 0.923077 -12.3311 considerations, 0.575758 -12.3312 components: 0.62963 -12.3312 ø 0.212 -12.3313 domain, 0.695652 -12.3313 protein names 0.62069 -12.3313 kay 0.381579 -12.3313 sources. 0.62963 -12.3314 cox 0.923077 -12.3314 wellknown 0.28169 -12.3314 form. 0.238579 -12.3315 returned 0.866667 -12.3315 (z 0.923077 -12.3316 beginning, 0.866667 -12.3318 resolved. 0.923077 -12.3319 english, spanish, 0.315315 -12.3322 translations from 0.923077 -12.3322 wise 0.923077 -12.3322 unresolved 0.923077 -12.3322 213 0.269737 -12.3323 abstract this paper 0.45283 -12.3323 good results 0.777778 -12.3324 best known 0.653846 -12.3325 heuristic. 0.923077 -12.3325 two types: 0.923077 -12.3327 supervised, 0.923077 -12.3327 positioning 0.666667 -12.333 pa, 0.923077 -12.333 1977; 0.75 -12.3333 fish 0.923077 -12.3333 dashes 0.923077 -12.3333 electron 0.224215 -12.3334 ] 0.923077 -12.3336 canadian 0.300813 -12.3338 1999; 0.923077 -12.3338 lfs 0.923077 -12.3338 suicide notes 0.310345 -12.3339 s = 0.415385 -12.3342 frequent words 0.575758 -12.3343 must not 0.666667 -12.3344 scored by 0.923077 -12.3344 clues, 0.42623 -12.3347 statistic 0.923077 -12.3347 of” 0.923077 -12.3347 jeremiah 0.923077 -12.3347 shortened 0.923077 -12.3347 (bow) 0.512195 -12.3348 numbers. 0.8125 -12.3349 phrase pair, 0.777778 -12.3352 63.8 0.923077 -12.3352 reusable 0.923077 -12.3352 sad 0.923077 -12.3352 emitting 0.923077 -12.3352 55.0 0.8125 -12.3354 optimization problem. 0.0407951 -12.3355 different 0.777778 -12.3356 operating on 0.923077 -12.3357 jelinek-mercer 0.695652 -12.3361 expectation-maximization 0.488889 -12.3361 700 0.277778 -12.3365 this process 0.8125 -12.3366 old, 0.923077 -12.3366 scarcity 0.923077 -12.3368 straightforward: 0.653846 -12.3369 mae 0.62069 -12.337 camera 0.923077 -12.3371 pooled 0.6 -12.3372 deaf 0.3125 -12.3373 available at 0.525 -12.3374 scenarios, 0.666667 -12.3374 jazzy 0.923077 -12.3374 schwenk, 0.923077 -12.3374 interrupted 0.714286 -12.3375 unsupervised parsing 0.666667 -12.3377 craft 0.923077 -12.3377 tokens), 0.923077 -12.3379 assemble 0.923077 -12.3379 ln 0.8125 -12.338 wmt08 0.5 -12.338 pos-tagging 0.8125 -12.3385 currently being 0.8125 -12.3385 player’s 0.923077 -12.3385 minimized 0.264151 -12.3387 stored 0.923077 -12.3388 85.8 0.923077 -12.339 correlated, 0.1125 -12.3392 hypothesis 0.409091 -12.3392 ny 0.214876 -12.3393 seen in 0.923077 -12.3393 (berg-kirkpatrick et 0.923077 -12.3393 textbook 0.8125 -12.3395 (lower 0.653846 -12.3396 supervised classification 0.923077 -12.3396 columns, 0.575758 -12.3398 dbpedia 0.653846 -12.3398 can reduce 0.336735 -12.3401 to exploit 0.8125 -12.3402 organizations, 0.8125 -12.3404 diminishing 0.8125 -12.3404 79.0 0.8125 -12.3404 −→ 0.923077 -12.3404 uss 0.461538 -12.3405 function: 0.126255 -12.3406 al., 2010). 0.512195 -12.3406 given. 0.8125 -12.3407 mpqa corpus 0.777778 -12.3409 americans 0.923077 -12.341 singular values 0.923077 -12.3412 w., 0.8125 -12.3414 72.6 0.923077 -12.3415 symmetric, 0.923077 -12.3415 commented 0.923077 -12.3418 cutoffs 0.923077 -12.3418 short-range 0.257485 -12.342 central 0.777778 -12.3421 department, 0.42623 -12.3422 1997) 0.923077 -12.3424 stanford’s 0.555556 -12.3425 an ambiguous 0.777778 -12.3426 v3 0.923077 -12.3426 adaptor 0.356322 -12.3427 = 1, 0.62963 -12.3427 demonstrated by 0.232673 -12.343 logistic 0.558824 -12.343 joint learning 0.857143 -12.3431 mac 0.857143 -12.3431 hold, 0.923077 -12.3432 6000 0.923077 -12.3432 seriously 0.923077 -12.3432 gathers 0.8125 -12.3433 78.6 0.857143 -12.3434 6a 0.402985 -12.3435 1the 0.923077 -12.3435 10we 0.923077 -12.3435 w1,w2 0.923077 -12.3437 cw 0.923077 -12.3437 drive, 0.8125 -12.3438 87.6 0.512195 -12.3438 koehn, 0.857143 -12.3439 biocreative 0.857143 -12.3439 outlier 0.479167 -12.3439 a compact 0.923077 -12.344 component-wise 0.6 -12.3441 poon and 0.857143 -12.3442 vw 0.857143 -12.3442 ethnicity 0.857143 -12.3444 2see 0.555556 -12.3444 prior distribution 0.333333 -12.3446 sides 0.923077 -12.3446 212 0.923077 -12.3448 target: 0.923077 -12.3448 underline 0.857143 -12.3449 dimple 0.469388 -12.3452 infinite 0.8125 -12.3455 agenda, 0.479167 -12.3456 are removed 0.857143 -12.3457 enlarged 0.857143 -12.3457 hopes 0.857143 -12.3457 2007): 0.540541 -12.3458 25% of 0.8125 -12.3458 decca 0.923077 -12.3459 aphasic 0.923077 -12.3459 oceanic 0.857143 -12.346 < .001 0.857143 -12.346 medicine, 0.714286 -12.3461 dead 0.857143 -12.3462 (relative 0.857143 -12.3465 writes 0.714286 -12.3465 turney, 0.777778 -12.3465 law, 0.62963 -12.3466 (2001). 0.857143 -12.3467 t∗ 0.923077 -12.3468 longdistance 0.857143 -12.347 16.7 0.419355 -12.347 0.03 0.8125 -12.347 28% 0.32381 -12.3471 are used for 0.8125 -12.3472 (zettlemoyer 0.438596 -12.3473 labeled data. 0.923077 -12.3473 target’s 0.8125 -12.3475 at, 0.857143 -12.3478 tokens: 0.525 -12.3478 synsets. 0.714286 -12.348 china, 0.777778 -12.3482 xt 0.857143 -12.3483 2009): 0.666667 -12.3483 open-domain 0.575758 -12.3484 pittsburgh 0.228571 -12.3487 feature. 0.857143 -12.3491 181 0.777778 -12.3491 initialized with 0.857143 -12.3493 ’09 0.8125 -12.3494 81.6 0.857143 -12.3496 typedm 0.857143 -12.3496 (extracted 0.857143 -12.3496 copular 0.666667 -12.3497 zhao, 0.62963 -12.3497 paraphrase candidates 0.479167 -12.3499 supplied 0.333333 -12.3499 report results 0.857143 -12.3503 an ordinary 0.137694 -12.3503 (i.e., 0.857143 -12.3506 ll 0.857143 -12.3509 top-n 0.227488 -12.3509 this problem 0.666667 -12.3513 aic 0.857143 -12.3514 production, 0.3 -12.3514 are assigned 0.857143 -12.3516 302 0.525 -12.3516 parts: 0.857143 -12.3519 225 0.291339 -12.3522 6.2 0.540541 -12.3526 srilm toolkit 0.8125 -12.3526 1.0, 0.714286 -12.3527 57.1 0.148297 -12.3528 remaining 0.714286 -12.3529 by concatenating 0.575758 -12.3531 build on 0.736842 -12.3532 .14 0.12259 -12.3533 difficult 0.580645 -12.3533 prediction accuracy 0.857143 -12.3534 gaze 0.512195 -12.3535 see section 0.340426 -12.3536 group. 0.857143 -12.3537 lgfn 0.419355 -12.3539 arg0 0.8125 -12.3539 graham 0.15222 -12.3541 module 0.857143 -12.3542 cities, 0.857143 -12.3542 union, 0.0663984 -12.3543 well 0.526316 -12.3547 more generally, 0.6 -12.3548 als 0.857143 -12.355 arguments: 0.8125 -12.3551 wfsa 0.402985 -12.3552 name, 0.326733 -12.3552 dependencies. 0.857143 -12.3553 4.3.3 0.423729 -12.3553 ar 0.8125 -12.3553 for each topic, 0.0556183 -12.3556 may 0.736842 -12.3557 to uncover 1 -12.3557 cbc 1 -12.3557 satisfactory. 1 -12.3557 cultural heritage 1 -12.3557 background. 1 -12.3557 1998. 1 -12.3557 subramanya 1 -12.3557 periods. 1 -12.3557 prevalent. 1 -12.3557 length). 1 -12.3557 vp. 1 -12.3557 keyboard. 1 -12.3557 discourses 1 -12.3557 durations. 1 -12.3557 profile. 1 -12.3557 improve. 1 -12.3557 discrepancies. 1 -12.3557 σ∗. 1 -12.3557 (graff, 1 -12.3557 i.e 1 -12.3557 others). 1 -12.3557 procedures. 1 -12.3557 regularity. 1 -12.3557 productions. 1 -12.3557 century. 1 -12.3557 nes. 1 -12.3557 comprehension. 1 -12.3557 verb). 1 -12.3557 determiner. 1 -12.3557 φ. 1 -12.3557 mark ov 1 -12.3557 breath group 1 -12.3557 predict. 1 -12.3557 (§4). 1 -12.3557 negligible. 1 -12.3557 potential. 1 -12.3557 reversed. 1 -12.3557 summarizers. 1 -12.3557 recommending 1 -12.3557 jwpl 1 -12.3557 infrastructure. 1 -12.3557 feature) 1 -12.3557 klein. 1 -12.3557 floor. 1 -12.3557 .93 1 -12.3557 construct. 1 -12.3557 exact. 1 -12.3557 meaningful. 1 -12.3557 expressed. 1 -12.3557 instruction. 1 -12.3557 identity. 1 -12.3557 sds. 1 -12.3557 results). 1 -12.3557 produce. 1 -12.3557 regulation. 1 -12.3557 2.2). 1 -12.3557 yi. 1 -12.3557 lossless 1 -12.3557 lab. 1 -12.3557 splits. 1 -12.3557 ccgbank (hockenmaier 1 -12.3557 (zens 1 -12.3557 expresions. 1 -12.3557 labeler. 1 -12.3557 1b. 1 -12.3557 2008)). 1 -12.3557 (jeon 1 -12.3557 variety. 1 -12.3557 bacteria. 1 -12.3557 narrative. 1 -12.3557 variance. 1 -12.3557 implementations. 1 -12.3557 perform. 1 -12.3557 (ex. 1 -12.3557 progress. 1 -12.3557 ranker. 1 -12.3557 italy. 1 -12.3557 gigaword. 1 -12.3557 make. 1 -12.3557 not). 1 -12.3557 self-training. 1 -12.3557 hand crafted 1 -12.3557 remarks. 1 -12.3557 207). 1 -12.3557 sortal restrictions 1 -12.3557 class). 1 -12.3557 (mcnamee 1 -12.3557 density. 1 -12.3557 days. 1 -12.3557 left). 1 -12.3557 syntactically. 1 -12.3557 overlaps. 1 -12.3557 walk. 1 -12.3557 rest. 1 -12.3557 likely. 1 -12.3557 prague, czech 1 -12.3557 replacement. 1 -12.3557 deletions. 1 -12.3557 collocation. 1 -12.3557 predictor. 1 -12.3557 alternative. 1 -12.3557 tutoring. 1 -12.3557 drift. 1 -12.3557 inferences. 1 -12.3557 frequently. 1 -12.3557 europarl. 1 -12.3557 possibilities. 1 -12.3557 value). 1 -12.3557 linguists. 1 -12.3557 miltsakaki 1 -12.3557 trees). 1 -12.3557 interesting. 1 -12.3557 gloss. 1 -12.3557 frameworks. 1 -12.3557 desired. 1 -12.3557 e1 e2 1 -12.3557 animation 1 -12.3557 ends. 1 -12.3557 (stamatatos, 1 -12.3557 reasoning. 1 -12.3557 exploration. 1 -12.3557 ratios. 1 -12.3557 ungrammatical. 0.8125 -12.3558 more meaningful 0.8125 -12.3558 36.8 0.916667 -12.356 behaviour. 0.916667 -12.356 pasca 0.916667 -12.356 handled. 0.916667 -12.356 portuguese varieties 0.916667 -12.356 hit. 0.580645 -12.3561 eval 0.857143 -12.3563 5in 0.916667 -12.3563 encodings 0.916667 -12.3563 cross-language linked 0.916667 -12.3563 shared neighbours 0.174648 -12.3564 our work 0.777778 -12.3564 dni 0.857143 -12.3566 “more 0.916667 -12.3566 beam-width prediction 0.228571 -12.3568 been shown 0.777778 -12.3569 enhanced with 0.3125 -12.357 permutation 0.916667 -12.3571 pas¸ca 0.916667 -12.3571 claims, 0.0890014 -12.3574 being 0.916667 -12.3574 co-training, 0.857143 -12.3576 tenth 0.736842 -12.3577 first introduced 0.916667 -12.3577 adjacency pairs 0.666667 -12.3578 very much 0.62963 -12.3578 that. 0.857143 -12.3579 lis 0.916667 -12.3579 hierarchical reinforcement 0.575758 -12.358 tilburg 0.419355 -12.358 after each 0.8125 -12.358 (id) 0.857143 -12.3581 this issue. 0.575758 -12.3581 test instances 0.171662 -12.3581 impact of 0.916667 -12.3582 random restarts 0.857143 -12.3584 a50 0.526316 -12.3585 robustness of 0.916667 -12.3585 finite automaton 0.916667 -12.3588 morphological analyzer. 0.857143 -12.3589 predicates: 0.371795 -12.3589 have already 0.857143 -12.3592 (r2 0.916667 -12.3593 succeeds 0.916667 -12.3593 evidence propagation 0.916667 -12.3593 .53 0.8125 -12.3595 adjunction 0.857143 -12.3597 nlp applications, 0.173554 -12.3597 relationship 0.8125 -12.3598 rae 0.857143 -12.3599 prices 0.736842 -12.3601 (3 0.62963 -12.3601 terrorist 0.314815 -12.3602 an overview 0.916667 -12.3602 college park, 0.916667 -12.3602 2009b; 0.916667 -12.3602 reliability, 0.916667 -12.3602 scattered 0.777778 -12.3602 inf 0.736842 -12.3604 usp 0.857143 -12.3605 interacted 0.916667 -12.3605 picked up 0.512195 -12.3605 entailment. 0.916667 -12.3607 lexicographers 0.46 -12.3607 are now 0.558824 -12.3608 transferred 0.6 -12.3608 to reorder 0.857143 -12.361 13th 0.857143 -12.361 an algebra 0.714286 -12.361 mmil 0.580645 -12.3612 are updated 0.231527 -12.3613 increase in 0.916667 -12.3613 phrased 0.916667 -12.3613 (hatzivassiloglou 0.916667 -12.3613 precisely what 0.916667 -12.3616 repetition, 0.916667 -12.3616 non-pronominal 0.777778 -12.3616 (i,j) 0.777778 -12.3616 will increase 0.580645 -12.3616 for comparison, we 0.8125 -12.3618 ipus 0.916667 -12.3619 nonprojective 0.714286 -12.3621 joseph 0.120053 -12.3621 need to 0.714286 -12.3623 grammatical relation 0.62963 -12.3624 conll-2011 0.916667 -12.3627 (allauzen 0.916667 -12.3627 clause boundary 0.916667 -12.3627 machine learning-based 0.916667 -12.3627 meaningfully 0.916667 -12.3627 (figure 2) 0.857143 -12.3628 246 0.777778 -12.363 equations. 0.383562 -12.3631 retrieval, 0.916667 -12.3632 luo 0.916667 -12.3632 86.5 0.240642 -12.3632 caused 0.916667 -12.3635 lend 0.916667 -12.3635 alternates between 0.857143 -12.3639 ini 0.302521 -12.3639 transform 0.195035 -12.3641 be applied 0.916667 -12.3641 true positives. 0.916667 -12.3641 .39 0.5 -12.3641 concatenating 0.326733 -12.3642 yields a 0.6 -12.3643 require: 0.259259 -12.3643 oregon, usa, 0.857143 -12.3644 (vapnik, 0.916667 -12.3644 keyword-based 0.916667 -12.3644 two-phase 0.8125 -12.3647 (f1) 0.8125 -12.3647 buttons 0.916667 -12.3649 subtask, 0.580645 -12.365 phenomenon. 0.16624 -12.3651 but also 0.857143 -12.3652 bolded 0.916667 -12.3655 abovementioned 0.251462 -12.3657 graph, 0.8125 -12.3657 tough 0.777778 -12.3659 npj 0.736842 -12.3662 policy: 0.736842 -12.3662 angbracketrightbig 0.916667 -12.3663 htmm 0.916667 -12.3663 64.8 0.444444 -12.3665 figure 1). 0.916667 -12.3666 isomorphic 0.857143 -12.3668 fruit 0.478261 -12.3668 2004. 0.916667 -12.3669 densely 0.916667 -12.3669 extensibility 0.916667 -12.3669 32.7 0.916667 -12.3669 bionlp-st 2011 0.857143 -12.3675 9we 0.857143 -12.3675 rank labels 0.916667 -12.3675 voice search 0.666667 -12.3676 uncommon 0.736842 -12.3676 noah a. 0.62963 -12.3676 nullt 0.916667 -12.3677 sls 0.916667 -12.3677 adjudicated 0.916667 -12.368 25,000 0.857143 -12.3681 sst 0.857143 -12.3681 outdegree 0.736842 -12.3682 30.0 0.580645 -12.3682 theory. 0.916667 -12.3686 randomised 0.736842 -12.3687 cardinality of 0.6 -12.3687 period. 0.857143 -12.3689 search: 0.736842 -12.3689 (p = 0.916667 -12.3689 multi-party 0.916667 -12.3689 weekly 0.857143 -12.3691 lookup, 0.64 -12.3691 2010b) 0.681818 -12.3691 esuli 0.916667 -12.3691 slack 0.916667 -12.3694 question-answer pairs 0.916667 -12.3694 subtasks, 0.916667 -12.3694 arg min 0.916667 -12.3694 (stoyanov 0.916667 -12.3694 14.0 0.857143 -12.3696 figure 9: 0.714286 -12.3697 as stated 0.681818 -12.3697 g., 0.916667 -12.3697 surprised 0.916667 -12.3697 context-aware 0.916667 -12.3697 48.4 0.916667 -12.37 calculation, 0.916667 -12.37 hand-aligned 0.580645 -12.37 further. 0.64 -12.3701 identities 0.666667 -12.3702 budget 0.666667 -12.3702 opinion expressions 0.736842 -12.3703 + l0 0.681818 -12.3703 information sources 0.478261 -12.3703 where, 0.916667 -12.3703 restore 0.916667 -12.3703 presumed 0.916667 -12.3703 (2008a), 0.62963 -12.3705 dorr, 0.209016 -12.3706 al., 2006). 0.681818 -12.3707 (lavie 0.8125 -12.3707 emit 0.165816 -12.3711 language processing 0.542857 -12.3711 ukwac 0.916667 -12.3714 86.2 0.916667 -12.3714 fuzzy-tree to 0.916667 -12.3714 50.3 0.916667 -12.3717 34.2 0.916667 -12.3717 greek, 0.916667 -12.3717 latest version 0.8125 -12.3717 likes 0.916667 -12.372 36.3 0.311927 -12.3722 authorship 0.916667 -12.3722 74.3 0.916667 -12.3722 intent, 0.736842 -12.3723 bos 0.540541 -12.3723 binary feature 0.289062 -12.3723 we conclude 0.607143 -12.3725 accent 0.916667 -12.3725 abstracted 0.916667 -12.3725 92.3 0.916667 -12.3725 opinion-bearing 0.916667 -12.3725 parties 0.681818 -12.3726 nuclear 0.45098 -12.3726 harder to 0.681818 -12.3728 svmlight 0.916667 -12.3728 section) 0.916667 -12.3728 70.9 0.916667 -12.3728 wmpf 0.666667 -12.3729 laboratory, 0.916667 -12.3731 best-scoring 0.916667 -12.3731 boundary, 0.916667 -12.3734 domain-dependent 0.916667 -12.3734 overfit 0.916667 -12.3737 exceeding 0.916667 -12.3739 76.1 0.916667 -12.3739 others’ 0.857143 -12.3741 ontology-based 0.916667 -12.3742 peerreview 0.241758 -12.3742 pp. 0.0371462 -12.3743 language 0.916667 -12.3745 regularizers 0.62963 -12.3748 annotated data, 0.666667 -12.3749 .01 0.714286 -12.375 øj 0.916667 -12.3751 45.0 0.916667 -12.3751 20%, 0.916667 -12.3751 shortly 0.391304 -12.3753 downloaded 0.314815 -12.3755 testing. 0.916667 -12.3756 slide 0.857143 -12.3757 royal 0.857143 -12.3762 non-english responses 0.916667 -12.3762 unspecified 0.916667 -12.3762 definitional 0.916667 -12.3762 gradient-based 0.681818 -12.3764 remote 0.62963 -12.3764 edge weight 0.916667 -12.3765 nodes) 0.916667 -12.3765 pairs; 0.916667 -12.3765 gaps, 0.916667 -12.3765 crf-d 0.916667 -12.3768 32.5 0.916667 -12.3768 50%, 0.916667 -12.3771 computer-assisted 0.916667 -12.3773 manually-crafted 0.558824 -12.3775 siblings 0.857143 -12.3776 au 0.666667 -12.3776 highest-scoring 0.916667 -12.3779 they, 0.916667 -12.3779 listener’s 0.916667 -12.3779 just-in-time 0.4 -12.378 restaurant 0.8125 -12.378 yan 0.857143 -12.3781 specific 0.270833 -12.3784 translation models 0.916667 -12.3785 weld, 0.916667 -12.3785 60.2 0.8125 -12.3787 mkcls 0.916667 -12.3788 library, 0.916667 -12.3788 51% 0.607143 -12.3789 lucy 0.916667 -12.379 worthy 0.916667 -12.379 contradict 0.916667 -12.379 popular, 0.916667 -12.379 79.2 0.468085 -12.3791 preparation 0.8125 -12.3792 59.1 0.916667 -12.3796 87.71 0.916667 -12.3796 emerged 0.299145 -12.3796 plays 0.8125 -12.3797 ou 0.916667 -12.3799 fragmented 0.916667 -12.3799 ruling 0.542857 -12.3802 bunescu 0.916667 -12.3802 resource-rich 0.916667 -12.3802 topic-related 0.916667 -12.3805 transitivity constraints 0.916667 -12.3805 machine-readable 0.916667 -12.3805 blunsom, 0.916667 -12.3805 2001, 0.916667 -12.3805 .10 0.62963 -12.3805 l-bfgs 0.8125 -12.3805 mexico 0.526316 -12.3808 were created 0.916667 -12.3808 prasad 0.916667 -12.3808 81.5 0.916667 -12.3808 ˆpintp 0.916667 -12.3808 (lee, 0.916667 -12.381 built upon 0.916667 -12.381 verifies 0.916667 -12.381 .66 0.916667 -12.381 extractions, 0.916667 -12.381 billions 0.916667 -12.381 kondrak, 0.8125 -12.381 25.6 0.666667 -12.3811 those who 0.857143 -12.3813 substitution grammars 0.8125 -12.3813 interacts 0.736842 -12.3814 30.5 0.23913 -12.3814 experiments with 0.62963 -12.3816 partial parse 0.512821 -12.3817 important aspects 0.916667 -12.3819 84.8 0.916667 -12.3819 classifier 0.714286 -12.3821 .95 0.64 -12.3821 new york, ny 0.857143 -12.3824 layered 0.916667 -12.3825 explicit, 0.916667 -12.3825 stuck 0.916667 -12.3825 strength, 0.916667 -12.3825 timeout 0.916667 -12.3827 focal 0.916667 -12.3827 (2008; 0.916667 -12.3827 aligner, 0.916667 -12.383 (ittycheriah 0.8125 -12.383 wr 0.8125 -12.383 sudden 0.916667 -12.3833 58.0 0.916667 -12.3833 (albeit 0.916667 -12.3833 space) 0.916667 -12.3833 resume 0.916667 -12.3833 treeto-string 0.8125 -12.3833 voted 0.666667 -12.3834 0.01, 0.8125 -12.3835 multinomial distributions 0.916667 -12.3836 extractive summaries 0.916667 -12.3836 1200 0.916667 -12.3836 .32 0.916667 -12.3839 predetermined 0.916667 -12.3839 76.5 0.857143 -12.3842 urls, 0.916667 -12.3842 intruder 0.64 -12.3845 xi, 0.916667 -12.3845 77.3 0.916667 -12.3845 (bird 0.916667 -12.3845 institution 0.8125 -12.3846 t-unit 0.857143 -12.3848 gnppa 0.916667 -12.385 queue, 0.916667 -12.385 vectors) 0.916667 -12.385 undergraduate 0.607143 -12.3851 defined. 0.764706 -12.3853 tim 0.916667 -12.3853 2009a) 0.916667 -12.3853 discern 0.916667 -12.3853 effects, 0.916667 -12.3856 parsimony 0.916667 -12.3856 topic) 0.916667 -12.3856 etc), 0.916667 -12.3856 signaling 0.8125 -12.3856 n’t 0.764706 -12.3857 despite their 0.857143 -12.3858 vmm 0.444444 -12.3858 latest 0.916667 -12.3859 sense) 0.764706 -12.386 hosts 0.764706 -12.386 ravichandran, 0.46 -12.3861 difficult. 0.681818 -12.3862 shortcomings of 0.916667 -12.3862 70.8 0.916667 -12.3862 87.4 0.736842 -12.3863 bojar 0.8125 -12.3863 fleiss’ 0.526316 -12.3865 italian, 0.64 -12.3865 ˆ 0.916667 -12.3865 cisp 0.681818 -12.3868 advance. 0.916667 -12.3868 timeconsuming 0.8125 -12.3868 conferences 0.488372 -12.387 our primary 0.916667 -12.387 wash 0.916667 -12.387 .58 0.916667 -12.3873 detector, 0.916667 -12.3873 neutral) 0.916667 -12.3873 human, 0.8125 -12.3873 at level 0.268966 -12.3876 as mentioned 0.0318702 -12.3876 word 0.916667 -12.3876 characters) 0.916667 -12.3876 63.2 0.916667 -12.3876 oldest 0.409836 -12.3877 we created a 0.736842 -12.3879 consumers 0.916667 -12.3879 cluster) 0.916667 -12.3879 false-positive 0.916667 -12.3879 members, 0.916667 -12.3882 94.0 0.916667 -12.3882 bigger than 0.916667 -12.3882 .68 0.64 -12.3883 each participant 0.607143 -12.3883 importance. 0.148536 -12.3884 the size of 0.681818 -12.3885 ravi 0.580645 -12.3887 for details. 0.916667 -12.3888 1999, 0.916667 -12.389 prescribed 0.916667 -12.389 deciding whether 0.64 -12.3895 2004), which 0.45098 -12.3895 algorithm 2 0.468085 -12.3896 reported results 0.916667 -12.3896 formulaic 0.409836 -12.3897 two methods 0.607143 -12.3899 in comparison, 0.916667 -12.3899 impoverished 0.857143 -12.3901 at present, 0.64 -12.3901 content-only 0.916667 -12.3902 cases) 0.916667 -12.3902 graph-partitioning 0.916667 -12.3902 non-literal 0.916667 -12.3902 s0, 0.190972 -12.3902 portland, oregon, june 19-24, 0.5 -12.3902 wan 0.916667 -12.3905 credibility 0.916667 -12.3905 guessed 0.916667 -12.3905 reasonable, 0.916667 -12.3905 (candito 0.736842 -12.3906 parsing complexity 0.857143 -12.3907 ensured that 0.916667 -12.3908 35.6 0.916667 -12.3908 baker 0.916667 -12.3911 texts; 0.916667 -12.3911 opinionfinder 0.916667 -12.3911 loses 0.916667 -12.3911 randomlow 0.916667 -12.3911 prototype-based 0.4 -12.3913 conversely, 0.916667 -12.3913 information-seeking 0.916667 -12.3913 advocated 0.916667 -12.3913 23.4 0.151515 -12.3914 start 0.8125 -12.3914 subsumption 0.764706 -12.3915 likelihoods 0.488372 -12.3915 type-based 0.916667 -12.3916 carpenter 0.916667 -12.3916 count-min 0.857143 -12.3917 verb: 0.681818 -12.3917 sag 0.8125 -12.3917 saldo 0.916667 -12.3919 situational 0.916667 -12.3919 cleaner 0.916667 -12.3919 allophones 0.916667 -12.3925 all-link 0.916667 -12.3925 parallelized 0.916667 -12.3925 locked 0.916667 -12.3925 reimplemented 0.120219 -12.3925 position 0.916667 -12.3928 graph’s 0.916667 -12.3928 .17 0.916667 -12.3928 intact 0.916667 -12.3928 27.7 0.8125 -12.3929 unseen words 0.916667 -12.3931 2000, 0.0744235 -12.3933 association 0.916667 -12.3934 extrapolated 0.916667 -12.3934 sentimental 0.580645 -12.3934 learners. 0.580645 -12.3936 seventh 0.916667 -12.3937 claiming 0.916667 -12.3937 .23 0.916667 -12.3937 submodularity 0.916667 -12.3937 38.3 0.916667 -12.3937 corectly 0.64 -12.3939 lda+hellinger 0.681818 -12.3939 transactions on 0.916667 -12.3939 transformation-based 0.736842 -12.3941 interleaved 0.681818 -12.3941 susan 0.916667 -12.3942 (2008a) 0.916667 -12.3942 summarization: 0.916667 -12.3942 signature, 0.916667 -12.3942 i(j) 0.916667 -12.3942 equivalently 0.857143 -12.3944 (crammer et 0.857143 -12.3944 nih 0.916667 -12.3945 bonferroni 0.916667 -12.3945 word-topic 0.916667 -12.3945 (t,h) 0.916667 -12.3945 94.8 0.916667 -12.3945 semi-automatically 0.764706 -12.3946 (gildea 0.416667 -12.3946 can still 0.40625 -12.3949 (n 0.8125 -12.395 well-studied 0.8125 -12.395 cite 0.764706 -12.3951 al$yx 0.916667 -12.3951 mapreduce 0.916667 -12.3951 portal 0.916667 -12.3951 84.1 0.916667 -12.3951 (baccianella 0.857143 -12.3953 for clarity, 0.736842 -12.3953 in our experiments, we use 0.916667 -12.3954 mentioned above. 0.916667 -12.3954 matter, 0.916667 -12.3954 congruent 0.916667 -12.3954 nissim’s feature 0.916667 -12.3957 footnote 0.916667 -12.3957 88.8 0.916667 -12.3957 logically 0.916667 -12.3957 mm+ 0.478261 -12.3958 holds for 0.409836 -12.3959 formulation of 0.247059 -12.3959 2011, 0.916667 -12.396 repetitive 0.916667 -12.396 compared, 0.916667 -12.396 boxed 0.212121 -12.3961 tags. 0.916667 -12.3963 schabes, 0.8125 -12.3963 1980). 0.681818 -12.3964 at: 0.916667 -12.3965 bypass 0.916667 -12.3965 imdb 0.857143 -12.3966 fa 0.681818 -12.3966 second experiment 0.336957 -12.3967 treebank, 0.916667 -12.3968 tag) 0.916667 -12.3971 254 0.916667 -12.3971 geo 0.916667 -12.3971 (syntactic 0.916667 -12.3971 concepts: 0.580645 -12.3973 locate 0.282443 -12.3974 vice 0.916667 -12.3974 gricean 0.916667 -12.3974 reproducing 0.764706 -12.3975 (denero 0.436364 -12.3975 higher accuracy 0.234375 -12.3975 we add 0.916667 -12.3977 periods, 0.916667 -12.3977 well: 0.916667 -12.3977 semantic relatedness. 0.916667 -12.3977 prlm-based 0.916667 -12.3977 essence 0.764706 -12.398 constraint: 0.916667 -12.398 complicate 0.916667 -12.398 l1-transfer 0.916667 -12.398 popularity, 0.916667 -12.398 convincing 0.916667 -12.398 cai 0.736842 -12.3981 decoding process 0.916667 -12.3983 2(a) 0.916667 -12.3983 wn, 0.916667 -12.3986 91.0 0.916667 -12.3986 statistical dialog 0.916667 -12.3986 exceeded 0.916667 -12.3986 compositionally 0.607143 -12.3987 [a 0.586207 -12.3988 include, 0.916667 -12.3989 munro 0.916667 -12.3989 reinforces 0.916667 -12.3989 threads, 0.916667 -12.3989 cbdf 0.916667 -12.3989 interestingly 0.18543 -12.399 removed 0.916667 -12.3991 10.0 0.478261 -12.3993 when dealing 0.916667 -12.3994 flp 0.916667 -12.3994 27.5 0.857143 -12.3996 (with no 0.857143 -12.3996 clips 0.217195 -12.3996 behind 0.764706 -12.3997 patwardhan 0.916667 -12.3997 (epi) 0.916667 -12.3997 30.7 0.580645 -12.3998 partially funded 0.764706 -12.3999 48.3 0.916667 -12.4 /d/ 0.916667 -12.4 altering 0.736842 -12.4004 large vocabulary 0.916667 -12.4006 stripped 0.916667 -12.4006 this year’s 0.916667 -12.4006 uh 0.916667 -12.4006 beliefs, 0.916667 -12.4009 may, 0.916667 -12.4009 depiction 0.136283 -12.4009 2010), 0.736842 -12.4011 often contain 0.857143 -12.4012 marginalize 0.857143 -12.4012 low accuracy 0.607143 -12.4012 syntactic constraints 0.916667 -12.4012 lexicography, 0.764706 -12.4014 gaining 0.314286 -12.4015 our work is 0.916667 -12.4015 synonymy, 0.916667 -12.4015 66.4 0.916667 -12.4015 reasoning, 0.916667 -12.4015 organizing 0.916667 -12.4015 consent 0.0595399 -12.4017 pairs 0.586207 -12.4018 is replaced with 0.916667 -12.4018 folowed 0.586207 -12.402 indicate whether 0.764706 -12.4021 .94 0.916667 -12.4021 adjusts 0.916667 -12.4021 memorize 0.527778 -12.4023 and roth 0.916667 -12.4023 .09 0.916667 -12.4023 acquires 0.245614 -12.4024 a strong 0.857143 -12.4026 ceo 0.468085 -12.4026 data structure 0.916667 -12.4026 markert, 0.916667 -12.4026 shaping 0.916667 -12.4026 identify, 0.916667 -12.4029 ber 0.764706 -12.4031 55.6 0.7 -12.4032 high proportion 0.916667 -12.4032 60.3 0.916667 -12.4032 giga-fren 0.916667 -12.4032 tags) 0.916667 -12.4032 existed 0.916667 -12.4032 emergent 0.586207 -12.4033 source language, 0.242938 -12.4033 alignment. 0.271429 -12.4035 in fig. 0.916667 -12.4035 273 0.916667 -12.4035 (green 0.916667 -12.4035 causality, 0.857143 -12.4037 space: 0.764706 -12.4038 68.1 0.7 -12.4038 validation data 0.7 -12.4038 thrown 0.096861 -12.4039 provided 0.916667 -12.4041 porting 0.916667 -12.4041 27.0 0.368421 -12.4041 weights, 0.764706 -12.4045 decoders, 0.290323 -12.4046 walk 0.916667 -12.4047 edits-ga 0.916667 -12.4047 2011b) 0.368421 -12.4048 can make 0.542857 -12.4049 librarian 0.857143 -12.405 jing 0.916667 -12.405 11.0 0.916667 -12.405 27.2 0.681818 -12.4053 explanatory 0.916667 -12.4053 passonneau 0.916667 -12.4053 music, 0.916667 -12.4053 aplications 0.916667 -12.4053 267 0.736842 -12.4056 bracehtipupright 0.916667 -12.4056 nonterminal, 0.916667 -12.4056 1989), 0.916667 -12.4058 theoretically, 0.916667 -12.4058 hand-labeled 0.916667 -12.4058 gre 0.542857 -12.406 reproduce 0.916667 -12.4061 (target) 0.916667 -12.4061 judgment, 0.916667 -12.4061 canada, 0.916667 -12.4061 aligned, 0.916667 -12.4061 gnetic 0.916667 -12.4064 classifier; 0.916667 -12.4064 0.4% 0.916667 -12.4064 user profile 0.488372 -12.4065 scenarios. 0.64 -12.4067 forms: 0.916667 -12.4067 corrections, 0.916667 -12.4067 structures: 0.5625 -12.4069 us. 0.916667 -12.407 12.1 0.201581 -12.4072 organized 0.736842 -12.4072 dd 0.916667 -12.4073 psycholinguistics 0.916667 -12.4073 analyzers 0.916667 -12.4073 9.5 0.916667 -12.4076 placeholder 0.5 -12.4078 simplification. 0.736842 -12.4079 por 0.916667 -12.4079 nominalization 0.916667 -12.4079 initializer 0.916667 -12.4079 (having 0.916667 -12.4079 associativity 0.916667 -12.4082 ion 0.916667 -12.4082 berlin, 0.586207 -12.4085 have recently 0.916667 -12.4085 yih 0.916667 -12.4085 decomposition, 0.7 -12.4087 representation: 0.7 -12.4087 relatum 0.7 -12.4087 1k 0.916667 -12.4088 urdu, 0.916667 -12.4088 81.9 0.916667 -12.4088 trending 0.7 -12.409 event extraction, 0.916667 -12.4091 acros 0.916667 -12.4091 granted 0.916667 -12.4091 1984; 0.916667 -12.4091 12.8 0.477273 -12.4093 container 0.916667 -12.4094 i+ 0.916667 -12.4094 subpart 0.736842 -12.4095 focussed 0.916667 -12.4096 zeman 0.916667 -12.4096 ¯t 0.736842 -12.4098 m-step 0.916667 -12.4099 pictures, 0.916667 -12.4099 49.7 0.615385 -12.41 transferring 0.764706 -12.4101 translations: 0.764706 -12.4101 planner 0.916667 -12.4102 mark-up 0.916667 -12.4105 detection/correction 0.380282 -12.4107 answers. 0.916667 -12.4108 worry 0.916667 -12.4108 secret 0.916667 -12.4108 glossary 0.916667 -12.4111 options: 0.916667 -12.4111 tractable, 0.916667 -12.4111 coder 0.916667 -12.4111 exponentially many 0.607143 -12.4112 3.3). 0.916667 -12.4114 1s 0.916667 -12.4114 swedish, 0.916667 -12.4114 toponym 0.45098 -12.4114 previous section 0.764706 -12.4116 .3 0.916667 -12.4117 trends, 0.916667 -12.4117 66.5 0.916667 -12.4117 91.5 0.916667 -12.412 humanitarian 0.916667 -12.412 joachims, 0.916667 -12.412 examined, 0.8 -12.412 oovs. 0.764706 -12.4121 cliques 0.736842 -12.4121 mean-field 0.736842 -12.4124 table 8. 0.285714 -12.4126 we calculate 0.615385 -12.4128 multi-task 0.652174 -12.4128 ryan 0.5625 -12.4132 surprising that 0.916667 -12.4132 5-point 0.916667 -12.4132 triplet 0.916667 -12.4135 synthesize 0.916667 -12.4135 (low 0.0580561 -12.4138 phrase 0.916667 -12.4138 (full 0.916667 -12.4138 basically, 0.916667 -12.4138 righthand 0.736842 -12.414 .76 0.916667 -12.4141 pp, 0.100195 -12.4141 focus 0.652174 -12.4141 shown, 0.586207 -12.4144 1989). 0.916667 -12.4144 semh 0.916667 -12.4144 subcorpora 0.736842 -12.4145 involvement 0.916667 -12.4146 cart 0.916667 -12.4146 activity, 0.916667 -12.4146 expresed 0.681818 -12.4147 textual information 0.204918 -12.4147 collect 0.258065 -12.4147 formed 0.442308 -12.4147 infix 0.764706 -12.4148 inter-coder 0.764706 -12.4148 f¨ur 0.916667 -12.4149 riezler 0.8 -12.4149 (hereafter 0.764706 -12.415 by hand. 0.8 -12.4151 mbrsc 0.219626 -12.4156 the training set 0.857143 -12.4157 seconds) 0.681818 -12.4158 roget 0.192029 -12.4158 a linear 0.916667 -12.4158 cipher, 0.764706 -12.416 28.1 0.681818 -12.416 turian et 0.916667 -12.4161 by chance. 0.916667 -12.4161 governors 0.916667 -12.4161 subject’s 0.764706 -12.4163 3for 0.916667 -12.4164 merger 0.916667 -12.4164 roark, 0.916667 -12.4164 dubey 0.158151 -12.4166 a similar 0.916667 -12.4167 s2, 0.916667 -12.4167 (ge) 0.916667 -12.4167 kit 0.545455 -12.4167 1990; 0.764706 -12.4168 pour 0.764706 -12.417 sch¨utze 0.916667 -12.417 nullnull, 0.64 -12.4174 bigram features 0.545455 -12.4174 file, 0.916667 -12.4176 trigger-argument 0.916667 -12.4176 efficient algorithms 0.857143 -12.4177 nullj 0.916667 -12.4179 rob 0.916667 -12.4179 plural, 0.916667 -12.4179 algorithm’s 0.916667 -12.4179 ontext 0.916667 -12.4179 increments 0.329787 -12.4181 are currently 0.916667 -12.4182 generative probabilistic model 0.916667 -12.4182 dimension-specific 0.916667 -12.4182 26.4 0.512821 -12.4182 bottleneck 0.64 -12.4184 principal 0.857143 -12.4185 88.9 0.736842 -12.4185 after adding 0.916667 -12.4185 interactions among 0.916667 -12.4185 worker’s 0.409836 -12.419 detail. 0.916667 -12.4191 too long 0.916667 -12.4191 century, 0.916667 -12.4191 interval, 0.764706 -12.4192 “john 0.7 -12.4193 summit 0.615385 -12.4193 we can observe 0.916667 -12.4194 burstein, 0.916667 -12.4194 farther 0.764706 -12.4195 three steps 0.64 -12.4195 evaluations on 0.477273 -12.4197 the lrscore 0.916667 -12.4197 one-vs-all 0.0723751 -12.4199 out 0.764706 -12.42 disgust, 0.764706 -12.42 marsi 0.283465 -12.42 concepts. 0.8 -12.42 (yu 0.857143 -12.4201 thus allowing 0.916667 -12.4203 19.2 0.916667 -12.4203 establishment 0.916667 -12.4206 comlex 0.916667 -12.4206 3a 0.413793 -12.4209 λ = 0.916667 -12.4209 1988) 0.916667 -12.4209 .11 0.7 -12.4211 translation pairs. 0.916667 -12.4211 rewrites 0.5625 -12.4216 seemed to 0.916667 -12.4217 main-clause 0.916667 -12.4217 ssr 0.916667 -12.4217 flows 0.26 -12.4218 summaries. 0.916667 -12.422 max-margin 0.615385 -12.4221 0.18 0.7 -12.4223 machine-generated 0.916667 -12.4223 qpdg 0.916667 -12.4223 investors 0.764706 -12.4225 (gao et 0.527778 -12.4225 technologies, 0.477273 -12.4226 r3 0.916667 -12.4226 answer: 0.916667 -12.4226 fork 0.857143 -12.4229 strength. 0.916667 -12.4229 (ds) 0.916667 -12.4229 hands 0.916667 -12.4229 newton’s 0.143434 -12.4231 assumption 0.44898 -12.4233 it consists of 0.916667 -12.4235 orca 0.916667 -12.4235 1982; 0.764706 -12.4239 dry 0.409836 -12.424 an annotated 0.189474 -12.4244 validation 0.8 -12.4244 66.9 0.488372 -12.4245 to eliminate 0.652174 -12.4246 to our knowledge, this 0.527778 -12.4247 7 conclusion we 0.916667 -12.4247 check, 0.916667 -12.4247 86.9 0.8 -12.4247 stil 0.369863 -12.4248 responses. 0.916667 -12.425 (computed 0.916667 -12.425 like, 0.916667 -12.4253 token-role 0.916667 -12.4256 once per 0.916667 -12.4256 princeton, 0.167123 -12.4258 problem, 0.916667 -12.4262 ceiling 0.916667 -12.4262 editors, 0.916667 -12.4262 founded 0.916667 -12.4262 simcoach 0.265734 -12.4265 was trained 0.916667 -12.4265 syntax-augmented 0.916667 -12.4265 back-transliteration 0.7 -12.4266 ellen 0.652174 -12.4267 gaps. 0.7 -12.4268 sentence ordering 0.916667 -12.4268 em’s 0.916667 -12.4268 kmeans 0.545455 -12.4269 k) 0.916667 -12.4274 julia 0.916667 -12.428 thematically 0.916667 -12.428 440 0.916667 -12.428 aplying 0.15011 -12.4281 level of 0.916667 -12.4283 ram 0.916667 -12.4283 author-topic 0.916667 -12.4283 lemmatised 0.916667 -12.4283 “correct” 0.916667 -12.4283 390 0.8 -12.4283 qualify 0.7 -12.4286 relatively large 0.916667 -12.4286 iterations) 0.8 -12.4286 outlook 0.652174 -12.4286 in this context, 0.5 -12.4287 we write 0.8 -12.4288 long-form 0.916667 -12.4289 places, 0.916667 -12.4289 smt-based 0.513514 -12.4291 denero 0.319588 -12.4291 pointwise 0.8 -12.4291 contextual features, 0.916667 -12.4292 conducts 0.545455 -12.4294 3.2.1 0.916667 -12.4295 asserts 0.916667 -12.4295 learn, 0.916667 -12.4298 dated 0.916667 -12.4301 context; 0.916667 -12.4301 epochs 0.916667 -12.4301 (brill 0.7 -12.4302 thompson 0.8 -12.4302 surpass 0.916667 -12.4304 attachment, 0.916667 -12.4304 discovery, 0.916667 -12.4304 rounded 0.764706 -12.4307 wc 0.615385 -12.4307 lafferty, 0.916667 -12.4307 thinks 0.487805 -12.4307 quotes 0.615385 -12.4309 was implemented 0.298246 -12.4309 we denote 0.185185 -12.431 half 0.916667 -12.431 2.1.1 0.652174 -12.4312 coerced 0.916667 -12.4313 (kl) 0.916667 -12.4316 center, 0.764706 -12.4317 sandra 0.35443 -12.4317 sec. 0.8 -12.4317 aac users 0.7 -12.4318 rk 0.566667 -12.4318 derivative 0.916667 -12.4319 mutations 0.916667 -12.4322 mirrors 0.764706 -12.4324 mixture model, 0.8 -12.4325 (online) 0.413793 -12.4327 comprised 0.276923 -12.4329 selection. 0.8 -12.433 authoritativeness 0.349398 -12.433 can achieve 0.916667 -12.4331 sumbasic 0.8 -12.4333 74.8 0.5 -12.4333 jdpa 0.8 -12.4336 83.5 0.566667 -12.4337 ck 0.764706 -12.4339 corner 0.7 -12.4339 deictic 0.7 -12.4341 “she 0.916667 -12.4343 5.2.3 0.764706 -12.4344 desktop 0.487805 -12.4347 mentioned, 0.421053 -12.4348 dep 0.916667 -12.4349 co-occurrence statistics 0.916667 -12.4349 20.1 0.375 -12.435 new features 0.916667 -12.4352 television 0.764706 -12.4354 packages 0.105611 -12.4354 1, 0.8 -12.4354 77.9 0.7 -12.4355 in these cases, 0.916667 -12.4355 cn-lvcs 0.916667 -12.4358 negator 0.846154 -12.4361 p(x) 0.586207 -12.4361 locations. 0.376812 -12.4361 emotion words 0.8 -12.4362 stone 0.349398 -12.4362 applied. 0.846154 -12.4364 (high 0.181518 -12.4365 comparison of 0.5 -12.4366 reconstruct 0.846154 -12.4367 prototype, 0.846154 -12.4367 watermark 0.916667 -12.4367 slightly, 0.916667 -12.4367 genic 0.916667 -12.4367 purchase 0.232432 -12.4368 articles. 0.764706 -12.4369 katz 0.846154 -12.4372 0.916667 -12.4373 (ignoring 0.916667 -12.4373 frame-based 0.466667 -12.4373 open-source 0.846154 -12.4375 i; 0.566667 -12.4376 question-answer 0.916667 -12.4376 multi-instance 0.916667 -12.4376 marine 0.916667 -12.4379 expandrank 0.586207 -12.438 helper 0.158025 -12.438 supported by 0.35443 -12.438 selected by 0.8 -12.438 converter 0.586207 -12.4382 these algorithms 0.916667 -12.4382 single-document 0.916667 -12.4382 microphone 0.846154 -12.4383 las uas 0.846154 -12.4383 (bnc) 0.846154 -12.4383 figures, 0.38806 -12.4384 an svm 0.8 -12.4385 receiver 0.916667 -12.4388 complex predicates. 0.916667 -12.4388 were, 0.256579 -12.4388 kim 0.916667 -12.4391 sociolinguistics 0.5 -12.4391 post. 0.652174 -12.4393 (kudo 0.916667 -12.4394 xu, 0.916667 -12.4394 mprime 0.916667 -12.4394 learner, 0.223881 -12.4394 think 0.846154 -12.4397 we’ve 0.846154 -12.4397 klsh 0.846154 -12.4397 brief, 0.846154 -12.4397 domain-adaptation 0.846154 -12.44 ranking: 0.916667 -12.44 31.8 0.8 -12.4401 79.5 0.846154 -12.4403 ets 0.916667 -12.4403 (2010): 0.916667 -12.4409 metropolis-hastings 0.8 -12.4409 92.8 0.652174 -12.441 hay 0.846154 -12.4411 worst case 0.916667 -12.4412 restart 0.8 -12.4412 local, 0.8 -12.4414 non-projective parsing 0.31068 -12.4415 harder 0.916667 -12.4415 4.9 0.615385 -12.4416 developments 0.916667 -12.4418 user-defined 0.846154 -12.442 dir 0.8 -12.442 63.6 0.916667 -12.4421 (results 0.333333 -12.4421 difficulty of 0.652174 -12.4421 agnostic 0.425926 -12.4424 qg 0.916667 -12.4424 233 0.846154 -12.4425 transliteration pairs. 0.846154 -12.4425 e.g.: 0.8 -12.4425 60.7 0.916667 -12.4427 sentiments, 0.545455 -12.4427 bionlp-st 0.846154 -12.4428 301 0.846154 -12.4431 orf 0.7 -12.4431 (chang et 0.8 -12.4433 generalize well 0.846154 -12.4436 hole 0.545455 -12.4436 identical. 0.545455 -12.4436 controlling 0.8 -12.4436 amsterdam 0.8 -12.4436 80.2 0.527778 -12.4437 will explore 0.290598 -12.4438 t1 0.7 -12.444 an ilp 0.722222 -12.4443 .90 0.846154 -12.4445 technological 0.396825 -12.4445 jst 0.846154 -12.4447 toefl 0.7 -12.4447 labelled with 0.117241 -12.4448 allow 0.846154 -12.445 λ-meaning 0.40678 -12.4451 clusters are 0.846154 -12.4453 dominance 0.846154 -12.4453 assisted 0.513514 -12.4453 owl 0.433962 -12.4453 decides 0.529412 -12.4454 an mwe 0.8 -12.4454 63.3 0.35 -12.4457 phase, 0.0818243 -12.4458 applied 0.263889 -12.4459 summarization. 0.8 -12.4459 150k 0.764706 -12.446 terrorism 0.282258 -12.446 co-occur 0.846154 -12.4462 ssh 0.846154 -12.4462 rn 0.722222 -12.4462 70.5 0.8 -12.4462 automotive 0.764706 -12.4465 φ, 0.513514 -12.4466 exported 0.846154 -12.4467 call, 0.8 -12.4467 grams 0.652174 -12.4467 an open-source 0.313725 -12.4468 we combine 0.15869 -12.4469 2008) 0.846154 -12.447 could, 0.266187 -12.447 treebank. 0.846154 -12.4473 naively 0.764706 -12.4473 pij 0.7 -12.4473 following features: 0.586207 -12.4473 to enrich 0.8 -12.4475 formalisation 0.7 -12.4477 sadness, 0.586207 -12.4477 acknowledgement 0.846154 -12.4481 margaret 0.545455 -12.4481 pos information 0.151724 -12.4484 sum 0.337209 -12.4485 outlined 0.615385 -12.4488 questions: 0.433962 -12.4489 † 0.846154 -12.449 web interface 0.846154 -12.449 twice, 0.566667 -12.449 prove that 0.456522 -12.449 but instead 0.846154 -12.4492 tpr 0.764706 -12.4493 relevance score 0.304762 -12.4493 feedback. 0.218447 -12.45 section 5 0.846154 -12.4501 rbm 0.764706 -12.4501 rl-based 0.592593 -12.4505 zones 0.592593 -12.4505 syntactic structures 0.764706 -12.4506 50%. 0.846154 -12.4507 (discussed 0.846154 -12.4507 ily 0.846154 -12.4507 1000, 0.846154 -12.4509 plugin 1 -12.451 non-terminals. 1 -12.451 permutation. 1 -12.451 experimentation. 1 -12.451 full-fledged 1 -12.451 standard deviation. 1 -12.451 filters. 1 -12.451 indexing. 1 -12.451 oregon health 1 -12.451 cards 1 -12.451 80%. 1 -12.451 somewhat. 1 -12.451 bound. 1 -12.451 achieve. 1 -12.451 load. 1 -12.451 ´o s´eaghdha 1 -12.451 correspondences. 1 -12.451 difficulties, 1 -12.451 intuitions. 1 -12.451 novel. 1 -12.451 modularization 1 -12.451 views. 1 -12.451 hwee tou 1 -12.451 empirically. 1 -12.451 block. 1 -12.451 (201). 1 -12.451 1984). 1 -12.451 pres. 1 -12.451 60%. 1 -12.451 (björne 1 -12.451 analogously. 1 -12.451 work). 1 -12.451 svms. 1 -12.451 relation). 1 -12.451 transducers. 1 -12.451 blog. 1 -12.451 (+0.3) 1 -12.451 evidences. 1 -12.451 (bottom). 1 -12.451 ned. 1 -12.451 go. 1 -12.451 profiles. 1 -12.451 compatible. 1 -12.451 treatment. 1 -12.451 scopes. 1 -12.451 represent. 1 -12.451 level). 1 -12.451 between. 1 -12.451 move. 1 -12.451 tutor, 1 -12.451 tracks. 1 -12.451 davidov 1 -12.451 deutsche forschungsgemeinschaft 1 -12.451 sprime def⇐⇒ 1 -12.451 %). 1 -12.451 b). 1 -12.451 distinctions. 1 -12.451 york. 1 -12.451 1972). 1 -12.451 .96 1 -12.451 assumed. 1 -12.451 boldface. 1 -12.451 assisting 1 -12.451 treetagger (schmid, 1994) 1 -12.451 .91 1 -12.451 recorded. 1 -12.451 %. 1 -12.451 server. 1 -12.451 computations. 1 -12.451 twice. 1 -12.451 questionnaire. 1 -12.451 varied. 1 -12.451 sortal restriction 1 -12.451 benefit. 1 -12.451 cm. 1 -12.451 else. 1 -12.451 expectation. 1 -12.451 casting 1 -12.451 0.85. 1 -12.451 25. 1 -12.451 power law 1 -12.451 normalisation. 1 -12.451 possibilities, 1 -12.451 .84 1 -12.451 themes. 1 -12.451 turkers. 1 -12.451 carried out, 1 -12.451 agencies. 1 -12.451 500k 1 -12.451 (?). 1 -12.451 exceptions. 1 -12.451 appropriately. 1 -12.451 cardie. 1 -12.451 subtask. 1 -12.451 〈〈q,t,precedesequal,ap〉,int,imp,exp〉 1 -12.451 dan roth 1 -12.451 17. 1 -12.451 activities. 1 -12.451 predictors. 1 -12.451 formality. 1 -12.451 personalization. 1 -12.451 edits. 1 -12.451 music. 1 -12.451 dependents. 1 -12.451 disabilities. 1 -12.451 underlined. 1 -12.451 companies. 1 -12.451 macaon exchange 1 -12.451 0.9. 1 -12.451 weighted. 1 -12.451 highest. 1 -12.451 num. 1 -12.451 sent. 1 -12.451 (nlp). 1 -12.451 generalisations 1 -12.451 limits. 1 -12.451 segmenter. 1 -12.451 abbreviations. 1 -12.451 discounting. 1 -12.451 omitted. 1 -12.451 such. 1 -12.451 usages. 1 -12.451 signatures. 1 -12.451 100%. 1 -12.451 based. 1 -12.451 uk. 1 -12.451 subtypes. 1 -12.451 sentence-level. 1 -12.451 annotation) 1 -12.451 provide. 1 -12.451 scrolling 1 -12.451 entities). 1 -12.451 texts). 1 -12.451 coefficients. 1 -12.451 p(c/¬a) ≡ 1 -12.451 light. 1 -12.451 starting point. 1 -12.451 show. 1 -12.451 syllables. 1 -12.451 remain. 1 -12.451 worst. 1 -12.451 aa. 1 -12.451 partitions. 1 -12.451 9). 1 -12.451 korean. 1 -12.451 approximations. 1 -12.451 kain 1 -12.451 201). 1 -12.451 1995. 1 -12.451 device. 1 -12.451 bootstraps 1 -12.451 rst. 1 -12.451 ρ. 1 -12.451 dep. 1 -12.451 graphics. 1 -12.451 11). 1 -12.451 decades, 1 -12.451 building. 1 -12.451 glosses. 1 -12.451 previously. 1 -12.451 unavailable. 1 -12.451 organizations. 1 -12.451 (dredze 1 -12.451 compositionality (disco’2011), 0.615385 -12.451 prevented 0.846154 -12.4512 (100 0.7 -12.4512 generality, 0.233333 -12.4512 s, 0.909091 -12.4513 dramatically. 0.909091 -12.4513 markables. 0.909091 -12.4513 206). 0.909091 -12.4513 advantages. 0.909091 -12.4513 discovered. 0.909091 -12.4513 post hoc 0.909091 -12.4513 stock price 0.909091 -12.4513 anaphora, 0.8 -12.4513 contextfree 0.846154 -12.4515 captions 0.846154 -12.4515 lt-hmm 0.846154 -12.4515 is empty. 0.909091 -12.4516 benign hypertensive 0.909091 -12.4516 closely tied 0.909091 -12.4516 watson research center 0.846154 -12.4518 tesla-b 0.44 -12.4518 ball 0.566667 -12.4518 mbdp-1 0.909091 -12.4519 verbs). 0.909091 -12.4519 downside 0.909091 -12.4519 common subspace 0.909091 -12.4519 slot fillers 0.529412 -12.4522 second part 0.8 -12.4523 coffee 0.592593 -12.4523 effects. 0.35 -12.4524 is described in 0.909091 -12.4525 majority voting 0.909091 -12.4525 ek – 0.846154 -12.4526 others” 0.529412 -12.4526 nice 0.764706 -12.4529 npk, 0.909091 -12.4529 world cup 0.371429 -12.453 research projects 0.8 -12.4531 86.0 0.846154 -12.4532 ternary 0.40678 -12.4532 suited for 0.5 -12.4533 opinion lexicon 0.487805 -12.4534 event trigger 0.228723 -12.4536 0.8 0.666667 -12.4537 cold 0.909091 -12.4538 finnish, 0.909091 -12.4538 legitimate 0.909091 -12.4544 91.3 0.909091 -12.4544 nullnullnullnullnullnullnullnullnullnullnull 0.909091 -12.4544 nq 0.909091 -12.4544 false alarms 0.909091 -12.4547 user interests 0.270677 -12.4547 trained using 0.3 -12.4548 5). 0.846154 -12.4549 down, 0.909091 -12.455 binary codes 0.909091 -12.455 narrative schema 0.909091 -12.455 non-hierarchical 0.909091 -12.455 phrasal deletion 0.764706 -12.4552 india. 0.764706 -12.4552 statistical parsers 0.909091 -12.4553 sequence labelling 0.7 -12.4554 73.5 0.846154 -12.4555 (+ 0.35 -12.4555 for training, 0.8 -12.4558 239 0.909091 -12.4559 (i.e 0.909091 -12.4559 newman 0.909091 -12.4559 decomposition (svd) 0.413793 -12.456 2010b). 0.44 -12.4562 discussion. 0.846154 -12.4563 case), 0.909091 -12.4565 .13 0.44 -12.4568 pos-tag 0.44898 -12.457 sizes. 0.666667 -12.4571 cherry 0.909091 -12.4571 article’s 0.909091 -12.4571 disambiguates 0.909091 -12.4571 compensated 0.909091 -12.4571 3.2), 0.8 -12.4572 acid 0.8 -12.4572 life, 0.846154 -12.4574 strike 0.909091 -12.4574 bio-crf 0.8 -12.4574 31.5 0.190476 -12.4576 additionally, 0.909091 -12.4577 gec 0.909091 -12.4577 popescu-belis, 0.909091 -12.4577 randlm 0.487805 -12.4577 assumption. 0.487805 -12.4577 medlda 0.8 -12.4577 mbot 0.0491126 -12.4578 task 0.909091 -12.458 semi-automated 0.909091 -12.458 suit 0.625 -12.4581 identically 0.625 -12.4581 recurrent 0.909091 -12.4586 correlates well 0.909091 -12.4586 contradicting 0.909091 -12.4586 head-dependents relations 0.8 -12.4588 288 0.909091 -12.4592 unsatisfactory 0.909091 -12.4592 out-ofdomain 0.8 -12.4593 244 0.846154 -12.4594 categorised 0.153477 -12.4595 translation, 0.35 -12.4595 malt 0.909091 -12.4596 candito 0.909091 -12.4596 topic-based 0.8 -12.4596 answer questions 0.8 -12.4596 topic-specific 0.8 -12.4596 xy 0.625 -12.4598 extraction techniques 0.909091 -12.4599 browse 0.909091 -12.4599 (graesser 0.248408 -12.4601 tokens. 0.290598 -12.4601 section 6 0.909091 -12.4602 microblogs, 0.319588 -12.4602 attribution 0.233333 -12.4604 the result of 0.513514 -12.4605 implemented using 0.722222 -12.4607 maximum length 0.909091 -12.4608 61.0 0.909091 -12.4608 5gram 0.909091 -12.4608 part-of-speech tagging. 0.666667 -12.4611 final output 0.7 -12.4611 drew 0.909091 -12.4611 exact, 0.909091 -12.4611 0.391 0.909091 -12.4611 imagenet 0.909091 -12.4611 drives 0.909091 -12.4611 compounding, 0.909091 -12.4611 argument-composition 0.909091 -12.4611 contents, 0.846154 -12.4614 basedon 0.909091 -12.4614 prof. 0.909091 -12.4614 31.0 0.909091 -12.4614 80.9 0.217391 -12.4615 engine 0.236994 -12.4615 parameters. 0.8 -12.4615 45.8 0.722222 -12.4616 timothy 0.846154 -12.4617 by normalizing 0.909091 -12.4617 ponzetto, 0.909091 -12.4617 93.4 0.130208 -12.4618 required 0.846154 -12.462 (average 0.909091 -12.462 rudnicky, 0.147577 -12.4622 the quality of 0.846154 -12.4623 resulted from 0.909091 -12.4623 saves 0.909091 -12.4623 björne 0.652174 -12.4624 c&c parser 0.909091 -12.4626 attributive 0.47619 -12.4627 boxshadowdwn 0.846154 -12.4628 initiate 0.666667 -12.4629 head nouns 0.909091 -12.4629 musical 0.846154 -12.4631 subject span 0.142857 -12.4631 act 0.722222 -12.4631 lends 0.8 -12.4633 100k 0.8 -12.4633 traditional summarization 0.846154 -12.4634 automatic acquisition 0.846154 -12.4634 82.3 0.909091 -12.4635 .65 0.909091 -12.4635 fragment pair 0.909091 -12.4635 complementarity 0.909091 -12.4635 cancel 0.846154 -12.4637 an agenda 0.846154 -12.4637 clicking on 0.846154 -12.464 (say 0.909091 -12.4641 dialogues collected 0.909091 -12.4641 testbed 0.8 -12.4644 ccm 0.615385 -12.4645 fy 0.846154 -12.4646 linguists, 0.846154 -12.4648 (2): 0.909091 -12.4648 chiang (2007), 0.909091 -12.4648 45.1 0.909091 -12.4648 them) 0.909091 -12.4648 computations, 0.909091 -12.4648 rich morphology 0.47619 -12.4649 lexical knowledge 0.666667 -12.465 pulman, 0.18 -12.465 this method 0.846154 -12.4651 redefine 0.846154 -12.4651 underscore 0.909091 -12.4651 o’connor 0.418182 -12.4653 derivation. 0.909091 -12.4654 antonymic 0.234637 -12.4654 lapata 0.566667 -12.4657 research institute 0.909091 -12.4657 ms, 0.909091 -12.4657 best) 0.909091 -12.466 muc-6 0.0382213 -12.4661 were 0.129983 -12.4662 4.1 0.410714 -12.4662 topic distribution 0.846154 -12.4663 signifies 0.722222 -12.4663 1] 0.846154 -12.4666 (v 0.846154 -12.4666 misspelling 0.909091 -12.4666 42.6 0.8 -12.4666 1b 0.846154 -12.4669 facets 0.909091 -12.4669 badly 0.909091 -12.4669 segmented, 0.909091 -12.4669 63.9 0.909091 -12.4669 habash, 0.466667 -12.4671 precision at 0.7 -12.4672 inquirer 0.529412 -12.4672 academy 0.909091 -12.4672 trimming 0.909091 -12.4672 .41 0.909091 -12.4672 decision-making 0.909091 -12.4672 95.4 0.8 -12.4674 taiwan 0.909091 -12.4675 trillion 0.909091 -12.4675 minimum risk 0.909091 -12.4675 enhances 0.909091 -12.4675 readability, 0.909091 -12.4675 2006); 0.909091 -12.4675 intra-annotator 0.722222 -12.4675 i2b2/va 0.909091 -12.4678 (denis 0.666667 -12.4681 ocr 0.909091 -12.4681 suspected 0.909091 -12.4685 nigam, 0.722222 -12.4687 ibm-2 0.909091 -12.4688 underperforms 0.8 -12.469 triples, 0.846154 -12.4691 technologies: 0.909091 -12.4691 fare 0.909091 -12.4691 sensibility 0.909091 -12.4691 pivot-based 0.0476411 -12.4693 performance 0.846154 -12.4694 damage 0.909091 -12.4694 andrew, 0.909091 -12.4694 unifies 0.909091 -12.4694 volunteer 0.909091 -12.4694 1964; 0.209091 -12.4696 available. 0.846154 -12.4697 make, 0.909091 -12.4697 b-best 0.909091 -12.4697 (bentivogli 0.909091 -12.4697 affixes, 0.909091 -12.4697 satisfactory, 0.909091 -12.47 scholarly 0.909091 -12.47 35.7 0.909091 -12.47 forwardbackward 0.909091 -12.47 61.4 0.7 -12.4701 they reported 0.909091 -12.4703 depth-first 0.909091 -12.4703 generalisation 0.418182 -12.4703 head, 0.909091 -12.4706 .08 0.909091 -12.4706  0.722222 -12.4707 metric scores 0.846154 -12.4709 17th 0.141104 -12.4709 it has 0.513514 -12.4711 (first 0.456522 -12.4711 shen 0.909091 -12.4712 wordforms 0.909091 -12.4712 87.5 0.909091 -12.4712 examples) 0.909091 -12.4712 34.5 0.909091 -12.4712 invariant 0.8 -12.4712 w5 0.7 -12.4713 this paper presents an 0.7 -12.4713 orig. 0.846154 -12.4714 (rows 0.846154 -12.4714 by placing 0.47619 -12.4714 apparent 0.909091 -12.4715 oracles 0.909091 -12.4715 extractor, 0.909091 -12.4715 tumor 0.456522 -12.4717 proc. 0.909091 -12.4719 them; 0.909091 -12.4719 reactions, 0.0237163 -12.4724 our 0.548387 -12.4724 system description 0.909091 -12.4725 text), 0.846154 -12.4729 m=1 0.722222 -12.4729 visualizing 0.909091 -12.4731 non-adjacent 0.909091 -12.4731 language), 0.909091 -12.4731 headlines, 0.722222 -12.4731 scanning, 0.182759 -12.4734 discussed in 0.909091 -12.4734 29.0 0.909091 -12.4734 quantization 0.909091 -12.4734 plsa, 0.722222 -12.4734 quasi-synchronous 0.167147 -12.4737 processing, 0.909091 -12.4737 difficulty, 0.909091 -12.4737 36.0 0.909091 -12.4737 distributed, 0.8 -12.4737 research direction 0.909091 -12.474 supplying 0.909091 -12.474 commonalities 0.4 -12.4741 between them 0.846154 -12.4743 bold) 0.909091 -12.4743 population, 0.909091 -12.4743 tetreault, 0.909091 -12.4743 20.0 0.846154 -12.4746 entertainment 0.909091 -12.4746 thesauri 0.909091 -12.4746 germanc 0.909091 -12.4746 6.2% 0.909091 -12.4746 kos 0.846154 -12.4749 (make 0.566667 -12.475 achieve better 0.909091 -12.475 34.1 0.909091 -12.475 outcome, 0.8 -12.475 swedish. 0.8 -12.475 language; 0.625 -12.4751 roukos, 0.36 -12.4752 parses. 0.107059 -12.4753 named 0.625 -12.4753 2008b; 0.909091 -12.4753 data sparsity, 0.47619 -12.4754 pan 0.625 -12.4755 spirit of 0.909091 -12.4756 stacked learning 0.909091 -12.4756 -0.01 0.909091 -12.4756 #2 0.722222 -12.4758 rozovskaya 0.909091 -12.4759 suppressed 0.446809 -12.4759 of all, 0.846154 -12.4761 genomics 0.666667 -12.4761 marcus, 0.355263 -12.4762 histograms 0.909091 -12.4762 agencies 0.909091 -12.4762 asimilation 0.8 -12.4764 peace 0.909091 -12.4765 learnable 0.909091 -12.4765 flip 0.909091 -12.4765 (other 0.909091 -12.4765 shuffled 0.909091 -12.4765 29.6 0.909091 -12.4765 impact, 0.909091 -12.4765 80.8 0.722222 -12.4766 formula. 0.909091 -12.4768 92.2 0.909091 -12.4768 progression 0.909091 -12.4768 42.4 0.909091 -12.4768 rosti 0.846154 -12.4769 formats, 0.548387 -12.4769 information extraction. 0.909091 -12.4771 29.4 0.909091 -12.4771 inversely 0.909091 -12.4771 82.7 0.722222 -12.4771 interspeech 0.8 -12.4772 α0 0.75 -12.4774 t-nl 0.909091 -12.4774 wizarded 0.909091 -12.4777 55.5 0.909091 -12.4777 18.2 0.909091 -12.4777 compositions 0.909091 -12.4777 comprehender’s 0.909091 -12.4784 manifested 0.909091 -12.4784 42.5 0.909091 -12.4784 44.7 0.909091 -12.4784 fluid 0.722222 -12.4786 terry 0.177049 -12.4786 above, 0.909091 -12.4787 speaker/writer 0.909091 -12.4787 spanish) 0.909091 -12.4787 48.8 0.722222 -12.4788 glow 0.36 -12.4789 shall 0.25 -12.4789 table 4. 0.909091 -12.479 pointer 0.909091 -12.479 sbmt 0.909091 -12.479 in–domain 0.909091 -12.479 tree-tostring 0.909091 -12.4793 8we 0.909091 -12.4793 mitigated 0.909091 -12.4793 trailing 0.909091 -12.4793 24.7 0.909091 -12.4796 context), 0.909091 -12.4796 adversely 0.909091 -12.4796 ascent 0.909091 -12.4796 translation; 0.8 -12.4797 dutch. 0.909091 -12.4799 training/test 0.909091 -12.4799 tier-2 0.909091 -12.4799 0.720 0.909091 -12.4799 (roth 0.722222 -12.48 tk 0.666667 -12.4802 various aspects 0.909091 -12.4802 achieved, 0.909091 -12.4802 variable-free 0.909091 -12.4802 slash 0.909091 -12.4802 detection module 0.8 -12.4802 modifier, 0.8 -12.4802 74.6 0.8 -12.4802 system combination task 0.666667 -12.4806 lexical semantics 0.666667 -12.4806 261 0.909091 -12.4806 60.9 0.909091 -12.4806 odd 0.909091 -12.4806 dunlop 0.909091 -12.4806 (pp) 0.909091 -12.4806 relations; 0.722222 -12.4808 89.4 0.0961538 -12.4809 computed 0.909091 -12.4809 antecedents, 0.909091 -12.4809 knowledge sources. 0.909091 -12.4809 30.6 0.909091 -12.4809 error-free 0.666667 -12.4811 gao et 0.566667 -12.4812 i want 0.371429 -12.4812 it uses 0.909091 -12.4812 81.0 0.909091 -12.4812 paradigmatic 0.846154 -12.4813 edu 0.846154 -12.4813 gt 0.8 -12.4813 phases: 0.514286 -12.4814 ’6 0.909091 -12.4815 resides 0.909091 -12.4818 null-calculus 0.909091 -12.4818 fit, 0.909091 -12.4818 risk, 0.846154 -12.4819 zk 0.846154 -12.4819 spearman correlation 0.846154 -12.4819 update semantics 0.514286 -12.4821 (called 0.625 -12.4821 pass, 0.909091 -12.4821 14.4 0.909091 -12.4821 2009)) 0.909091 -12.4821 sharing, 0.909091 -12.4821 default settings. 0.909091 -12.4821 form) 0.666667 -12.4822 subcorpus 0.722222 -12.4823 similarity functions 0.846154 -12.4824 info 0.566667 -12.4824 quickly. 0.909091 -12.4824 27.1 0.75 -12.4826 202 0.846154 -12.4827 ads data 0.909091 -12.4827 hk10 0.909091 -12.4831 inverted, 0.846154 -12.4833 think about 0.75 -12.4833 imperatives 0.909091 -12.4834 subordination 0.909091 -12.4834 maxine 0.909091 -12.4834 chai 0.909091 -12.4834 label) 0.909091 -12.4834 yule’s 0.909091 -12.4834 vc 0.909091 -12.4837 provenance 0.909091 -12.4837 load, 0.909091 -12.4837 denkowski 0.909091 -12.4837 originally designed 0.846154 -12.4839 semantic similarity measures 0.909091 -12.484 languages), 0.909091 -12.484 agirre et al. 0.909091 -12.484 1967) 0.909091 -12.484 elections 0.514286 -12.4841 candidate generation 0.571429 -12.4841 woz-08 0.909091 -12.4843 46.3 0.909091 -12.4843 76.3 0.909091 -12.4843 lattice-based 0.909091 -12.4843 .50 0.909091 -12.4843 0.7, 0.909091 -12.4843 .40 0.909091 -12.4843 behaviour, 0.722222 -12.4843 absolute error 0.139676 -12.4844 approach, 0.909091 -12.4846 trait 0.909091 -12.4846 4; 0.909091 -12.4846 neighborhoods 0.361111 -12.4847 some examples 0.258741 -12.4848 variables. 0.909091 -12.4849 lacked 0.909091 -12.4849 violating 0.909091 -12.4849 prototyping 0.8 -12.4849 first few 0.8 -12.4849 0.500 0.47619 -12.4851 proper noun 0.909091 -12.4852 also included. 0.909091 -12.4852 d´ejean 0.909091 -12.4852 analogical 0.47619 -12.4852 meant to 0.8 -12.4852 ahmed 0.31 -12.4853 roles. 0.909091 -12.4856 directl+ 0.909091 -12.4859 stack; 0.909091 -12.4859 reproducible 0.909091 -12.4859 270 0.592593 -12.4859 ranged 0.75 -12.4862 94% 0.909091 -12.4862 speculative 0.909091 -12.4862 grammatical, 0.909091 -12.4862 absent. 0.909091 -12.4862 tracing 0.909091 -12.4862 26.1 0.909091 -12.4862 ocur 0.909091 -12.4865 .44 0.909091 -12.4865 jargon 0.571429 -12.4865 described, 0.666667 -12.4866 covered. 0.456522 -12.4867 30, 0.909091 -12.4868 text’s 0.75 -12.4869 (talukdar 0.306931 -12.487 forms. 0.909091 -12.4871 75.9 0.75 -12.4872 24% 0.286957 -12.4872 correct. 0.909091 -12.4874 0.680 0.909091 -12.4874 aliases 0.909091 -12.4874 nc, 0.423077 -12.4877 spanish. 0.8 -12.4877 santa 0.8 -12.4877 inflections. 0.909091 -12.4878 h’s 0.571429 -12.4881 describes how 0.666667 -12.4884 proposition. 0.909091 -12.4884 texting 0.909091 -12.4884 sbv 0.909091 -12.4887 will, 0.909091 -12.4887 isahara, 0.909091 -12.4887 (jelinek 0.625 -12.4888 tb 0.722222 -12.4888 quotation 0.846154 -12.4889 alphabet matching 0.909091 -12.489 turktrain 0.722222 -12.489 punctuations 0.8 -12.489 mississippi 0.846154 -12.4892 diagnosed 0.909091 -12.4893 55.4 0.722222 -12.4893 en-es 0.228261 -12.4895 research. 0.666667 -12.4896 major difference 0.909091 -12.4896 dataset; 0.423077 -12.4898 will allow 0.909091 -12.49 bad, 0.909091 -12.4903 morpheme-level 0.909091 -12.4903 9.0 0.846154 -12.4906 clique 0.909091 -12.4906 errors; 0.909091 -12.4906 (2007); 0.909091 -12.4906 triplets 0.909091 -12.4906 72.9 0.909091 -12.4906 ostendorf, 0.909091 -12.4909 sbml 0.8 -12.491 homology-based 0.8 -12.491 filter. 0.909091 -12.4912 recognized, 0.909091 -12.4912 nederhof 0.909091 -12.4912 awarded 0.909091 -12.4912 wrapped 0.909091 -12.4912 sourceside 0.909091 -12.4912 editorials 0.75 -12.4914 grishman, 0.909091 -12.4915 dagan, 0.909091 -12.4915 surrogate 0.909091 -12.4915 wmt-small 0.666667 -12.4917 dependent, 0.4 -12.4917 application. 0.8 -12.4918 decay. 0.529412 -12.4919 parameters: 0.909091 -12.4919 probabilities) 0.909091 -12.4919 participants’ 0.909091 -12.4919 83% 0.267176 -12.4919 each node 0.571429 -12.492 agirre 0.592593 -12.492 spanish-english 0.909091 -12.4922 79.4 0.846154 -12.4924 probabilities: 0.909091 -12.4925 ancestral 0.666667 -12.4928 (yarowsky, 0.909091 -12.4928 4.2.3 0.909091 -12.4928 0.664 0.909091 -12.4928 allocated 0.909091 -12.4928 node) 0.909091 -12.4928 brew 0.909091 -12.4928 string-based 0.909091 -12.4931 manual, 0.909091 -12.4931 imposed on 0.909091 -12.4931 traum 0.487179 -12.4931 posts. 0.625 -12.4932 perfect, 0.722222 -12.4933 preposition errors 0.909091 -12.4934 boards 0.909091 -12.4934 mccord, 0.909091 -12.4934 few years 0.909091 -12.4934 values) 0.909091 -12.4934 overwhelmingly 0.341463 -12.4935 each token 0.909091 -12.4937 14.6 0.909091 -12.4937 druck 0.909091 -12.4937 decision trees. 0.909091 -12.4937 45.2 0.909091 -12.4937 disagreeing 0.0341214 -12.4938 one 0.666667 -12.4938 schemas 0.666667 -12.4938 she had 0.571429 -12.4938 keywords. 0.571429 -12.494 success. 0.909091 -12.4941 (lapata 0.909091 -12.4941 (carbonell 0.909091 -12.4941 celebrity 0.909091 -12.4941 top-5 0.529412 -12.4943 brands 0.684211 -12.4943 we formalize 0.684211 -12.4943 publications. 0.909091 -12.4944 (crfs) 0.909091 -12.4944 spam, 0.909091 -12.4944 connections, 0.909091 -12.4944 screen, 0.909091 -12.4944 canonicalization 0.909091 -12.4944 90.4 0.909091 -12.4944 resourced 0.846154 -12.4945 f1-scores 0.625 -12.4945 to reproduce 0.366197 -12.4946 perform better 0.909091 -12.4947 agrement 0.909091 -12.4947 morphophonemic 0.909091 -12.4947 sufficient, 0.909091 -12.4947 least) 0.846154 -12.495 (fung 0.361111 -12.495 are ranked 0.325843 -12.4951 constrain 0.909091 -12.4953 preference, 0.909091 -12.4953 0.604 0.666667 -12.4954 we gathered 0.303922 -12.4954 extended to 0.909091 -12.4956 use: 0.909091 -12.4956 approved 0.6 -12.4956 “this 0.75 -12.4958 11.7 0.75 -12.4958 i-prf 0.909091 -12.496 abstraction, 0.909091 -12.496 tagging) 0.75 -12.4961 nodef 0.431373 -12.4961 0.77 0.139059 -12.4962 variety of 0.909091 -12.4963 collaborator 0.909091 -12.4963 theresa 0.909091 -12.4963 assure 0.909091 -12.4966 figure 10: 0.909091 -12.4966 graphics, 0.846154 -12.4968 blueprint 0.909091 -12.4969 russia 0.909091 -12.4969 islands 0.909091 -12.4969 bag-of-words representation 0.909091 -12.4969 (kozareva 0.410714 -12.4971 preventing 0.75 -12.4971 nullnullnullnullnullnullnullnull 0.625 -12.4971 introduced. 0.909091 -12.4972 grant-in-aid for 0.909091 -12.4972 3.4) 0.909091 -12.4972 trees: 0.75 -12.4976 kirchhoff, 0.75 -12.4976 linguistic structure 0.5 -12.4976 dn 0.403509 -12.4977 realisation 0.0671078 -12.4977 present 0.0329157 -12.498 more 0.909091 -12.4982 guage 0.909091 -12.4982 clusters) 0.909091 -12.4985 bowl 0.909091 -12.4985 72.8 0.909091 -12.4985 social, 0.909091 -12.4988 identifinder 0.909091 -12.4988 1983; 0.846154 -12.4989 193 0.571429 -12.499 fortunately, 0.909091 -12.4991 34.7 0.909091 -12.4991 tree-structured 0.909091 -12.4991 58.8 0.185455 -12.4993 as: 0.625 -12.4995 no explicit 0.909091 -12.4995 chunked 0.684211 -12.4996 unambiguously 0.75 -12.4997 mercer, 0.846154 -12.4998 fung, 0.909091 -12.4998 middle, 0.909091 -12.4998 median9 0.909091 -12.4998 ordered list 0.346154 -12.5001 interact 0.909091 -12.5001 nonverbal 0.909091 -12.5001 options, 0.75 -12.5003 65.9 0.909091 -12.5004 kulkarni, 0.666667 -12.5005 table 1 presents the 0.75 -12.5005 can express 0.625 -12.5006 this happens 0.393443 -12.5006 arg2 0.846154 -12.5007 forrsooth 0.846154 -12.5007 creole sms 0.909091 -12.5007 engineered 0.909091 -12.5007 function) 0.684211 -12.5008 angbracketleftbig 0.21256 -12.5008 enables 0.454545 -12.5009 aiming 0.846154 -12.501 gim´enez and 0.75 -12.501 svm classifiers 0.909091 -12.501 452 0.909091 -12.501 0.362 0.909091 -12.501 shannon 0.909091 -12.501 subcategories 0.636364 -12.5011 wisdom of 0.625 -12.5012 quaero 0.220513 -12.5014 cases. 0.909091 -12.5014 stream, 0.909091 -12.5014 mp model 0.4375 -12.5015 to modify 0.237805 -12.5015 enough to 0.75 -12.5016 how does 0.909091 -12.5017 kondrak 0.909091 -12.5017 relation; 0.909091 -12.502 (hwa 0.909091 -12.502 26.0 0.909091 -12.502 intellectual 0.846154 -12.5021 define 0.909091 -12.5023 contiguously 0.909091 -12.5023 tacq 0.909091 -12.5026 1960) 0.909091 -12.5026 recovers 0.684211 -12.5027 nucle 0.909091 -12.503 detected, 0.909091 -12.503 realise 0.571429 -12.5032 mt04 0.846154 -12.5033 (martins et 0.909091 -12.5033 slices 0.909091 -12.5033 22.8 0.909091 -12.5033 handbook of 0.666667 -12.5035 ec 0.846154 -12.5036 by aggregating 0.909091 -12.5036 math 0.909091 -12.5036 emails, 0.909091 -12.5036 corresponded 0.909091 -12.5039 find 0.909091 -12.5039 lect 0.625 -12.5041 entailment relations 0.465116 -12.5044 on twitter 0.380952 -12.5045 with, 0.909091 -12.5046 dickinson 0.373134 -12.5047 generated. 0.909091 -12.5049 initializations 0.909091 -12.5049 alleviated 0.909091 -12.5049 interests, 0.846154 -12.5051 web search. 0.6 -12.5051 simplification, 0.909091 -12.5052 90.7 0.909091 -12.5052 .02 0.75 -12.5053 absolute value 0.909091 -12.5055 top, 0.909091 -12.5055 pssc 0.909091 -12.5055 news-commentary 0.909091 -12.5055 kilgarriff 0.625 -12.5056 (dyer 0.6 -12.5057 single-beam 0.15544 -12.5059 distinguish 0.75 -12.506 dprime 0.625 -12.5061 subsections 0.625 -12.5061 calbc 0.909091 -12.5062 checker 0.909091 -12.5062 53.8 0.53125 -12.5062 training, development 0.909091 -12.5065 reference) 0.909091 -12.5065 noun) 0.116379 -12.5066 impact 0.909091 -12.5068 bases, 0.909091 -12.5068 harness 0.909091 -12.5068 appraisal 0.625 -12.5069 r-2 0.722222 -12.5069 09] 0.909091 -12.5071 concurrent 0.909091 -12.5071 65.5 0.909091 -12.5071 (1987) 0.909091 -12.5071 contractions 0.684211 -12.5073 concretely, 0.684211 -12.5073 notation, 0.909091 -12.5074 354 0.909091 -12.5074 disappeared 0.380952 -12.5077 to judge 0.230337 -12.5077 one can 0.403509 -12.5078 couple 0.909091 -12.5078 eces 0.909091 -12.5078 patient, 0.909091 -12.5078 student, 0.909091 -12.5078 invented 0.909091 -12.5078 57% 0.909091 -12.5081 n(c) 0.909091 -12.5081 foundation, 0.909091 -12.5081 (gold) 0.909091 -12.5081 newer 0.684211 -12.5082 summarization systems 0.0842739 -12.5083 either 0.909091 -12.5084 2007b) 0.909091 -12.5084 mterater-plus 0.909091 -12.5084 t-based 0.22905 -12.5084 we investigate 0.423077 -12.5085 run on 0.846154 -12.5087 (chelba 0.75 -12.509 sorensen, 0.909091 -12.509 (linear) 0.909091 -12.509 on-going 0.909091 -12.509 a” 0.2 -12.509 extent 0.909091 -12.5094 feature hashing 0.846154 -12.5096 gene renaming 0.909091 -12.5097 (chiang 0.909091 -12.5097 consentcanvas 0.722222 -12.5097 feature values. 0.625 -12.5098 isolate 0.5 -12.5099 vbn 0.909091 -12.51 91.7 0.909091 -12.51 sv-lvcs 0.909091 -12.51 meronyms 0.909091 -12.51 name) 0.909091 -12.51 example), 0.909091 -12.51 (maxent) 0.487179 -12.51 set-up 0.625 -12.5103 decade, 0.246753 -12.5105 trend 0.909091 -12.5107 eisner’s 0.909091 -12.511 factorization, 0.909091 -12.511 g) 0.909091 -12.511 polysemous, 0.909091 -12.511 fsa 0.785714 -12.5113 (fn) 0.909091 -12.5113 (pcfg) 0.909091 -12.5113 two-word 0.846154 -12.5114 class: 0.909091 -12.5116 (pa) 0.909091 -12.5119 adition, 0.909091 -12.5119 20.5 0.909091 -12.5119 others) 0.666667 -12.512 properties: 0.909091 -12.5123 head-dependent 0.909091 -12.5123 monolinguals 0.722222 -12.5123 is necessary. 0.75 -12.5124 introductory 0.75 -12.5124 forums, 0.909091 -12.5126 reichart 0.909091 -12.5126 simple lateen 0.387097 -12.5127 for english, 0.465116 -12.5129 we illustrate 0.909091 -12.5129 lev 0.909091 -12.5129 0), 0.909091 -12.5129 reverb’s 0.909091 -12.5129 suitably 0.846154 -12.5132 trigrams. 0.666667 -12.5132 keyphrases. 0.909091 -12.5132 arguments; 0.909091 -12.5132 mccain 0.571429 -12.5133 ppt 0.666667 -12.5134 confidence intervals 0.548387 -12.5134 compression, 0.846154 -12.5135 guided models 0.846154 -12.5135 59.7 0.636364 -12.5135 we outline 0.909091 -12.5136 gun 0.909091 -12.5136 headers 0.909091 -12.5136 pretest 0.160665 -12.5137 intended 0.428571 -12.5138 by making 0.178082 -12.5139 linked 0.909091 -12.5139 1(b) 0.909091 -12.5139 mult 0.684211 -12.514 m3 0.625 -12.514 this allows for 0.265152 -12.5141 2011), 0.4375 -12.5141 parser’s 0.909091 -12.5142 rules), 0.909091 -12.5142 (ace) 0.909091 -12.5142 performance degradation 0.909091 -12.5145 lemmatizer 0.53125 -12.5146 learner’s 0.164706 -12.5147 allowing 0.666667 -12.5148 section presents 0.909091 -12.5148 interaction: 0.846154 -12.515 characterized as 0.548387 -12.515 assignments. 0.909091 -12.5152 justifying 0.333333 -12.5152 an overall 0.909091 -12.5155 situ 0.909091 -12.5155 few, 0.318681 -12.5156 were obtained 0.909091 -12.5158 with/without 0.909091 -12.5158 volden 0.909091 -12.5158 richly 0.909091 -12.5158 twets 0.909091 -12.5161 per-class 0.909091 -12.5161 rhetoric 0.909091 -12.5161 father 0.909091 -12.5161 tre 0.684211 -12.5165 j48 0.909091 -12.5165 why, 0.909091 -12.5165 senior 0.548387 -12.5167 provide useful 0.909091 -12.5168 census 0.909091 -12.5168 psychology. 0.909091 -12.5168 ill 0.551724 -12.5169 clicking 0.909091 -12.5171 to maximise 0.909091 -12.5174 (translated 0.909091 -12.5174 only). 0.909091 -12.5178 theory: 0.0350256 -12.518 was 0.909091 -12.5181 bay 0.909091 -12.5181 fig 0.909091 -12.5181 0.146 0.380952 -12.5182 our study 0.909091 -12.5184 0.576 0.909091 -12.5184 ace-2004 0.909091 -12.5187 (containing 0.909091 -12.5187 48% 0.909091 -12.5187 vsms 0.846154 -12.5189 character-level language 0.684211 -12.5189 (clark, 0.846154 -12.5192 cu-tecto 0.0564103 -12.5192 similar 0.219895 -12.5194 our approach is 0.909091 -12.5194 widdows 0.909091 -12.5194 triangle 0.909091 -12.5194 tremendous 0.666667 -12.5196 our dataset, 0.785714 -12.5197 blog post 0.846154 -12.5198 yang,heeman,andkun 0.846154 -12.5198 l(s) 0.909091 -12.52 (just 0.666667 -12.5201 hovy. 0.909091 -12.5203 extraction-based 0.909091 -12.5203 superscript 0.909091 -12.5203 meta-metric 0.909091 -12.5203 plenty 0.846154 -12.5204 lexical sample 0.684211 -12.5204 lmf 0.487179 -12.5204 we also present 0.6 -12.5206 bracketleftbig 0.75 -12.5207 word tokens, 0.909091 -12.5207 glass 0.684211 -12.5209 macquarie 0.5 -12.5209 criterion. 0.909091 -12.521 devising 0.909091 -12.521 1see 0.6 -12.5212 co-referring 0.684211 -12.5213 pink 0.909091 -12.5213 wizards’ 0.909091 -12.5213 typology, 0.909091 -12.5213 mrs. 0.551724 -12.5216 and lapata (2010) 0.909091 -12.5216 postposition 0.909091 -12.5216 lengthened 0.909091 -12.5216 narrower 0.909091 -12.5216 condensation 0.909091 -12.5216 0.001) 0.53125 -12.5216 inputs. 0.785714 -12.5217 (wan 0.75 -12.522 indications 0.909091 -12.522 accelerate 0.909091 -12.522 alphanumeric 0.909091 -12.522 µ, 0.909091 -12.522 perpetrator 0.684211 -12.5221 surdeanu 0.207547 -12.5223 encoded 0.909091 -12.5223 3); 0.909091 -12.5223 record, 0.909091 -12.5223 expectations, 0.909091 -12.5223 5.1.2 0.909091 -12.5223 bulletcircle 0.909091 -12.5223 27, 0.909091 -12.5223 cross-cutting 0.909091 -12.5223 bpc 0.909091 -12.5226 in) 0.846154 -12.5228 ext. 0.846154 -12.5228 min. 0.846154 -12.5228 degradation. 0.909091 -12.523 adverbials 0.909091 -12.523 executive 0.909091 -12.523 thereafter, 0.785714 -12.5231 http:// 0.909091 -12.5233 255 0.396552 -12.5235 source domain 0.785714 -12.5236 “if 0.75 -12.5236 mann and 0.20362 -12.5236 target word 0.909091 -12.5236 ¢ 0.909091 -12.5236 inventory, 0.636364 -12.5237 cpos 0.367647 -12.5238 be interpreted as 0.221053 -12.5239 subsets 0.785714 -12.5242 (nouns, 0.909091 -12.5243 435 0.909091 -12.5243 coord 0.909091 -12.5243 eos 0.909091 -12.5243 turn-level 0.909091 -12.5243 true: 0.75 -12.5244 victims 0.275 -12.5244 opinions, 0.362319 -12.5245 we achieve 0.909091 -12.5246 shelf 0.909091 -12.5246 8), 0.551724 -12.5248 labeled examples 0.909091 -12.5249 very sensitive to 0.909091 -12.5252 doctoral 0.909091 -12.5252 (mean 0.909091 -12.5252 prize 0.909091 -12.5252 graph: 0.909091 -12.5252 easy, 0.684211 -12.5253 palestinian 0.909091 -12.5256 300,000 0.909091 -12.5256 (0) 0.909091 -12.5259 -ed 0.909091 -12.5262 using, 0.909091 -12.5262 solvable 0.909091 -12.5262 479 0.53125 -12.5262 0.61 0.785714 -12.5264 80.5 0.380952 -12.5265 evaluated using 0.909091 -12.5265 buffer, 0.785714 -12.5267 enjoyed 0.785714 -12.5267 .97 0.144144 -12.5267 achieves 0.514286 -12.5268 have investigated 0.6 -12.5268 cases: 0.909091 -12.5269 concentrates 0.909091 -12.5269 ditransitive 0.909091 -12.5269 possession 0.909091 -12.5269 for pscfgs 0.909091 -12.5269 pascal 0.785714 -12.527 good enough 0.785714 -12.527 venue 0.551724 -12.5271 renal 0.636364 -12.5273 government, 0.909091 -12.5275 course description 0.785714 -12.5276 29.9 0.909091 -12.5278 argm-prx 0.785714 -12.5279 weight vectors 0.75 -12.5282 agr 0.75 -12.5282 steven 0.684211 -12.5285 border 0.75 -12.5288 vehicles 0.909091 -12.5288 (wong 0.909091 -12.5288 francis 0.909091 -12.5288 email, 0.318681 -12.5289 we will use 0.909091 -12.5292 free, 0.205479 -12.5293 we could 0.684211 -12.5295 sg 0.909091 -12.5295 reminiscent of 0.785714 -12.5296 ctpr 0.909091 -12.5298 rwth’s 0.684211 -12.5299 zu 0.909091 -12.5301 interjections 0.6 -12.5303 reports on 0.909091 -12.5305 facility 0.909091 -12.5305 ew 0.909091 -12.5305 efficiently, 0.909091 -12.5305 query’s 0.909091 -12.5305 bold indicates 0.909091 -12.5305 prerequisite for 0.75 -12.5306 woman 0.141304 -12.5307 suggests 0.909091 -12.5308 rehabilitation 0.909091 -12.5308 ldc’s 0.75 -12.5309 microblogging 0.6 -12.5309 proteins, 0.190476 -12.531 we plan 0.909091 -12.5311 confidence scores. 0.909091 -12.5311 comunicate 0.909091 -12.5311 emitted 0.208531 -12.5312 maximize 0.75 -12.5312 (2010) show that 0.909091 -12.5315 coordinations 0.75 -12.5317 determiner, 0.785714 -12.5318 query-aggr 0.909091 -12.5324 375 0.636364 -12.5326 feng 0.242038 -12.5328 attempts to 0.909091 -12.5328 (wikipedia, 0.909091 -12.5328 1-1 0.909091 -12.5328 agents, 0.571429 -12.533 intransitive 0.0415414 -12.533 they 0.909091 -12.5331 correlations, 0.909091 -12.5331 roll 0.909091 -12.5331 tradaboost 0.6 -12.5333 (goldwater 0.909091 -12.5334 latvian 0.909091 -12.5334 authoritative 0.909091 -12.5338 iij 0.684211 -12.5339 anna 0.785714 -12.5344 previous approaches, 0.909091 -12.5347 haven’t 0.346667 -12.5353 a log-linear 0.909091 -12.5354 typos 0.909091 -12.5357 conventions, 0.75 -12.5358 `a 0.636364 -12.5359 australia 0.909091 -12.5361 bod, 0.909091 -12.5361 county 0.909091 -12.5364 15.8 0.909091 -12.5367 pwn 0.909091 -12.5367 λ-expression 0.454545 -12.5368 an abstract 0.6 -12.5368 setup. 0.785714 -12.5372 31% 0.53125 -12.5372 crucially 0.909091 -12.5374 51.0 0.909091 -12.5374 cfgs 0.475 -12.5374 pinyin 0.705882 -12.5375 impression values 0.705882 -12.5377 mstparser (mcdonald 0.909091 -12.538 223 0.833333 -12.5383 hebrew, 0.475 -12.5383 were assigned 0.195745 -12.5383 be considered 0.785714 -12.5384 candidate keyphrases 0.785714 -12.5384 35.2 0.636364 -12.5384 translation candidates 0.909091 -12.5384 distinct, 0.833333 -12.5386 governs 0.785714 -12.5387 conclusions, 0.909091 -12.5387 derivationally 0.909091 -12.5387 localized 0.785714 -12.5389 abstractive sentence 0.909091 -12.539 (1994), 0.909091 -12.539 455 0.909091 -12.539 matched, 0.909091 -12.539 metacost 0.909091 -12.539 arguably, 0.833333 -12.5392 seat 0.833333 -12.5392 8-form 0.454545 -12.5393 product. 0.909091 -12.5394 strategies: 0.909091 -12.5394 updates, 0.486486 -12.5394 europarl corpus 0.833333 -12.5395 framenet (baker 0.6 -12.5396 books, 0.909091 -12.5397 sister 0.909091 -12.5397 q(·) 0.909091 -12.5397 selections 0.785714 -12.5398 au4 0.576923 -12.5398 ner. 0.833333 -12.5399 capitalized, 0.909091 -12.54 cognitive-science 0.53125 -12.5401 difference, 0.6 -12.5403 asr system 0.5 -12.5404 every pair 0.280702 -12.5406 wordnet. 0.785714 -12.5407 label assignment 0.833333 -12.5408 0.909091 -12.541 r-bleu 0.833333 -12.5411 “≥ 0.833333 -12.5411 typicaly 0.145882 -12.5413 penn 0.909091 -12.5414 test: 0.785714 -12.5415 1800 0.352113 -12.5415 is probably 0.909091 -12.5417 224 0.909091 -12.5417 interviews 0.909091 -12.5417 asterisk 0.389831 -12.5418 pereira 0.909091 -12.542 dpmm 0.833333 -12.542 adverb, 0.909091 -12.5423 wikis 0.833333 -12.5423 well-established 0.369231 -12.5423 room for 0.705882 -12.5424 tropical 0.705882 -12.5424 international journal 0.705882 -12.5426 kenneth 0.486486 -12.5426 technology. 0.75 -12.5429 excerpt from 0.551724 -12.5429 (4). 0.833333 -12.5429 utilise 0.833333 -12.5429 lag 0.909091 -12.543 notorious 0.785714 -12.5432 (han 0.833333 -12.5432 in proceedings of acl 0.53125 -12.5434 alignments between 0.833333 -12.5435 native language, 0.833333 -12.5435 poetic 0.833333 -12.5435 garment 0.833333 -12.5435 vr 0.551724 -12.5436 steve 0.909091 -12.5437 follows; 0.909091 -12.5437 0.2, 0.785714 -12.5438 simplification rules 0.064695 -12.5439 paper 0.75 -12.544 98% 0.909091 -12.544 non-root 0.909091 -12.544 c4 0.909091 -12.544 user interaction 0.785714 -12.5441 girju 0.833333 -12.5441 i’ll 0.833333 -12.5441 off, 0.909091 -12.5443 andy 0.909091 -12.5443 among others 0.909091 -12.5443 1–5 0.684211 -12.5446 bracehtipdownrightbracehtipdownleft 0.785714 -12.5447 vector) 0.833333 -12.5447 s/o 0.833333 -12.5447 topic-word 0.833333 -12.5447 arthur 0.684211 -12.5449 .2 0.486486 -12.5449 cb 0.785714 -12.545 22.2 0.909091 -12.545 animal, 0.909091 -12.545 submissions, 0.909091 -12.545 hoped 0.362319 -12.545 same. 0.515152 -12.5451 heldout 0.909091 -12.5453 mada 0.684211 -12.5454 res 0.785714 -12.5455 +coref 0.311828 -12.5457 weights are 0.705882 -12.5457 abstracting 0.909091 -12.546 upcoming 0.833333 -12.5462 247 0.833333 -12.5462 cause-effect 0.833333 -12.5462 evaluation campaign. 0.833333 -12.5462 r+ 0.636364 -12.5463 substitutions, 0.909091 -12.5463 “provide 0.909091 -12.5463 we’re 0.785714 -12.5464 stable, 0.833333 -12.5466 else, 0.168224 -12.5467 mutual 0.333333 -12.5468 some other 0.833333 -12.5469 suported 0.833333 -12.5469 3in 0.0175576 -12.5469 with 0.785714 -12.5473 (o) 0.833333 -12.5475 willing to 0.833333 -12.5475 orders, 0.65 -12.5478 accurately. 0.833333 -12.5478 synthesis, 0.486486 -12.5479 tales 0.833333 -12.5481 gene, 0.833333 -12.5481 tbl 0.126786 -12.5484 furthermore, 0.785714 -12.5484 test10 0.833333 -12.5484 93.9 0.294118 -12.5486 referent 0.5 -12.5486 case-insensitive 0.833333 -12.5487 bt 0.75 -12.5489 comp 0.6 -12.5491 guidelines. 0.383333 -12.5492 small. 0.428571 -12.5493 srl-aware 0.833333 -12.5496 2010] 0.407407 -12.5496 dept. of 0.785714 -12.5499 undesired 0.833333 -12.5499 bindings 0.0232767 -12.5501 an 0.0858545 -12.5502 associated 0.833333 -12.5502 hierarchical fusion 0.352113 -12.5504 whether they 0.785714 -12.5505 ascending 0.785714 -12.5505 phr 0.636364 -12.5505 unlabeled data, 0.833333 -12.5505 tiny 0.833333 -12.5505 deep similarity 0.236025 -12.5506 equivalent to 0.833333 -12.5508 sub-corpus 0.833333 -12.5508 ments 0.75 -12.5509 3.3.1 0.785714 -12.551 91.2 0.785714 -12.551 56% 0.833333 -12.5511 bnc, 0.75 -12.5512 1st-tier 0.705882 -12.5517 good performance. 0.833333 -12.5518 nou 0.833333 -12.5518 delicious 0.608696 -12.5518 gravano 0.434783 -12.5519 one might 0.833333 -12.5521 (its 0.785714 -12.5522 90.2 0.684211 -12.5522 feature space. 0.251799 -12.5524 entity recognition 0.75 -12.5525 markup. 0.146283 -12.5526 tables 0.207729 -12.5529 needs to 0.705882 -12.553 family. 0.329268 -12.5531 r = 0.551724 -12.5531 heuristics. 0.785714 -12.5533 found out 0.833333 -12.5533 cd, 0.65 -12.5535 insertions and 0.833333 -12.5536 medical entities 0.636364 -12.5538 each cell 0.705882 -12.554 marco 0.463415 -12.5542 each row 0.833333 -12.5545 phrase: 0.705882 -12.5546 sight 0.833333 -12.5551 dtu 0.576923 -12.5551 turns. 0.636364 -12.5552 published by 0.0125678 -12.5556 for 0.785714 -12.5557 brent 0.785714 -12.5557 22.4 0.576923 -12.5557 increment 0.75 -12.5559 have equal 0.785714 -12.556 are planning 0.189516 -12.5561 relatedness 0.705882 -12.5561 ptb, 0.785714 -12.5563 (1989) 0.184615 -12.5564 a larger 0.75 -12.5564 expert. 1 -12.5564 positive). 1 -12.5564 iit bombay 1 -12.5564 substring. 1 -12.5564 bd. 1 -12.5564 attachments. 1 -12.5564 4676 admiralty 1 -12.5564 apart. 1 -12.5564 her. 1 -12.5564 textgraphs-6 workshop, 1 -12.5564 variability. 1 -12.5564 movement. 1 -12.5564 tense. 1 -12.5564 microtexts. 1 -12.5564 completeness. 1 -12.5564 non-understanding. 1 -12.5564 death. 1 -12.5564 tagged. 1 -12.5564 tone. 1 -12.5564 sketch. 1 -12.5564 boxes. 1 -12.5564 coherent. 1 -12.5564 bouchard-cˆot´e 1 -12.5564 republic. 1 -12.5564 benchmark. 1 -12.5564 (10). 1 -12.5564 full. 1 -12.5564 terminals. 1 -12.5564 apple. 1 -12.5564 thereof. 1 -12.5564 zipf’s law 1 -12.5564 bo pang, 1 -12.5564 anotation librarian 1 -12.5564 avoided. 1 -12.5564 consolidate 1 -12.5564 wintner, 1 -12.5564 (207). 1 -12.5564 evaluators. 1 -12.5564 (amig´o 1 -12.5564 state-of-the-art. 1 -12.5564 enabled. 1 -12.5564 bases. 1 -12.5564 landmark. 1 -12.5564 debate. 1 -12.5564 2006)). 1 -12.5564 snowfall 1 -12.5564 paterns. 1 -12.5564 0.001. 1 -12.5564 discourse-level 1 -12.5564 complements. 1 -12.5564 throughout. 1 -12.5564 productive, 1 -12.5564 detectors. 1 -12.5564 1979). 1 -12.5564 against. 1 -12.5564 fly. 1 -12.5564 assessments. 1 -12.5564 agreements. 1 -12.5564 order). 1 -12.5564 dropped. 1 -12.5564 conversion. 1 -12.5564 vowels. 1 -12.5564 htk 1 -12.5564 inaccurate. 1 -12.5564 sms. 1 -12.5564 centroid. 1 -12.5564 mission 4636 1 -12.5564 stems. 1 -12.5564 data? 1 -12.5564 (budanitsky 1 -12.5564 (+0.7) 1 -12.5564 explained. 1 -12.5564 extreme. 1 -12.5564 extensibility. 1 -12.5564 month. 1 -12.5564 stars. 1 -12.5564 ce. 1 -12.5564 converge. 1 -12.5564 suites. 1 -12.5564 2011b). 1 -12.5564 lexica. 1 -12.5564 cn. 1 -12.5564 scalable. 1 -12.5564 germany. 1 -12.5564 vertex. 1 -12.5564 cycle. 1 -12.5564 degrees. 1 -12.5564 separated. 1 -12.5564 edit. 1 -12.5564 difficulties. 1 -12.5564 women. 1 -12.5564 regular. 1 -12.5564 upward closure 1 -12.5564 projections. 1 -12.5564 della pietra, 1 -12.5564 simulations. 1 -12.5564 (+0.5) 1 -12.5564 occurrence. 1 -12.5564 cross-source 1 -12.5564 compute. 1 -12.5564 hlt. 1 -12.5564 them). 1 -12.5564 filtered. 1 -12.5564 integration. 1 -12.5564 accuracy). 1 -12.5564 experiments). 1 -12.5564 principle. 1 -12.5564 verb-argument 1 -12.5564 increasing. 1 -12.5564 weischedel. 1 -12.5564 miyao 1 -12.5564 novels. 1 -12.5564 s1. 1 -12.5564 (8). 1 -12.5564 audience. 1 -12.5564 thumbs 1 -12.5564 effectiveness. 1 -12.5564 .ie 1 -12.5564 players. 1 -12.5564 measured. 1 -12.5564 original. 1 -12.5564 minipar 1 -12.5564 ssc. 1 -12.5564 eg. 1 -12.5564 algebra. 1 -12.5564 i). 1 -12.5564 208). 1 -12.5564 13. 1 -12.5564 k-means, 1 -12.5564 inter-sentential 1 -12.5564 typology. 1 -12.5564 1(b). 1 -12.5564 (12 1 -12.5564 underspecified. 1 -12.5564 modelling, 1 -12.5564 digits. 1 -12.5564 paired t-test. 1 -12.5564 happens. 1 -12.5564 clarity. 1 -12.5564 (§5). 1 -12.5564 1997. 1 -12.5564 slovak. 1 -12.5564 afinitor she 1 -12.5564 say. 1 -12.5564 supportive 1 -12.5564 impractical. 1 -12.5564 laborious 1 -12.5564 fes. 1 -12.5564 voting. 1 -12.5564 million. 1 -12.5564 (tsuruoka 1 -12.5564 line). 1 -12.5564 hp/np 1 -12.5564 conscious 1 -12.5564 failure. 1 -12.5564 homotopy tracking 1 -12.5564 subtype. 1 -12.5564 medline. 1 -12.5564 nominated headline 1 -12.5564 diagram. 1 -12.5564 affected. 1 -12.5564 1a. 1 -12.5564 affinities 1 -12.5564 parts-of-speech. 1 -12.5564 windows. 1 -12.5564 interactivity 1 -12.5564 received. 1 -12.5564 encouraging. 1 -12.5564 clusterings. 1 -12.5564 (lm). 1 -12.5564 categories). 1 -12.5564 sentiwordnet 3.0 1 -12.5564 epsilon1. 1 -12.5564 well-founded 1 -12.5564 checking. 1 -12.5564 mallet (mccallum, 1 -12.5564 programs. 1 -12.5564 (platt 1 -12.5564 david chiang, 1 -12.5564 anger, fear, 1 -12.5564 weak. 1 -12.5564 d2. 1 -12.5564 prescription 1 -12.5564 segmented. 1 -12.5564 granularities. 1 -12.5564 textrank (mihalcea 1 -12.5564 listeners. 1 -12.5564 ibw plus 1 -12.5564 train. 1 -12.5564 4.3). 1 -12.5564 isr. 1 -12.5564 post-processing. 1 -12.5564 storage. 1 -12.5564 ramat-gan, israel 1 -12.5564 mismatches. 1 -12.5564 conjunctions. 1 -12.5564 gold. 1 -12.5564 typing. 1 -12.5564 hypernymy. 1 -12.5564 varieties. 1 -12.5564 easy. 1 -12.5564 grounding. 1 -12.5564 persons. 1 -12.5564 (206). 1 -12.5564 keys. 1 -12.5564 1%. 1 -12.5564 transcription. 1 -12.5564 adjacent. 1 -12.5564 taira 1 -12.5564 memorized 1 -12.5564 reactions. 1 -12.5564 state-of-art 1 -12.5564 sub lattices 1 -12.5564 deviant, 1 -12.5564 considerably. 1 -12.5564 body. 1 -12.5564 confusion. 1 -12.5564 grishman. 1 -12.5564 significant). 1 -12.5564 bitexts. 1 -12.5564 lexemes. 1 -12.5564 pool. 1 -12.5564 game-based 1 -12.5564 cognitive science, 1 -12.5564 apparent. 1 -12.5564 though. 1 -12.5564 side). 1 -12.5564 gestures. 1 -12.5564 hebrew. 1 -12.5564 preterminal. 1 -12.5564 perceptron. 1 -12.5564 performance). 1 -12.5564 diseases. 0.6 -12.5564 treetagger (schmid, 0.533333 -12.5565 informatics, 0.705882 -12.5567 reconstructed 0.833333 -12.5567 187 0.833333 -12.5567 document d. 0.9 -12.5568 today. 0.9 -12.5568 information). 0.9 -12.5568 drawbacks. 0.9 -12.5568 inflections, 0.9 -12.5568 japan. 0.9 -12.5568 binary. 0.9 -12.5568 combinatorial explosion 0.9 -12.5568 beforehand. 0.9 -12.5568 evaluation (lrec 0.9 -12.5568 hazard ratio 0.9 -12.5568 duc 2006 0.9 -12.5568 needing 0.9 -12.5568 share. 0.9 -12.5568 email. 0.9 -12.5568 row) 0.9 -12.5568 pp pp 0.9 -12.5568 lecture notes 0.9 -12.5568 cross-event inference 0.9 -12.5568 decreased. 0.9 -12.5568 facto 0.9 -12.5568 events). 0.9 -12.5568 conducted. 0.9 -12.5568 k-best lists. 0.9 -12.5568 column). 0.833333 -12.557 [vp 0.9 -12.5571 signal. 0.833333 -12.5573 241 0.9 -12.5574 artificial intelligence. 0.785714 -12.5574 three major 0.342105 -12.5575 remainder of this 0.705882 -12.5577 series, 0.9 -12.5578 talked about 0.9 -12.5581 so on). 0.9 -12.5581 multiword expression 0.684211 -12.5582 compliance 0.833333 -12.5582 wd 0.785714 -12.5583 an entity. 0.785714 -12.5583 β1 0.636364 -12.5583 court 0.9 -12.5584 (luo, 0.389831 -12.5584 reasons for 0.833333 -12.5585 biboot 0.833333 -12.5585 departing 0.369231 -12.5586 this procedure 0.608696 -12.5586 film 0.9 -12.5588 concept map 0.9 -12.5588 permuting 0.705882 -12.5588 71.8 0.551724 -12.5588 (top 0.236025 -12.5589 vectors. 0.833333 -12.5589 triangulation 0.9 -12.5591 (sutton 0.9 -12.5591 sutton 0.9 -12.5591 (table 1) 0.785714 -12.5592 pie 0.833333 -12.5592 188 0.705882 -12.5593 75.4 0.9 -12.5594 (guo 0.785714 -12.5595 8.5 0.65 -12.5595 action-value 0.9 -12.5598 (stymne, 0.833333 -12.5598 zh 0.785714 -12.56 51.6 0.9 -12.5601 lightly-supervised training 0.833333 -12.5601 animate 0.833333 -12.5601 analytical 0.705882 -12.5603 erotic 0.9 -12.5604 necessitating 0.833333 -12.5604 rochester 0.65 -12.5605 alan 0.684211 -12.5605 (m 0.377049 -12.5606 participants, 0.65 -12.5607 in nlp, 0.833333 -12.5607 waiting for 0.347222 -12.5608 our parser 0.785714 -12.5609 85.6 0.785714 -12.5609 signaled 0.785714 -12.5609 cv 0.9 -12.5611 unless otherwise 0.9 -12.5611 easy-to-read 0.9 -12.5615 equally, 0.9 -12.5615 difficult–to–read 0.9 -12.5615 shieber, 0.833333 -12.5616 conception 0.785714 -12.5618 4-class 0.392857 -12.562 each step 0.833333 -12.562 painted 0.9 -12.5621 narayanan, 0.9 -12.5621 1/3 0.636364 -12.5623 laws 0.9 -12.5625 conformity 0.9 -12.5625 yt(i) 0.9 -12.5625 curve, 0.9 -12.5625 lemmatize 0.833333 -12.5626 ord 0.5 -12.5627 total. 0.9 -12.5628 opposite direction. 0.9 -12.5628 matusov 0.684211 -12.5628 grid. 0.833333 -12.5629 lei 0.833333 -12.5629 question mark 0.0735115 -12.563 defined 0.9 -12.5631 isozaki 0.9 -12.5631 unsupervised self-trained 0.142529 -12.5635 except 0.705882 -12.5637 ),( 0.9 -12.5638 depths 0.157303 -12.5639 row 0.705882 -12.564 kernelized 0.9 -12.5641 {dpi, 0.9 -12.5641 1...n, 0.9 -12.5641 probabilistic synchronous 0.9 -12.5641 28.8 0.9 -12.5641 relaxation, 0.785714 -12.5641 plakias 0.785714 -12.5641 celikyilmaz 0.9 -12.5645 (izumi 0.785714 -12.5647 original numerical 0.833333 -12.5647 activate 0.9 -12.5648 sds, 0.9 -12.5651 repositories. 0.103912 -12.5653 frequent 0.608696 -12.5654 not allowed 0.9 -12.5655 future research, 0.9 -12.5655 auxiliary tasks 0.705882 -12.5656 text normalization 0.9 -12.5658 definitive 0.9 -12.5658 0.875 0.65 -12.5658 system). 0.608696 -12.5658 we also observe 0.785714 -12.5659 msnbc 0.4 -12.5659 partially supported 0.684211 -12.5661 an old 0.298969 -12.5661 the ith 0.9 -12.5662 114,501 0.9 -12.5662 maximum activation 0.833333 -12.5663 fight 0.9 -12.5665 as previously mentioned, 0.9 -12.5665 googleweb 0.9 -12.5665 bwsa 0.636364 -12.5666 fields. 0.833333 -12.5666 nowadays, 0.9 -12.5668 programing 0.9 -12.5668 0.625 0.9 -12.5668 total), 0.392857 -12.5669 lrscore 0.9 -12.5672 aftermath 0.9 -12.5672 11.2 0.9 -12.5672 (in addition 0.9 -12.5672 bibtex 0.833333 -12.5672 2006): 0.705882 -12.5674 annotation process. 0.533333 -12.5674 (miller 0.9 -12.5675 annotation guidelines. 0.9 -12.5675 normalisation rules 0.9 -12.5675 described earlier. 0.833333 -12.5676 multiplier 0.9 -12.5678 benefits 0.9 -12.5678 2/3 0.833333 -12.5679 251 0.705882 -12.568 (bangalore 0.65 -12.5682 62.9 0.65 -12.5682 pos-tagger 0.684211 -12.5682 rappoport, 0.785714 -12.5683 running example 0.515152 -12.5683 and moschitti, 0.9 -12.5685 referring expressions, 0.9 -12.5685 ?; 0.9 -12.5685 (aho 0.833333 -12.5685 e′ 0.833333 -12.5685 exponentially large 0.434783 -12.5687 include: 0.608696 -12.5688 .98 0.608696 -12.569 0.31 0.785714 -12.5691 br87 0.833333 -12.5691 (same 0.9 -12.5692 mutual information (mi) 0.9 -12.5692 0.25, 0.9 -12.5692 m¨oller 0.9 -12.5695 (di 0.9 -12.5695 tailor 0.9 -12.5695 computer-generated 0.9 -12.5695 consume 0.9 -12.5695 interrupt 0.9 -12.5695 unedited 0.251799 -12.5696 occurred 0.9 -12.5699 0.337 0.555556 -12.5699 section 3.1 0.205742 -12.5699 nor 0.230303 -12.5702 grammar. 0.9 -12.5702 90.8 0.9 -12.5702 problem; 0.9 -12.5702 english→french 0.9 -12.5702 similarity) 0.9 -12.5702 0.390 0.9 -12.5702 gispert 0.9 -12.5702 factive 0.9 -12.5702 pos) 0.9 -12.5702 sacrifice 0.9 -12.5702 self-paced 0.9 -12.5702 analysis) 0.785714 -12.5703 67.3 0.463415 -12.5704 unigrams, 0.9 -12.5705 0.278 0.9 -12.5705 attributing 0.9 -12.5705 english→german 0.9 -12.5705 bypassing 0.9 -12.5705 treelets, 0.9 -12.5705 advantage, 0.9 -12.5705 re-use 0.347222 -12.5705 polarity. 0.9 -12.5709 harms 0.9 -12.5709 (quinlan, 0.785714 -12.5709 electrical 0.833333 -12.571 cy 0.9 -12.5712 cpl 0.9 -12.5712 (see table 1) 0.9 -12.5712 (mimno 0.9 -12.5712 relevancy 0.9 -12.5716 parsing-based 0.833333 -12.5716 prosodic break 0.833333 -12.5716 systematic comparison 0.9 -12.5719 (sennrich, 0.9 -12.5719 transfer-based 0.9 -12.5719 yangarber 0.9 -12.5719 cucerzan 0.515152 -12.572 dependency treebank 0.785714 -12.5721 an attractive 0.785714 -12.5721 4-grams 0.9 -12.5722 two-thirds 0.9 -12.5722 93.8 0.9 -12.5722 23.3 0.9 -12.5722 grammaticality improvement 0.9 -12.5722 29.8 0.705882 -12.5722 1 million 0.321429 -12.5723 6.3 0.785714 -12.5724 normalized, 0.9 -12.5726 sx 0.321429 -12.5726 p2 0.9 -12.5729 beams 0.9 -12.5729 unification-based 0.576923 -12.5729 topic shift 0.9 -12.5732 kuhn, 0.9 -12.5732 shiftreduce 0.9 -12.5732 tracked 0.533333 -12.5732 agent’s 0.785714 -12.5733 as expected. 0.705882 -12.5733 section 3.1). 0.176678 -12.5735 next, 0.9 -12.5736 sfc 0.785714 -12.5736 arg1, 0.785714 -12.5736 eeg 0.321429 -12.5736 both of these 0.186508 -12.5736 process, 0.116641 -12.5737 likely to 0.9 -12.5739 correlating 0.9 -12.5739 strengthens 0.9 -12.5739 berant 0.608696 -12.574 till 0.9 -12.5743 kindly 0.9 -12.5743 (gamon 0.9 -12.5743 anchors, 0.576923 -12.5743 bcubed 0.9 -12.5746 32k 0.9 -12.5746 (hoang 0.9 -12.5746 (chandrasekar 0.65 -12.5748 guideline 0.9 -12.5749 error-prone 0.9 -12.5749 mdi 0.9 -12.5749 53.1 0.452381 -12.5749 feature: 0.288462 -12.5751 japan 0.833333 -12.5751 (unlike 0.833333 -12.5751 respondents 0.833333 -12.5751 eisner, 2009) 0.230303 -12.5753 an input 0.9 -12.5753 upper stage 0.9 -12.5753 senmpqa 0.9 -12.5753 treebank-style 0.9 -12.5753 less, 0.785714 -12.5753 telic 0.705882 -12.5754 for deriving 0.9 -12.5756 (qa) 0.9 -12.5756 l-types 0.9 -12.5756 distinction, 0.9 -12.5756 53.2 0.533333 -12.5758 fragment. 0.9 -12.576 coecke 0.288462 -12.5765 reliability of 0.705882 -12.5765 macro-average 0.9 -12.5766 lsa-based 0.9 -12.5766 shriberg 0.9 -12.5766 84.3 0.9 -12.5766 styles, 0.9 -12.5766 61.9 0.9 -12.5766 22.7 0.9 -12.5766 (choi 0.9 -12.5766 4,000 0.9 -12.577 bolh 0.9 -12.5773 turntaking 0.9 -12.5773 target-independent 0.9 -12.5773 revu-nl 0.515152 -12.5773 we noticed 0.411765 -12.5775 we measured 0.608696 -12.5775 classification tasks. 0.9 -12.5777 memes 0.9 -12.578 resident 0.9 -12.578 sub-sequence 0.411765 -12.5782 mention, 0.9 -12.5783 yr 0.9 -12.5787 88.3 0.9 -12.5787 start/end 0.9 -12.5787 schone 0.9 -12.5787 efort 0.785714 -12.5789 bruce 0.833333 -12.5789 clausal 0.9 -12.579 2009)), 0.9 -12.579 yij 0.9 -12.579 beukelman 0.9 -12.579 (callison-burch, 0.9 -12.579 recordings, 0.576923 -12.5793 phd 0.9 -12.5794 84.7 0.9 -12.5794 selling 0.9 -12.5794 argument’s 0.9 -12.5797 copestake, 0.9 -12.5797 utterance’s 0.9 -12.5797 0.843 0.9 -12.5797 system-user 0.9 -12.5797 highest-ranked 0.705882 -12.5797 manchester 0.785714 -12.5798 bart 0.9 -12.5801 backed 0.9 -12.5801 (koller 0.9 -12.5801 deep parser 0.9 -12.5801 grammar’s 0.9 -12.5801 computable 0.785714 -12.5801 analyzer, 0.833333 -12.5801 medco 0.9 -12.5804 17.4 0.9 -12.5804 n-gram posterior 0.515152 -12.5804 was run 0.833333 -12.5804 abandoned 0.9 -12.5807 yuret, 0.9 -12.5807 rule extraction, 0.9 -12.5807 grounding, 0.9 -12.5807 wh-question 0.358209 -12.5807 transducer 0.9 -12.5811 typed, 0.9 -12.5811 “o” 0.9 -12.5811 conventionally 0.833333 -12.5811 two stages. 0.9 -12.5814 eigenvalue 0.9 -12.5814 nontrivial 0.9 -12.5814 guesser 0.833333 -12.5814 vad 0.785714 -12.5816 anticipation 0.9 -12.5818 (nakagawa 0.9 -12.5818 .79 0.370968 -12.582 purposes of 0.9 -12.5821 gramatical 0.9 -12.5821 plausibly 0.9 -12.5821 list) 0.9 -12.5821 (st) 0.833333 -12.5823 on: 0.9 -12.5824 ctb5 0.9 -12.5824 43.2 0.9 -12.5824 challenges: 0.9 -12.5824 inexact 0.403846 -12.5826 person. 0.9 -12.5828 memm 0.9 -12.5828 (1, 0.302083 -12.5832 were trained 0.234177 -12.5832 2001). 0.576923 -12.5832 5 times 0.9 -12.5835 73.6 0.9 -12.5835 copy-free 0.833333 -12.5836 5.3.1 0.9 -12.5838 understand, 0.9 -12.5838 advance, 0.733333 -12.5839 gapped 0.9 -12.5842 pos-tagging, 0.785714 -12.5843 < 0.05). 0.576923 -12.5843 appearance 0.733333 -12.5845 p10 0.9 -12.5845 closure, 0.9 -12.5845 lexicosyntactic 0.9 -12.5845 arsl 0.785714 -12.5846 tectomt 0.9 -12.5848 re-ordering 0.9 -12.5848 winston 0.9 -12.5848 postulate 0.9 -12.5848 restaurants 0.9 -12.5848 32-bit 0.9 -12.5848 steady 0.785714 -12.5849 maximum, 0.0478067 -12.585 however, 0.9 -12.5852 complexes 0.9 -12.5852 17.7 0.9 -12.5852 foundational 0.785714 -12.5852 null1, 0.9 -12.5855 574 0.9 -12.5855 57.4 0.9 -12.5855 markup, 0.9 -12.5855 maps, 0.9 -12.5859 54.4 0.9 -12.5859 cost estimation 0.298969 -12.5859 they use 0.329114 -12.586 been developed 0.705882 -12.5861 existing resources 0.9 -12.5862 conclusion: 0.9 -12.5862 wordnet.ptglobal 0.9 -12.5862 consistently, 0.9 -12.5862 negotiate 0.9 -12.5862 min, 0.9 -12.5862 state-action 0.176895 -12.5865 annotation. 0.833333 -12.5865 (available 0.9 -12.5866 alignilp 0.9 -12.5866 aware, 0.358209 -12.5866 contributes 0.785714 -12.5867 approximate search 0.833333 -12.5868 automatically parsed 0.833333 -12.5868 more details on 0.833333 -12.5868 rmse 0.9 -12.5869 researched 0.9 -12.5869 well; 0.9 -12.5869 2.3% 0.9 -12.5869 signing 0.833333 -12.5871 computer generated 0.9 -12.5872 obtainable 0.9 -12.5872 quote/response 0.9 -12.5872 92.6 0.9 -12.5872 dual-classifier 0.9 -12.5872 87.2 0.9 -12.5872 normalise 0.452381 -12.5872 categorial 0.434783 -12.5873 serve as a 0.0631579 -12.5875 could 0.325 -12.5875 novel approach 0.9 -12.5876 29.7 0.9 -12.5876 negation/speculation 0.833333 -12.5877 219 0.44186 -12.5878 they contain 0.9 -12.5879 bilingually 0.9 -12.5879 rayson 0.329114 -12.588 quantify 0.234177 -12.588 table 1, 0.733333 -12.5881 tweet: 0.608696 -12.5881 zhu, 0.9 -12.5883 bold, 0.9 -12.5883 favoring 0.65 -12.5884 natural, 0.533333 -12.5884 own. 0.833333 -12.5884 medical language 0.833333 -12.5884 randomly sample 0.9 -12.5886 85.1 0.833333 -12.5887 the advent 0.9 -12.589 distribution; 0.9 -12.589 bk 0.9 -12.589 (1996), 0.733333 -12.5893 pyysalo 0.9 -12.5893 convergent 0.833333 -12.5893 est 0.833333 -12.5893 adaboost 0.65 -12.5894 performance, but 0.9 -12.5897 extract, 0.9 -12.59 gates 0.9 -12.59 inc., 0.9 -12.59 sauper 0.9 -12.59 oovs, 0.785714 -12.59 weapons 0.485714 -12.59 composition. 0.733333 -12.5901 did. 0.44186 -12.5901 consequence of 0.9 -12.5903 categorisation 0.9 -12.5903 44.1 0.9 -12.5903 snippets, 0.9 -12.5903 infixes 0.9 -12.5903 sorted list 0.9 -12.5907 24.0 0.209184 -12.5907 illustrates 0.733333 -12.5909 18, 0.533333 -12.5909 associates 0.833333 -12.5909 source: 0.9 -12.591 glarf 0.9 -12.591 content-expert 0.608696 -12.5912 govern 0.9 -12.5914 subjects’ 0.9 -12.5914 permissive 0.9 -12.5914 (co) 0.9 -12.5914 65.3 0.9 -12.5914 re-estimation 0.9 -12.5914 interpretable, 0.9 -12.5914 statement, 0.155367 -12.5914 we propose a 0.65 -12.5914 src 0.785714 -12.5915 president, 0.9 -12.5917 (platt, 0.9 -12.5917 pauls 0.9 -12.5917 confounding 0.205882 -12.5918 distance between 0.304348 -12.5919 future work, we 0.452381 -12.5919 7). 0.555556 -12.592 n+1 0.733333 -12.5921 proves that 0.9 -12.5921 reader’s 0.9 -12.5921 83.6 0.9 -12.5921 ewtm 0.9 -12.5921 evert 0.9 -12.5921 micro-averaged accuracy 0.9 -12.5921 (nakov 0.785714 -12.5921 extra information 0.833333 -12.5922 l1, 0.833333 -12.5922 lsa, 0.833333 -12.5922 ruppenhofer 0.9 -12.5924 fusing 0.9 -12.5924 4.3) 0.9 -12.5924 twenty-five 0.833333 -12.5925 (fan et 0.785714 -12.5927 i− 0.261905 -12.5927 rule. 0.576923 -12.5927 stack, 0.9 -12.5928 themes, 0.9 -12.5928 (porter, 0.9 -12.5928 (β 0.9 -12.5931 proprietary 0.9 -12.5931 roles: 0.261905 -12.5931 probability. 0.785714 -12.5933 (among 0.9 -12.5935 23.6 0.425532 -12.5935 projecting 0.9 -12.5938 0.404 0.9 -12.5938 an extensible 0.9 -12.5938 sentiment) 0.9 -12.5938 (solid 0.833333 -12.5938 accepted, 0.785714 -12.5939 52.3 0.65 -12.5939 τb 0.177778 -12.5939 the test set 0.9 -12.5941 taskar 0.9 -12.5941 bulb 0.9 -12.5941 transitive, 0.9 -12.5941 56.3 0.9 -12.5945 81.4 0.9 -12.5945 product reviews, 0.9 -12.5945 tang 0.9 -12.5945 tures 0.9 -12.5945 unnecessarily 0.9 -12.5948 analysis; 0.705882 -12.5948 hirschberg, 0.9 -12.5952 338 0.9 -12.5952 parse-based 0.9 -12.5952 93.34 0.9 -12.5952 teh 0.9 -12.5952 construed as 0.9 -12.5955 59.6 0.9 -12.5955 0.4, 0.9 -12.5955 8.4 0.9 -12.5955 ramshaw 0.705882 -12.5956 wsd method 0.608696 -12.5956 svs 0.9 -12.5959 tts-based 0.9 -12.5959 input-output 0.9 -12.5959 kyoto, japan 0.666667 -12.596 thater 0.833333 -12.596 institutes 0.9 -12.5962 targetside 0.9 -12.5962 emergence 0.9 -12.5962 4.2% 0.301075 -12.5962 not able 0.733333 -12.5963 r5 0.785714 -12.5963 another line 0.619048 -12.5963 yet. 0.65 -12.5964 current approaches 0.833333 -12.5964 table 5 presents 0.833333 -12.5964 comparedto 0.9 -12.5966 movie, 0.0576258 -12.5966 show 0.9 -12.5969 tendencies 0.9 -12.5969 handwriting 0.9 -12.5969 round, 0.9 -12.5969 guevara, 0.9 -12.5969 enhancement 0.785714 -12.5969 concludes. 0.785714 -12.5969 resp. 0.9 -12.5973 apropriate 0.9 -12.5973 warrant 0.9 -12.5973 impacting 0.833333 -12.5973 nk 0.65 -12.5974 readers, 0.583333 -12.5975 narration 0.9 -12.5976 memorizing 0.9 -12.5976 nist, 0.9 -12.5976 saga 0.9 -12.5976 capitalize 0.127619 -12.5979 far 0.9 -12.598 [+ 0.619048 -12.598 10 iterations 0.733333 -12.5983 outer 0.9 -12.5983 pad´o, 0.9 -12.5983 acoustics 0.9 -12.5983 (sentences 0.9 -12.5983 (baker, 0.9 -12.5983 containers 0.461538 -12.5983 also contains 0.0989761 -12.5984 we present 0.555556 -12.5984 satta 0.833333 -12.5986 10.4 0.9 -12.5987 before: 0.9 -12.5987 231 0.9 -12.5987 hyphenated 0.9 -12.5987 248 0.9 -12.5987 five-point 0.9 -12.599 subclasof 0.9 -12.599 interchangeably 0.9 -12.599 modi 0.9 -12.599 decaying 0.9 -12.599 46.5 0.9 -12.599 climbing 0.19214 -12.5991 scores are 0.9 -12.5993 between, 0.9 -12.5993 markovitch, 0.9 -12.5993 av 0.9 -12.5993 1981; 0.9 -12.5993 io 0.9 -12.5993 teacher, 0.9 -12.5993 leave-one-out 0.9 -12.5993 impulsive 0.485714 -12.5993 committee 0.705882 -12.5994 using 10-fold 0.608696 -12.5994 it remains 0.833333 -12.5996 explanation, 0.9 -12.5997 succession 0.9 -12.5997 46.4 0.833333 -12.5999 eliminated from 0.833333 -12.5999 275 0.733333 -12.6 transfer learning 0.9 -12.6 tutor’s 0.9 -12.6 sequence-based 0.9 -12.6004 minimal subtree 0.9 -12.6004 improvement: 0.9 -12.6004 far too 0.9 -12.6004 deviates 0.9 -12.6004 museum 0.666667 -12.6004 cssr 0.733333 -12.6005 79.6 0.619048 -12.6006 record. 0.9 -12.6007 less informative 0.9 -12.6007 correct) 0.9 -12.6007 janyce wiebe, 0.9 -12.6007 intersect 0.733333 -12.6008 %, 0.9 -12.6011 multi-dimensional 0.9 -12.6011 r’s 0.9 -12.6011 15,000 0.9 -12.6011 non-isomorphism 0.9 -12.6011 0.7% 0.619048 -12.6011 bed 0.370968 -12.6012 pearson’s 0.9 -12.6014 monotone, 0.359375 -12.6016 daniel 0.9 -12.6018 delimited 0.9 -12.6018 pecina 0.9 -12.6018 42.9 0.9 -12.6018 52.6 0.9 -12.6018 mh 0.9 -12.6018 aviation 0.5 -12.6018 paris 0.9 -12.6021 bender 0.9 -12.6021 npov 0.9 -12.6021 maxlanguage 0.9 -12.6021 (null = 0.733333 -12.6022 bright 0.833333 -12.6022 bcpo 0.9 -12.6025 e2, 0.9 -12.6025 16.5 0.9 -12.6025 conll’09 0.9 -12.6025 396 0.9 -12.6025 attract 0.9 -12.6025 vso 0.9 -12.6025 so(tp) 0.301075 -12.6025 view, 0.833333 -12.6025 implementations, 0.9 -12.6028 equivalently, 0.9 -12.6028 over, 0.9 -12.6028 46.8 0.65 -12.6029 crude 0.9 -12.6032 (teh, 0.65 -12.6032 may differ 0.9 -12.6035 boulder, 0.9 -12.6035 71.7 0.9 -12.6035 tricks 0.9 -12.6035 89% 0.9 -12.6035 2010)) 0.833333 -12.6038 an underspecified 0.5 -12.6038 we can conclude 0.9 -12.6039 unsurprisingly, 0.9 -12.6039 47.6 0.9 -12.6039 newswire, 0.9 -12.6039 subsumed 0.9 -12.6039 inflection, 0.583333 -12.6041 is located 0.9 -12.6042 micro-blogs 0.9 -12.6042 opponent 0.9 -12.6042 briscoe 0.705882 -12.6043 jointly. 0.259843 -12.6045 when they 0.9 -12.6046 determined, 0.9 -12.6046 basics 0.705882 -12.6046 zodiac-408 0.733333 -12.6048 two modules 0.9 -12.6049 deterioration 0.9 -12.6049 confusions 0.9 -12.6049 buy, 0.9 -12.6049 graphic, 0.9 -12.6049 gap, 0.9 -12.6049 specialization 0.9 -12.6053 (refer 0.9 -12.6053 underlies 0.9 -12.6053 (xia 0.9 -12.6056 factoring 0.9 -12.6056 conn 0.9 -12.6056 effectively, 0.9 -12.6056 (1980) 0.147208 -12.6056 reason 0.9 -12.606 retrieved, 0.9 -12.606 54.2 0.9 -12.606 fillmore 0.9 -12.606 idiosyncrasies 0.9 -12.606 searle’s 0.666667 -12.606 helpful discussions 0.9 -12.6063 metadata, 0.9 -12.6063 feature vectors, 0.65 -12.6065 title. 0.0972376 -12.6066 effect 0.9 -12.6067 botstraping 0.9 -12.6067 62.50 0.9 -12.6067 briefly, 0.9 -12.6067 91.8 0.9 -12.6067 great potential 0.9 -12.6067 3note that 0.9 -12.607 algorithmically 0.9 -12.607 uszkoreit 0.583333 -12.607 degree, 0.833333 -12.607 buildings 0.733333 -12.6071 consecutive words 0.705882 -12.6071 revilp 0.425532 -12.6073 using a single 0.9 -12.6074 differentiates 0.9 -12.6074 pcfg-la 0.9 -12.6074 504 0.555556 -12.6075 (moschitti, 0.65 -12.6078 4.1.2 0.733333 -12.6079 derivation trees 0.9 -12.6081 13.6 0.9 -12.6081 combinators 0.9 -12.6081 adversative 0.9 -12.6081 5.2, 0.44186 -12.6083 (lafferty 0.65 -12.6083 rewritten as 0.9 -12.6084 3.7% 0.9 -12.6084 performer 0.5 -12.6084 this phenomenon 0.45 -12.6086 evaluated. 0.9 -12.6088 marcu. 0.9 -12.6088 pl-mrf 0.733333 -12.6091 sara 0.9 -12.6091 aw 0.9 -12.6091 arguing 0.9 -12.6091 cepstral 0.535714 -12.6094 respectively; 0.535714 -12.6094 were removed 0.9 -12.6095 website, 0.9 -12.6095 morphological properties 0.9 -12.6095 lunch 0.485714 -12.6095 singular value 0.9 -12.6098 multi-layered 0.9 -12.6098 biographical 0.65 -12.6098 is performed using 0.535714 -12.61 prevalence 0.0881543 -12.6101 instead 0.733333 -12.6102 jp 0.9 -12.6102 382 0.9 -12.6102 initiative, 0.9 -12.6102 repertoire 0.9 -12.6102 41.1 0.9 -12.6102 (nlp), 0.9 -12.6102 believe, 0.9 -12.6102 divisions 0.9 -12.6102 questionable 0.9 -12.6102 dyslexia 0.9 -12.6102 operates over 0.9 -12.6102 40.8 0.9 -12.6102 microbiology 0.619048 -12.6103 wikipedia page 0.9 -12.6105 six, 0.9 -12.6105 smote 0.535714 -12.6107 enough. 0.666667 -12.6107 downspeak 0.733333 -12.6108 missed by 0.9 -12.6109 hvs model 0.9 -12.6109 (plakias 0.9 -12.6109 ψ, 0.9 -12.6113 casing 0.9 -12.6113 postorder 0.9 -12.6113 42.2 0.9 -12.6113 recognise 0.9 -12.6113 380 0.733333 -12.6114 sandeep 0.733333 -12.6114 wasn’t 0.27193 -12.6114 first two 0.9 -12.6116 23.8 0.65 -12.6116 modeled by 0.833333 -12.6116 mpf 0.191304 -12.6117 2005), 0.9 -12.612 1.7% 0.619048 -12.6122 these. 0.284314 -12.6122 semantic analysis 0.833333 -12.6122 negative evidence 0.833333 -12.6122 cnl 0.485714 -12.6122 sys sent 0.9 -12.6123 mainew 0.9 -12.6123 set), 0.9 -12.6123 translate, 0.583333 -12.6123 extracted rules 0.9 -12.6127 recompute 0.9 -12.6127 elect 0.9 -12.6127 wmt-large 0.9 -12.6127 imaginative 0.5 -12.6127 vbp 0.0174377 -12.6129 on 0.9 -12.613 cars, 0.666667 -12.6131 88.7 0.666667 -12.6131 2003), which 0.833333 -12.6132 das, 0.342857 -12.6133 without using 0.9 -12.6134 disco 0.9 -12.6134 invert 0.833333 -12.6135 data sharing 0.9 -12.6137 paraphrasable 0.21547 -12.6137 to extend 0.619048 -12.6139 family, 0.65 -12.6139 negative impact 0.9 -12.6141 ct, 0.9 -12.6141 semposes 0.9 -12.6141 agreed, 0.9 -12.6141 either, 0.833333 -12.6142 transcriptions, 0.44186 -12.6145 unclear 0.9 -12.6148 distortion, 0.9 -12.6148 provider 0.9 -12.6148 dt, 0.9 -12.6148 p2, 0.666667 -12.6149 probabilistic approach 0.666667 -12.6149 o(n3) 0.9 -12.6151 infinitival 0.9 -12.6151 516 0.705882 -12.6151 lfg 0.9 -12.6155 323 0.9 -12.6155 29.5 0.9 -12.6155 burn-in 0.9 -12.6155 judge, 0.9 -12.6155 cross-product 0.9 -12.6155 modern-day 0.733333 -12.6157 transliteration system 0.733333 -12.6157 table 2: example 0.9 -12.6158 44.4 0.9 -12.6158 p-kla 0.9 -12.6158 journalistic 0.9 -12.6158 finitely 0.9 -12.6158 zhai 0.9 -12.6158 13.7 0.9 -12.6162 bep 0.9 -12.6162 (stolcke 0.9 -12.6162 1998, 0.9 -12.6165 t-h 0.9 -12.6165 dataset: 0.9 -12.6165 88.1 0.9 -12.6165 compressor 0.666667 -12.6165 2007a; 0.833333 -12.6165 can rewrite 0.9 -12.6169 co-hyponyms 0.9 -12.6169 annotate, 0.9 -12.6169 human-produced 0.9 -12.6169 joint, 0.516129 -12.6169 propagated 0.9 -12.6173 21218, 0.9 -12.6173 paper), 0.833333 -12.6175 section 2 reviews 0.9 -12.6176 (suchanek 0.9 -12.6176 exploratory study 0.9 -12.6176 contextsensitive 0.733333 -12.6177 mproductdisplay 0.733333 -12.6177 angeles 0.833333 -12.6178 rhymes 0.733333 -12.618 keller, 0.9 -12.618 families, 0.9 -12.618 flower 0.9 -12.618 injecting 0.9 -12.618 already know 0.9 -12.618 wrt 0.9 -12.618 submodular, 0.9 -12.618 degre 0.9 -12.6183 structure) 0.9 -12.6183 invite 0.9 -12.6183 right) 0.9 -12.6183 omd 0.461538 -12.6183 game. 0.555556 -12.6184 (% 0.9 -12.6187 ranked, 0.9 -12.6187 infinitely 0.9 -12.6187 4note that 0.769231 -12.6187 connectivity 0.5 -12.6188 in-depth 0.733333 -12.6189 diane 0.9 -12.619 97.0 0.9 -12.619 fish, 0.9 -12.619 highly predictive 0.769231 -12.619 the  0.666667 -12.6191 blue, 0.833333 -12.6191 56.2 0.65 -12.6192 seeded 0.209424 -12.6193 tried 0.9 -12.6194 historian 0.9 -12.6194 dq 0.9 -12.6194 (a, 0.9 -12.6194 covered, 0.666667 -12.6194 corpus creation 0.65 -12.6195 }. 0.666667 -12.6196 coreference links 0.535714 -12.6197 candidate translations 0.9 -12.6197 occupies 0.9 -12.6197 h3 0.9 -12.6197 rule) 0.5 -12.62 ¸@ 0.9 -12.6201 taste 0.9 -12.6201 celldesigner 0.9 -12.6201 32% 0.666667 -12.6202 specialists 0.9 -12.6204 atoms 0.9 -12.6204 disease, 0.9 -12.6204 mechanics 0.9 -12.6204 frequency) 0.9 -12.6204 2.14 0.365079 -12.6204 many of these 0.583333 -12.6205 fires 0.20398 -12.6205 gigaword 0.9 -12.6208 sads 0.297872 -12.6208 (2009). 0.833333 -12.6211 edge detection 0.9 -12.6212 sag, 0.9 -12.6212 1.58 0.9 -12.6212 361 0.9 -12.6212 indifferent 0.733333 -12.6215 pair: 0.9 -12.6215 lang, 0.9 -12.6215 beside 0.769231 -12.6215 multi-taskingdialogues 0.769231 -12.6215 77.7 0.833333 -12.6217 gimpel 0.9 -12.6219 -0.03 0.9 -12.6219 has-property 0.583333 -12.6219 biblical 0.472222 -12.6221 plate 0.472222 -12.6221 height 0.9 -12.6222 signature modules. 0.9 -12.6222 contradictions 0.9 -12.6222 (1988) 0.365079 -12.6225 interacting 0.9 -12.6226 fifa 0.9 -12.6226 barak 0.9 -12.6226 1975; 0.306818 -12.6226 components. 0.666667 -12.6228 2010b), 0.9 -12.6229 moment, 0.9 -12.6229 r) 0.9 -12.6229 bioframenet 0.9 -12.6233 italicized 0.535714 -12.6235 genome 0.229299 -12.6236 algorithms. 0.9 -12.6237 f-b 0.9 -12.6237 suports 0.9 -12.6237 0.829 0.9 -12.6237 zeit 0.9 -12.6237 199 0.9 -12.6237 pressing 0.9 -12.6237 idioms, 0.22093 -12.6237 expected to 0.833333 -12.6237 consistently outperforms 0.9 -12.624 main ideas 0.9 -12.624 schwartz, 0.733333 -12.6241 )( 0.485714 -12.6243 documents into 0.485714 -12.6243 wild 0.9 -12.6244 maltparser, 0.9 -12.6244 player, 0.236486 -12.6244 we applied 0.833333 -12.6244 sentimentality 0.619048 -12.6246 ii. 0.9 -12.6247 (negative 0.9 -12.6247 nmfcontext 0.9 -12.6247 outweigh 0.173285 -12.6248 we did 0.833333 -12.625 ws 0.9 -12.6251 berkeleylm 0.9 -12.6251 40,000 0.9 -12.6251 lexicon) 0.5 -12.6251 multiple-beam 0.9 -12.6254 end-user 0.9 -12.6254 shuffle 0.9 -12.6254 itm 0.769231 -12.6254 those whose 0.833333 -12.6254 nvda 0.733333 -12.6255 exactly two 0.769231 -12.6257 (brent, 0.9 -12.6258 11.5 0.9 -12.6258 bnews 0.294737 -12.6259 resolution. 0.769231 -12.626 (hereafter, 0.733333 -12.6261 manual evaluation. 0.9 -12.6261 2008a) 0.769231 -12.6263 moat 0.833333 -12.6264 non-nil 0.535714 -12.6265 a114 0.9 -12.6265 chan, 0.9 -12.6265 exchange, 0.324675 -12.6267 random fields 0.9 -12.6269 asher 0.9 -12.6269 tig 0.9 -12.6269 2nd-tier 0.9 -12.6269 approval 0.9 -12.6269 coin 0.769231 -12.627 8.6 0.9 -12.6272 preorder 0.9 -12.6272 shifter 0.733333 -12.6273 banner 0.516129 -12.6273 ciphertext 0.733333 -12.6276 micro-blogging 0.733333 -12.6276 9.1 0.9 -12.6276 (again 0.9 -12.6276 third row 0.619048 -12.6276 eskenazi, 0.833333 -12.6277 xl 0.9 -12.6279 tutoring, 0.9 -12.6279 e-mail 0.9 -12.6279 disorders 0.9 -12.6279 organisers 0.769231 -12.6279 (compared 0.583333 -12.6279 standard spanish 0.485714 -12.6279 n=1 0.291667 -12.6279 sciences 0.102922 -12.628 build 0.269565 -12.628 labels are 0.253968 -12.628 is therefore 0.431818 -12.6283 a central 0.9 -12.6283 contracted 0.9 -12.6283 c 0.9 -12.6283 subsampled 0.9 -12.6283 rival 0.9 -12.6283 hwang 0.9 -12.6283 template, 0.666667 -12.6283 67.4 0.9 -12.6287 elhadad’s 0.9 -12.6287 1987) 0.9 -12.6287 φt 0.619048 -12.6288 cross-cultural 0.516129 -12.6288 beijing 0.9 -12.629 ρ, 0.9 -12.629 marginalization 0.149333 -12.629 2003). 0.769231 -12.6291 update, 0.114241 -12.6292 used as 0.9 -12.6294 charles, 0.9 -12.6294 explored, 0.769231 -12.6294 pt, 0.769231 -12.6294 valid, 0.9 -12.6297 formalisms, 0.9 -12.6297 co. 0.9 -12.6297 america, 0.9 -12.6297 flipping 0.9 -12.6297 44.8 0.769231 -12.63 de-en es-en 0.555556 -12.63 gmm 0.583333 -12.63 substitute for 0.9 -12.6301 680 0.9 -12.6301 anyway 0.9 -12.6301 song, 0.9 -12.6301 −1, 0.9 -12.6301 315 0.9 -12.6301 original, 0.833333 -12.6303 aprime 0.9 -12.6304 arg2, 0.9 -12.6304 sharply 0.388889 -12.6304 for example, consider 0.733333 -12.6305 figure. 0.733333 -12.6308 word boundaries. 0.9 -12.6308 favre 0.9 -12.6308 counterpart, 0.9 -12.6308 medium, 0.215909 -12.6309 to assign 0.9 -12.6312 487 0.9 -12.6312 peculiar 0.733333 -12.6314 ae 0.56 -12.6314 sweden 0.9 -12.6315 two: 0.9 -12.6315 knowledge-driven 0.733333 -12.6317 semeval-2010 task 0.138889 -12.6318 we find 0.9 -12.6319 dyads 0.9 -12.6319 multiplication, 0.9 -12.6319 interestingness 0.9 -12.6319 test-set 0.769231 -12.6319 90%. 0.733333 -12.632 uniondisplay 0.666667 -12.6321 uttered by 0.9 -12.6322 convention, 0.769231 -12.6325 58.3 0.9 -12.6326 caught 0.833333 -12.6327 from: 0.185654 -12.6329 noted 0.9 -12.633 weka’s 0.9 -12.633 rambow 0.9 -12.6333 r(%) 0.9 -12.6333 sale 0.9 -12.6333 philosophy 0.9 -12.6333 verbs; 0.833333 -12.6333 explicit. 0.833333 -12.6333 stable. 0.833333 -12.6333 nullcc =+ 0.833333 -12.6333 society. 0.833333 -12.6333 home. 0.833333 -12.6333 .74 0.9 -12.6337 psychologists 0.9 -12.6337 softly 0.9 -12.6337 ttr 0.9 -12.6337 nict 0.9 -12.6337 93.6 0.666667 -12.6339 intrusion 0.9 -12.634 sdss 0.9 -12.634 (nlg) 0.9 -12.634 amendment 0.9 -12.634 p) 0.769231 -12.634 ebay 0.197183 -12.634 7: 0.619048 -12.6342 child-directed 0.619048 -12.6342 they found that 0.472222 -12.6343 relevant documents 0.9 -12.6344 non-compositionality 0.9 -12.6344 rp 0.9 -12.6344 haptic 0.769231 -12.6346 formosan 0.666667 -12.6347 word overlap 0.422222 -12.6347 semeval-2010 0.9 -12.6348 speedups 0.9 -12.6351 manager, 0.9 -12.6351 radical 0.733333 -12.6352 saffran 0.666667 -12.6353 adjp 0.9 -12.6355 count-based 0.9 -12.6355 f1% 0.9 -12.6358 o(nlogn) 0.9 -12.6358 nouns; 0.9 -12.6358 ostendorf 0.9 -12.6358 lab, 0.9 -12.6358 coefficients, 0.5 -12.6359  0.666667 -12.6361 generating new 0.9 -12.6362 (to be 0.9 -12.6362 understands 0.9 -12.6362 consciously 0.9 -12.6362 (esuli 0.769231 -12.6362 hb 0.769231 -12.6362 to fuse 0.9 -12.6366 (p,q) 0.9 -12.6366 micr 0.9 -12.6366 w′ 0.9 -12.6369 possible: 0.9 -12.6369 81.7 0.9 -12.6369 marathi 0.516129 -12.637 the last column 0.9 -12.6373 integration, 0.769231 -12.6374 (erk 0.733333 -12.6376 sent-par 0.0870761 -12.6376 define 0.9 -12.6377 and qun 0.666667 -12.6379 strapparava, 0.9 -12.638 286 0.9 -12.638 generator, 0.431818 -12.6383 amazon’s mechanical 0.769231 -12.6383 64.3 0.9 -12.6384 boat 0.0488439 -12.6384 source 0.4 -12.6385 all cases, 0.733333 -12.6387 anonymized 0.9 -12.6387 64% 0.9 -12.6387 abuse 0.9 -12.6391 n+ 0.9 -12.6391 presentation, 0.516129 -12.6391 stanford, 0.733333 -12.6393 jiang, 0.221557 -12.6393 this feature 0.408163 -12.6394 search algorithm 0.9 -12.6395 (mehdad 0.9 -12.6395 performance; 0.9 -12.6395 3; 0.9 -12.6395 idealized 0.769231 -12.6396 cows 0.769231 -12.6396 kyoto university 0.733333 -12.6399 (8 0.769231 -12.6399 arises from 0.9 -12.6402 icwr 0.9 -12.6402 soup 0.9 -12.6402 10.9 0.9 -12.6405 uc50q 0.666667 -12.6406 columbus, 0.9 -12.6409 skip-1 0.769231 -12.6411 pushdown 0.516129 -12.6412 measured using 0.9 -12.6413 19% 0.9 -12.6413 monotonous 0.9 -12.6413 m: 0.769231 -12.6414 table 2 gives 0.56 -12.6415 ˆe 0.9 -12.6416 markedness 0.583333 -12.6417 30 35 0.9 -12.642 lscript, 0.619048 -12.642 .28 0.733333 -12.6423 cookbook 0.9 -12.6424 going beyond 0.9 -12.6424 resorted 0.9 -12.6424 (and, 0.9 -12.6424 nulld 0.9 -12.6424 clever 0.769231 -12.6424 beam, 0.769231 -12.6424 21.6 0.9 -12.6427 (209) 0.9 -12.6427 ∆t 0.9 -12.6427 spotted 0.9 -12.6427 tamil, 0.9 -12.6427 b), 0.9 -12.6427 2.2), 0.360656 -12.6427 decompose 0.769231 -12.643 unify 0.360656 -12.643 are drawn 0.9 -12.6431 generalizable 0.9 -12.6431 \ 0.251969 -12.6433 senses. 0.9 -12.6435 pprefixg 0.9 -12.6435 apartment 0.9 -12.6435 alternative analyses 0.9 -12.6438 alignment; 0.9 -12.6438 german–english 0.6875 -12.644 hmm-lda 0.733333 -12.6441 (2007)) 0.0655824 -12.6441 the following 0.9 -12.6442 (along 0.9 -12.6442 proving 0.9 -12.6442 tuna 0.769231 -12.6442 streams, 0.9 -12.6445 ˆes 0.9 -12.6449 watching 0.9 -12.6449 borrowed 0.666667 -12.6452 hhmm 0.9 -12.6453 1/2 0.9 -12.6453 techniques: 0.9 -12.6453 break, 0.6875 -12.6454 92.0 0.769231 -12.6455 server-side 0.9 -12.6456 coreference-annotated 0.619048 -12.6457 rahman and 0.769231 -12.6458 mislabeled 0.9 -12.646 florian 0.9 -12.646 mike 0.167808 -12.646 properties of 0.9 -12.6464 two-component 0.9 -12.6464 subjectivity classification 0.9 -12.6464 dat 0.9 -12.6464 hyphens 0.9 -12.6464 boston, 0.769231 -12.6464 present an unsupervised 0.9 -12.6467 chambers 0.769231 -12.647 parsing algorithms. 0.9 -12.6475 (traum 0.9 -12.6475 complicates 0.253968 -12.6476 this type 0.769231 -12.6477 59% 0.9 -12.6478 europe 0.9 -12.6478 individual, 0.9 -12.6478 comparatives 0.081255 -12.6479 computational linguistics 0.6875 -12.6479 frequently, 0.9 -12.6482 predictor, 0.9 -12.6486 timestamp 0.9 -12.6486 anotators 0.535714 -12.6489 ensured 0.9 -12.6489 present study, 0.9 -12.6489 categorization, 0.769231 -12.6489 licenses 0.9 -12.6493 25.7 0.9 -12.6493 tabulate 0.484848 -12.6494 ann 0.9 -12.6497 (under 0.9 -12.6497 generic, 0.9 -12.6497 93.16 0.666667 -12.6498 hls 0.5 -12.6499 listings 0.333333 -12.65 performed. 0.9 -12.6504 corroborate 0.9 -12.6508 projectivity 0.631579 -12.651 researchers, 0.9 -12.6511 erp 0.9 -12.6511 §4 0.769231 -12.6511 urban 0.366667 -12.6511 report on 0.216374 -12.6511 trained with 0.631579 -12.6513 lwlm 0.9 -12.6515 non-specific 0.9 -12.6515 utilization 0.9 -12.6515 relevance, 0.9 -12.6515 alchemy 0.9 -12.6515 rat 0.9 -12.6519 query-based 0.9 -12.6519 word-predictor 0.818182 -12.6521 tong 0.818182 -12.6521 converge, 0.818182 -12.6521 discriminatory 0.56 -12.6522 becomes: 0.9 -12.6526 f1measure 0.9 -12.6526 postcondition 0.6875 -12.6526 easy-first 0.818182 -12.6527 kind, 0.818182 -12.6527 surge 0.818182 -12.6527 h(x) 0.631579 -12.6528 context-vector 0.631579 -12.6528 exposure to 0.590909 -12.6529 readers. 0.733333 -12.653 recent works 0.9 -12.653 denotations 0.9 -12.653 -5 0.769231 -12.653 hassan 0.9 -12.6533 antonyms, 0.9 -12.6533 syntbip 0.9 -12.6533 s,t 0.9 -12.6533 slave 0.381818 -12.6533 issue. 0.818182 -12.6534 displayed on 0.818182 -12.6534 afterwards 0.818182 -12.6534 continually 0.769231 -12.6536 69.0 0.769231 -12.6536 lasts 0.9 -12.6537 tactical 0.9 -12.6537 we’ll 0.9 -12.6537 right-corner 0.818182 -12.6537 commits 0.818182 -12.6537 n-1 0.666667 -12.6539 chance, 0.256198 -12.6539 minimize 0.6875 -12.654 (model 0.9 -12.6541 forget 0.9 -12.6541 cont 0.818182 -12.6541 brazil 0.818182 -12.6541 athletes 0.818182 -12.6541 them: 0.769231 -12.6542 hansard 0.9 -12.6544 neighbouring 0.818182 -12.6544 band 0.733333 -12.6545 uu 0.9 -12.6548 spoke 0.296703 -12.6548 performed on 0.818182 -12.6554 no-arc 0.818182 -12.6554 verb-specific 0.818182 -12.6557 originated 0.769231 -12.6558 formants 0.9 -12.6559 schlangen 0.9 -12.6559 290 0.733333 -12.656 italian. 0.818182 -12.6561 nj, 0.818182 -12.6561 nb-adapted 0.818182 -12.6561 203 0.818182 -12.6561 two-tiered 0.9 -12.6563 gprefix 0.9 -12.6563 218 0.56 -12.6563 brody 0.769231 -12.6564 crisis, 0.818182 -12.6564 c++ 0.818182 -12.6564 gaussier, 0.733333 -12.6566 transliterations. 0.9 -12.6566 ruled 0.9 -12.6566 ’6 – 0.9 -12.6566 (part 0.818182 -12.6568 “with 0.590909 -12.6568 competence 0.0266538 -12.6568 model 0.9 -12.657 unaware of 0.9 -12.657 interactional 0.818182 -12.6571 rouge-l 0.818182 -12.6571 stamp 0.818182 -12.6571 composed rules. 0.9 -12.6574 jaws 0.9 -12.6574 (here, 0.769231 -12.6574 90s 0.818182 -12.6574 input/output 0.131078 -12.6576 focused 0.9 -12.6577 finder 0.9 -12.6577 iraq 0.818182 -12.6578 x1,x2 0.428571 -12.6579 target word. 0.769231 -12.658 73.2 0.769231 -12.658 dale 0.769231 -12.658  0.9 -12.6581 ice-ireland 0.9 -12.6581 normalizes 0.9 -12.6581 before applying 0.818182 -12.6581 api, 0.6875 -12.6583 softmax 0.9 -12.6585 om 0.9 -12.6585 alvis 0.354839 -12.6586 (wang 0.413043 -12.6587 were manually 0.631579 -12.6587 we proceed 0.517241 -12.6587 snow 0.9 -12.6588 formerly 0.9 -12.6588 analysts 0.9 -12.6588 richardson, 0.818182 -12.6588 cloa 0.619048 -12.6588 of maryland 0.818182 -12.6591 electronically 0.9 -12.6592 69.4 0.9 -12.6592 ,) 0.769231 -12.6593 303 0.375 -12.6593 method proposed 0.818182 -12.6594 javascript 0.222222 -12.6594 al., 2007), 0.9 -12.6596 strategy: 0.9 -12.66 c), 0.9 -12.66 läsbart 0.381818 -12.6601 par 0.769231 -12.6602 would enable 0.9 -12.6603 i-qry 0.818182 -12.6604 partial match 0.769231 -12.6605 instructional 0.360656 -12.6605 two systems 0.9 -12.6607 x′ 0.9 -12.6607 14th 0.9 -12.6611 ftib 0.316456 -12.6611 zhang and clark 0.56 -12.6611 browser 0.769231 -12.6612 inserts 0.631579 -12.6613 aids 0.9 -12.6614 +, 0.9 -12.6614 divrank 0.392157 -12.6614 we also used 0.413043 -12.6614 obtained. 0.254098 -12.6614 state. 0.769231 -12.6615 designers 0.818182 -12.6615 instruments 0.818182 -12.6618 grand 0.254098 -12.6619 target words 0.413043 -12.662 litman, 0.818182 -12.6621 complemented 0.16129 -12.6621 to address 0.769231 -12.6624 inference: 0.818182 -12.6625 sub-corpora 0.818182 -12.6625 pv 0.818182 -12.6628 multi-label 0.142494 -12.6629 necessarily 0.631579 -12.6629 n: 0.818182 -12.6631 context-specific 0.484848 -12.6631 ¡ 0.733333 -12.6632 to realize 0.733333 -12.6635 62.8 0.459459 -12.6638 this constraint 0.769231 -12.6643 tony 0.733333 -12.6644 rebecca 0.368421 -12.6644 it will be 0.818182 -12.6645 exit 0.631579 -12.6645 the former, 0.177165 -12.6646 intuition 0.818182 -12.6648 couldn’t 0.5 -12.6648 also add 0.733333 -12.665 correctly labeled 0.733333 -12.665 word aligner 0.733333 -12.665 newstest2008 0.316456 -12.6652 were annotated 0.818182 -12.6658 1-4 0.733333 -12.6659 (sinha 0.769231 -12.6659 saarland 0.666667 -12.6659 thread, 0.206349 -12.6661 an event 0.538462 -12.6661 primarily on 0.769231 -12.6662 khapra 0.769231 -12.6662 ept 0.818182 -12.6662 snowball 0.263158 -12.6664 discusses 0.818182 -12.6665 τ, 0.818182 -12.6665 distributions: 0.590909 -12.6667 presidential 0.175781 -12.6668 al., 2006; 0.818182 -12.6669 222 0.818182 -12.6672 export 0.619048 -12.6673 venugopal 0.769231 -12.6675 wsj sections 0.769231 -12.6675 bits, 0.818182 -12.6675 forests, 0.6875 -12.6676 pedagogy 0.818182 -12.6679 2.1.2 0.818182 -12.6682 backwards 0.818182 -12.6682 takipi 0.818182 -12.6682 (ding 0.666667 -12.6684 ich 0.324324 -12.6686 validate 0.348485 -12.6687 papers. 0.666667 -12.6687 standard n-gram 0.818182 -12.6689 latitude 0.0821338 -12.6689 difference 0.56 -12.669 joint distribution 0.769231 -12.6691 contracts 0.818182 -12.6692 lscripti 0.818182 -12.6692 gloss, 0.6875 -12.6693 combining them 0.769231 -12.6694 pre-terminal 0.290323 -12.6696 it is worth 0.6875 -12.6696 eventuality 0.447368 -12.6699 submit 0.6875 -12.6699 13, 0.818182 -12.6702 earth’s 0.246154 -12.6703 computes 0.354839 -12.6706 applicability 0.184549 -12.6706 bionlp 0.818182 -12.6709 organizational 0.818182 -12.6709 renamed 0.818182 -12.6709 f4 0.818182 -12.6709 match: 0.23913 -12.6709 view of 0.666667 -12.6712 λm 0.769231 -12.6713 +2 0.818182 -12.6713 party, 0.818182 -12.6713 σ∗, 0.6875 -12.6713 management, 0.384615 -12.6715 lexicons. 0.375 -12.6715 be helpful 0.538462 -12.6726 classification tasks 0.6875 -12.6727 ho 0.769231 -12.6729 traditional approaches 0.818182 -12.673 352 0.818182 -12.673 di, 0.6875 -12.673 two-sided 0.769231 -12.6733 ve, 0.769231 -12.6736 np) 0.818182 -12.6736 challenging, 0.338235 -12.6737 modules. 0.254098 -12.6737 these models 0.666667 -12.6739 p.o. 0.666667 -12.6742 lin. 1 -12.6742 capitalized. 1 -12.6742 part-of-speech. 1 -12.6742 training). 1 -12.6742 cue. 1 -12.6742 appendix). 1 -12.6742 listener. 1 -12.6742 comments: 1 -12.6742 inst. 1 -12.6742 intuitive. 1 -12.6742 sampled. 1 -12.6742 (f). 1 -12.6742 ch. 1 -12.6742 bona 1 -12.6742 n). 1 -12.6742 (1989). 1 -12.6742 describe. 1 -12.6742 you? 1 -12.6742 d). 1 -12.6742 paris sud 1 -12.6742 therein. 1 -12.6742 cfs 1 -12.6742 bounds. 1 -12.6742 senti-mentality 1 -12.6742 finding. 1 -12.6742 rescoring. 1 -12.6742 recommendation. 1 -12.6742 encoding. 1 -12.6742 matters. 1 -12.6742 arbitrary. 1 -12.6742 trie. 1 -12.6742 (ilp). 1 -12.6742 twofold. 1 -12.6742 yield’s 1 -12.6742 unique. 1 -12.6742 freq. 1 -12.6742 has. 1 -12.6742 null; 1 -12.6742 james pustejovsky 1 -12.6742 etl committee 1 -12.6742 tomorrow morning 1 -12.6742 pres). 1 -12.6742 insufficient. 1 -12.6742 debates. 1 -12.6742 global/local 1 -12.6742 impossible. 1 -12.6742 implications. 1 -12.6742 invaluable 1 -12.6742 plot. 1 -12.6742 recently. 1 -12.6742 expert’s 1 -12.6742 costly, 1 -12.6742 already. 1 -12.6742 many. 1 -12.6742 browser. 1 -12.6742 ground. 1 -12.6742 appreciable 1 -12.6742 s49s48s48s48 s50s48s48s48 1 -12.6742 (kumar et. 1 -12.6742 quaero programme, 1 -12.6742 (pos). 1 -12.6742 kluwer academic publishers. 1 -12.6742 wsi/wsd 1 -12.6742 assertions. 1 -12.6742 npk. 1 -12.6742 established. 1 -12.6742 p2. 1 -12.6742 square brackets 1 -12.6742 narratives. 1 -12.6742 (vickrey 1 -12.6742 bing liu. 1 -12.6742 t3. 1 -12.6742 title-based 1 -12.6742 histograms. 1 -12.6742 mixed. 1 -12.6742 chair. 1 -12.6742 mlns 1 -12.6742 authority. 1 -12.6742 coreferences. 1 -12.6742 complications 1 -12.6742 re-sampling 1 -12.6742 california. 1 -12.6742 hypernyms. 1 -12.6742 automaton. 1 -12.6742 room. 1 -12.6742 word–colour associations 1 -12.6742 lexicalization. 1 -12.6742 bo pang 1 -12.6742 lscript. 1 -12.6742 18. 1 -12.6742 pathway. 1 -12.6742 (+0.8) 1 -12.6742 sarcasm. 1 -12.6742 referees. 1 -12.6742 availability, 1 -12.6742 preferable. 1 -12.6742 n+1. 1 -12.6742 entity). 1 -12.6742 discriminative. 1 -12.6742 nucleus. 1 -12.6742 taking. 1 -12.6742 videos. 1 -12.6742 5%. 1 -12.6742 wrt. 1 -12.6742 domain-adapted 1 -12.6742 transport. 1 -12.6742 km. 1 -12.6742 stepping 1 -12.6742 nr. 1 -12.6742 lau 1 -12.6742 lillian lee. 1 -12.6742 vice versa), 1 -12.6742 205). 1 -12.6742 unambiguous. 1 -12.6742 20%. 1 -12.6742 const. 1 -12.6742 overview. 1 -12.6742 arise. 1 -12.6742 ‘fell’ 1 -12.6742 ncs. 1 -12.6742 stored. 1 -12.6742 chalenging 1 -12.6742 comprehensible 1 -12.6742 unaligned. 1 -12.6742 off. 1 -12.6742 inter-domain 1 -12.6742 reranker. 1 -12.6742 speed/accuracy 1 -12.6742 consumption. 1 -12.6742 nuclei. 1 -12.6742 75.0% 1 -12.6742 documentation. 1 -12.6742 200. 1 -12.6742 at. 1 -12.6742 relative). 1 -12.6742 both). 1 -12.6742 triangles 1 -12.6742 counted. 1 -12.6742 evaluate, 1 -12.6742 species. 1 -12.6742 (smt), 1 -12.6742 dtks 1 -12.6742 capabilities. 1 -12.6742 http://www. 1 -12.6742 written. 1 -12.6742 da. 1 -12.6742 parameter estimation, 1 -12.6742 } order 1 -12.6742 decrease. 1 -12.6742 finished. 1 -12.6742 course. 1 -12.6742 compensating 1 -12.6742 drops. 1 -12.6742 dev-set 1 -12.6742 environments, 1 -12.6742 eliminated. 1 -12.6742 like. 1 -12.6742 different 1 -12.6742 (1994). 1 -12.6742 authors’ alone. 1 -12.6742 vocalized 1 -12.6742 powerful. 1 -12.6742 ssrs. 1 -12.6742 ?? 1 -12.6742 (e.g, 1 -12.6742 iteratively. 1 -12.6742 12). 1 -12.6742 sentiwordnet. 1 -12.6742 decoupled 1 -12.6742 av. 1 -12.6742 re-training 1 -12.6742 lvcs. 1 -12.6742 shortcomings. 1 -12.6742 informativeness. 1 -12.6742 competitors. 1 -12.6742 multitask 1 -12.6742 inflection. 1 -12.6742 columns). 1 -12.6742 language? 1 -12.6742 signs. 1 -12.6742 solve. 1 -12.6742 psc-based 1 -12.6742 disagree. 1 -12.6742 scarce. 1 -12.6742 opposite. 1 -12.6742 minimized. 1 -12.6742 tighten 1 -12.6742 1b). 1 -12.6742 zhou. 1 -12.6742 clia 1 -12.6742 you deliver 1 -12.6742 rec. 1 -12.6742 makes. 1 -12.6742 command-line 1 -12.6742 l1. 1 -12.6742 n1. 1 -12.6742 mrfs 1 -12.6742 step-by-step 1 -12.6742 translate. 1 -12.6742 breaks. 1 -12.6742 morpheme. 1 -12.6742 last. 1 -12.6742 belongs. 1 -12.6742 dimensionality. 1 -12.6742 regimes. 1 -12.6742 accur acy 1 -12.6742 item-response item 1 -12.6742 permitted. 1 -12.6742 allocation. 1 -12.6742 loop. 1 -12.6742 1.5. 1 -12.6742 determination. 1 -12.6742 paraphrase-based 1 -12.6742 french→english 1 -12.6742 subclasses. 1 -12.6742 origins. 1 -12.6742 subgraph. 1 -12.6742 rus 1 -12.6742 probable. 1 -12.6742 recalculated 1 -12.6742 executed. 1 -12.6742 contiguous. 1 -12.6742 recall). 1 -12.6742 readable. 1 -12.6742 phd thesis, 1 -12.6742 validity. 1 -12.6742 intact. 1 -12.6742 any. 1 -12.6742 labelling. 1 -12.6742 fide 1 -12.6742 rewriting. 1 -12.6742 tease apart 1 -12.6742 color. 1 -12.6742 violated. 1 -12.6742 hrs 1 -12.6742 neighbor. 1 -12.6742 capability. 1 -12.6742 prosody. 1 -12.6742 contradiction. 1 -12.6742 prefixes. 1 -12.6742 formalisms. 1 -12.6742 noted. 1 -12.6742 stsg. 1 -12.6742 recipient. 1 -12.6742 (huang, 1 -12.6742 l1 regularization 1 -12.6742 exclusive. 1 -12.6742 weston, 1 -12.6742 adp det 1 -12.6742 computer. 1 -12.6742 mstparser. 1 -12.6742 spectrum. 1 -12.6742 (burkett 1 -12.6742 other). 1 -12.6742 categorizes 1 -12.6742 pr. 1 -12.6742 1.2. 1 -12.6742 unattached subtrees[] 1 -12.6742 corpora) 1 -12.6742 interact. 1 -12.6742 acronyms. 1 -12.6742 calls. 1 -12.6742 “... 1 -12.6742 newswire. 1 -12.6742 pcfgs. 1 -12.6742 ts. 1 -12.6742 molecular biology 1 -12.6742 tag). 1 -12.6742 footprint. 1 -12.6742 ...... 1 -12.6742 tackles 1 -12.6742 concept–colour 1 -12.6742 21. 1 -12.6742 phrase). 1 -12.6742 rating. 1 -12.6742 darpa communicator 1 -12.6742 generator. 1 -12.6742 symmetric. 1 -12.6742 bin. 1 -12.6742 the,sg 1 -12.6742 bottom. 1 -12.6742 calbc cii 1 -12.6742 arg-comp aux+v 1 -12.6742 fiction. 1 -12.6742 init. 1 -12.6742 tagset. 1 -12.6742 organism. 1 -12.6742 esp. 1 -12.6742 pino 1 -12.6742 (voorhees, 1 -12.6742 feature). 1 -12.6742 path-to-root 1 -12.6742 iii. 1 -12.6742 indirectly. 1 -12.6742 obtain. 1 -12.6742 rapid growth 1 -12.6742 wmt. 1 -12.6742 afr. 1 -12.6742 organizers. 1 -12.6742 1973). 1 -12.6742 sciences. 1 -12.6742 man. 1 -12.6742 free. 1 -12.6742 (lexeme, template) 1 -12.6742 advantages, 1 -12.6742 “similar” 1 -12.6742 monitored 1 -12.6742 displayed. 1 -12.6742 corpus-specific 1 -12.6742 w1. 1 -12.6742 world heritage sites 1 -12.6742 (ii). 1 -12.6742 (null). 1 -12.6742 ) as,t,u,v 1 -12.6742 sebastiani 1 -12.6742 pak 0.818182 -12.6743 nowadays 0.769231 -12.6745 0.008 0.6875 -12.6745 carlson 0.888889 -12.6746 leaves. 0.888889 -12.6746 one). 0.888889 -12.6746 tuple. 0.888889 -12.6746 meanwhile 0.888889 -12.6746 (2009)). 0.888889 -12.6746 (nadeau 0.888889 -12.6746 hashtags. 0.888889 -12.6746 pioneered 0.888889 -12.6746 deletion-based 0.888889 -12.6746 (ter) 0.888889 -12.6746 subsample 0.888889 -12.6746 1965). 0.888889 -12.6746 third party 0.888889 -12.6746 earthquake. 0.888889 -12.6746 oot 0.888889 -12.6746 agents. 0.888889 -12.6746 exclamation marks 0.888889 -12.6746 abs. 0.888889 -12.6746 50.1 0.6875 -12.6747 paraphrase ranking 0.888889 -12.6749 technical report, 0.888889 -12.6749 (bohus 0.888889 -12.6749 co-referring expressions 0.888889 -12.6749 large-scale unannotated 0.769231 -12.6752 mihai 0.888889 -12.6753 hirst, 2006). 0.888889 -12.6753 grow-diag-final-and heuristic 0.818182 -12.6754 “yes” 0.56 -12.6755 al., 1998; 0.6875 -12.6756 58.2 0.888889 -12.6757 (jurafsky 0.888889 -12.6761 despite its simplicity, 0.888889 -12.6761 taken together, 0.888889 -12.6764 coreferent, 0.888889 -12.6764 teams submitted 0.888889 -12.6764 linked, 0.818182 -12.6767 annotation effort. 0.354839 -12.6767 reported by 0.439024 -12.6771 operators. 0.470588 -12.6772 σ∗ 0.888889 -12.6772 (section 4.1). 0.517241 -12.6772 their relative 0.6875 -12.6773 liefern:deliver 0.818182 -12.6774 implausible 0.888889 -12.6775 diversity, 0.888889 -12.6775 objects cannot be 0.888889 -12.6775 segmenter, 0.769231 -12.6778 unigram baseline 0.484848 -12.6778 savings 0.888889 -12.6779 bracketing structures 0.888889 -12.6779 projected production 0.888889 -12.6783 error propagation. 0.888889 -12.6783 virtual prototype 0.769231 -12.6784 continuity 0.769231 -12.6784 hoping 0.818182 -12.6784 operations: 0.888889 -12.6787 nonenglish 0.888889 -12.6787 incongruent 0.818182 -12.6788 preferring 0.888889 -12.679 (briscoe et 0.888889 -12.679 acceptable head-dependents 0.888889 -12.6794 close proximity 0.888889 -12.6794 weka (hall 0.818182 -12.6795 callison-burch, 2005; 0.126789 -12.6796 results on 0.888889 -12.6798 nominal predicates 0.888889 -12.6798 pair) 0.470588 -12.6799 goodman, 0.769231 -12.68 quest 0.413043 -12.68 we formulate 0.232877 -12.6801 al., 2005) 0.211429 -12.6802 classifier, 0.6875 -12.6802 will perform 0.769231 -12.6803 usr 0.194175 -12.6804 easier 0.818182 -12.6805 advertising 0.888889 -12.6805 salience measure 0.102667 -12.6807 represents 0.818182 -12.6808 jim 0.888889 -12.6809 co-training (blum 0.888889 -12.6809 newly created 0.888889 -12.6809 linearity 0.56 -12.6809 was tested 0.277228 -12.6811 this work was supported 0.888889 -12.6813 (venugopal 0.888889 -12.6813 phillips, 0.818182 -12.6815 materials, 0.631579 -12.6815 metrics: 0.769231 -12.6816 equation 7 0.888889 -12.6816 bottlenecks 0.56 -12.6816 line. 0.818182 -12.6819 (100%) 0.888889 -12.682 39.4 0.888889 -12.682 cheung, 0.888889 -12.682 standard melm 0.888889 -12.682 wordnet 3.0. 0.888889 -12.682 0.737 0.888889 -12.682 specificaly 0.888889 -12.682 zuv 0.392157 -12.6821 de-en 0.818182 -12.6822 linear, 0.769231 -12.6823 ks 0.888889 -12.6824 automatically-extracted 0.818182 -12.6825 pk 0.818182 -12.6825 1); 0.818182 -12.6825 actions: 0.590909 -12.6827 basic types 0.888889 -12.6828 misclassifying 0.888889 -12.6828 difficult” 0.888889 -12.6828 2005)), 0.6875 -12.6828 rue 0.242424 -12.6831 is used for 0.888889 -12.6831 yatskar 0.888889 -12.6831 cheap, 0.888889 -12.6831 easier, 0.769231 -12.6832 jrip 0.888889 -12.6835 precision) 0.888889 -12.6835 fit together 0.888889 -12.6835 programming languages, 0.268519 -12.6836 predicates. 0.125749 -12.6839 found that 0.888889 -12.6839 hollingshead 0.6875 -12.684 stereotypes 0.6875 -12.684 learntlow 0.215569 -12.6841 (2010). 0.590909 -12.6842 survival function 0.888889 -12.6843 influence, 0.888889 -12.6843 extraction; 0.888889 -12.6843 garrod, 0.888889 -12.6843 precision-recall curve 0.888889 -12.6843 globally, 0.769231 -12.6845 partly due 0.818182 -12.6846 english-arabic 0.888889 -12.6846 (ueffing 0.888889 -12.6846 strong indicators 0.538462 -12.6846 chance. 0.631579 -12.6847 admissible 0.888889 -12.685 continue, 0.888889 -12.685 (section 5), 0.888889 -12.685 (rozovskaya 0.888889 -12.685 community-based 0.590909 -12.685 one per 0.6875 -12.6851 to attach 0.818182 -12.6853 schmid 0.888889 -12.6854 heeman 0.888889 -12.6854 hybrid n-grams 0.5 -12.6857 concluding 0.769231 -12.6858 infon 0.888889 -12.6858 signatures, 0.888889 -12.6858 gatt 0.439024 -12.6859 the first time 0.818182 -12.686 rose-rank 0.888889 -12.6861 parent, 0.888889 -12.6861 (crego 0.368421 -12.6864 ← 0.769231 -12.6865 gerund 0.888889 -12.6865 cluster’s 0.888889 -12.6865 bertoldi 0.888889 -12.6865 (da) 0.888889 -12.6865 conducted, 0.888889 -12.6869 counterparts, 0.888889 -12.6869 bionlp09st 0.888889 -12.6869 decisive 0.888889 -12.6869 2010)), 0.888889 -12.6869 brute 0.769231 -12.6871 until convergence 0.631579 -12.6871 id task 0.888889 -12.6873 simplification strategies 0.888889 -12.6873 piano 0.888889 -12.6873 manages 0.769231 -12.6874 hp 0.888889 -12.6876 satta, 0.888889 -12.6876 snover 0.888889 -12.6876 english-to-chinese 0.888889 -12.6876 produced, 0.888889 -12.6876 resource-light 0.418605 -12.6877 suggested that 0.565217 -12.6877 bedroom 0.888889 -12.688 compelling 0.888889 -12.688 segmented decoder 0.888889 -12.688 association measures. 0.818182 -12.6881 rest, 0.565217 -12.6882 session. 0.769231 -12.6884 barcelona 0.888889 -12.6884 63.0 0.888889 -12.6884 transducers, 0.888889 -12.6884 participation, 0.484848 -12.6884 our intuition 0.888889 -12.6888 33.6 0.888889 -12.6888 38.4 0.888889 -12.6888 nicely 0.888889 -12.6888 0.335 0.769231 -12.6891 normalizer 0.565217 -12.6891 kok 0.213018 -12.6893 is calculated 0.484848 -12.6894 noise model 0.117544 -12.6894 implementation 0.888889 -12.6895 canceled 0.888889 -12.6895 continuations 0.395833 -12.6899 can refer 0.888889 -12.6899 goldberg, 0.888889 -12.6899 q+l+g 0.888889 -12.6899 question retrieval. 0.6875 -12.69 pp@2null1] 0.818182 -12.6901 39.0 0.304878 -12.6901 women 0.888889 -12.6903 affairs 0.888889 -12.6903 voice quality 0.769231 -12.6904 sold 0.32 -12.6904 testing, 0.888889 -12.6906 (tag) 0.888889 -12.6906 we re-implement 0.888889 -12.6906 about 1% 0.6875 -12.6906 gothic 0.28125 -12.6908 word’s 0.565217 -12.6908 3.2.2 0.888889 -12.691 predicators 0.888889 -12.691 arrange 0.565217 -12.691 contradiction 0.338235 -12.6912 evaluates 0.769231 -12.6914 35k 0.818182 -12.6915 50 times 0.888889 -12.6918 speaker-specific 0.888889 -12.6918 seamlessly 0.888889 -12.6918 18th 0.888889 -12.6918 www 0.888889 -12.6918 0.308 0.484848 -12.6919 nw 0.565217 -12.692 primary evaluation 0.888889 -12.6921 linguisticallymotivated 0.888889 -12.6921 0.654 0.888889 -12.6921 38.6 0.888889 -12.6921 32.9 0.888889 -12.6921 crossed 0.818182 -12.6922 low: 0.888889 -12.6925 5.5% 0.888889 -12.6925 0.826 0.818182 -12.6926 bullets, 0.34375 -12.6926 giza++ (och 0.565217 -12.6927 follows from 0.818182 -12.6929 (xi,xj) 0.888889 -12.6929 tentative 0.888889 -12.6929 doubles 0.888889 -12.6929 ours: 0.888889 -12.6929 .57 0.292135 -12.693 came 0.565217 -12.693 (miwa 0.714286 -12.6931 journals 0.769231 -12.6933 benefit 0.888889 -12.6933 opca, 0.888889 -12.6933 0.350 0.888889 -12.6933 qtb 0.888889 -12.6933 finishes 0.20904 -12.6933 prefix 0.631579 -12.6933 73.3 0.818182 -12.6936 steyvers 0.888889 -12.6936 hu, 0.888889 -12.6936 0.676 0.888889 -12.6936 beneath 0.888889 -12.6936 64.9 0.888889 -12.6936 encountering 0.888889 -12.6936 crossdomain 0.888889 -12.6936 machinelearning 0.888889 -12.6936 all-words 0.254237 -12.6937 reason for 0.140665 -12.694 is very 0.888889 -12.694 inter-punctuation 0.888889 -12.694 (vilain 0.888889 -12.694 (paul 0.888889 -12.694 lexicalized, 0.888889 -12.694 dataset) 0.349206 -12.6941 these experiments 0.295455 -12.6942 human language 0.818182 -12.6943 334 0.418605 -12.6943 how can 0.28 -12.6944 reference. 0.888889 -12.6944 multi-task learning 0.888889 -12.6944 wordnet’s 0.6875 -12.6944 [a1,a2] 0.888889 -12.6948 27.4 0.888889 -12.6948 realization, 0.888889 -12.6948 skill 0.888889 -12.6948 instance-based 0.888889 -12.6948 engineers 0.888889 -12.6948 noun synsets 0.888889 -12.6948 386 0.888889 -12.6948 fillmore, 0.818182 -12.695 were discarded. 0.888889 -12.6952 fires when 0.888889 -12.6952 1btpd 0.888889 -12.6952 non-monotone 0.888889 -12.6952 i/o 0.888889 -12.6955 (verhagen 0.888889 -12.6955 navigator 0.888889 -12.6955 quote, 0.888889 -12.6955 meetings, 0.888889 -12.6955 (pr) 0.888889 -12.6955 challenge: 0.888889 -12.6955 realworld 0.888889 -12.6955 860 0.6875 -12.6956 are becoming 0.151335 -12.6959 4.3 0.714286 -12.6959 ↓ 0.714286 -12.6959 bilingual dictionary. 0.888889 -12.6959 clustering algorithms, 0.888889 -12.6959 intriguing 0.888889 -12.6959 (recasens 0.888889 -12.6959 justifies 0.888889 -12.6959 specificities 0.19898 -12.6962 orientation 0.769231 -12.6963 266 0.769231 -12.6963 49.9 0.888889 -12.6963 nondeterministic 0.888889 -12.6963 mcintosh 0.888889 -12.6963 tive 0.888889 -12.6963 (hobbs, 0.888889 -12.6963 upper case 0.818182 -12.6964 211 0.818182 -12.6964 metavariable 0.714286 -12.6965 402 0.818182 -12.6967 (e,a) 0.0655647 -12.6967 pair 0.818182 -12.6971 further reduce 0.818182 -12.6971 (over all 0.888889 -12.6971 do-pst-dec 'he 0.888889 -12.6971 text simplification, 0.888889 -12.6971 deficiencies 0.888889 -12.6974 obscure 0.888889 -12.6974 score), 0.888889 -12.6974 (hinton 0.769231 -12.6976 krippendorff’s 0.333333 -12.6976 where they 0.590909 -12.6977 consortium 0.888889 -12.6978 d4 0.888889 -12.6978 superficially 0.888889 -12.6978 (seki 0.0905263 -12.6979 obtain 0.368421 -12.6981 g = 0.888889 -12.6982 67.8% 0.888889 -12.6982 (uszkoreit 0.888889 -12.6982 90.20 0.888889 -12.6986 1note that 0.888889 -12.6986 deserves 0.818182 -12.6988 transfer rule 0.470588 -12.6988 girl 0.888889 -12.699 picture, 0.888889 -12.699 inauguration 0.888889 -12.699 spanishenglish 0.888889 -12.699 wible 0.888889 -12.699 ’06 0.888889 -12.699 reorders 0.6875 -12.6991 68.5 0.888889 -12.6993 (sch¨utze, 0.888889 -12.6993 senellart, 0.888889 -12.6993 callisonburch, 0.888889 -12.6993 suggestive 0.888889 -12.6993 running times 0.888889 -12.6993 313 0.888889 -12.6997 mimacr 0.6875 -12.6997 [np 0.888889 -12.7001 (dyer, 0.888889 -12.7001 websites, 0.888889 -12.7001 manually specified 0.888889 -12.7001 3b 0.888889 -12.7001 ˇzabokrtsk´y 0.888889 -12.7001 detokenized 0.818182 -12.7002 (le 0.888889 -12.7005 okumura 0.888889 -12.7005 milk, 0.888889 -12.7005 pattvsm 0.888889 -12.7005 csds 0.888889 -12.7005 state-of-the-art, 0.194175 -12.7008 correlated 0.631579 -12.7008 now consider 0.818182 -12.7009 overal 0.888889 -12.7009 76.6 0.888889 -12.7009 10.6 0.888889 -12.7009 necesary 0.714286 -12.7011 .81 0.631579 -12.7011 messages were 0.769231 -12.7012 2; 0.769231 -12.7012 1 0 . 0.888889 -12.7012 one-best 0.818182 -12.7013 mfc 0.258621 -12.7014 looked 0.517241 -12.7014 (daum´e 0.818182 -12.7016 ”the 0.888889 -12.7016 phrasing 0.888889 -12.702 values; 0.888889 -12.702 hltcoe 0.888889 -12.702 apposition 0.888889 -12.702 ronaldo is 0.888889 -12.702 ibm4 0.484848 -12.7022 central to 0.818182 -12.7023 ℓ0 0.818182 -12.7023 lexicon-based sentiment 0.888889 -12.7024 showcase 0.888889 -12.7024 2007a), 0.888889 -12.7024 participles 0.818182 -12.7027 uoy: 0.631579 -12.7027 topic graph 0.888889 -12.7028 compare, 0.888889 -12.7028 3.3) 0.888889 -12.7028 follower’s 0.888889 -12.7031 21.0 0.888889 -12.7031 partitions, 0.888889 -12.7031 audio, 0.888889 -12.7035 krenn 0.888889 -12.7035 12.2 0.888889 -12.7035 63.1 0.888889 -12.7035 (levin, 0.888889 -12.7035 39.5 0.888889 -12.7035 deduced from 0.888889 -12.7035 providers 0.888889 -12.7035 noisiness 0.888889 -12.7035 80.1 0.888889 -12.7035 (celikyilmaz 0.818182 -12.7037 (en) 0.769231 -12.7038 0.5) 0.6 -12.7038 harper, 0.888889 -12.7039 giga corpus 0.888889 -12.7039 deficient 0.888889 -12.7039 (punyakanok 0.888889 -12.7039 0.610 0.888889 -12.7039 0.692 0.769231 -12.7042 jigsaw 0.288889 -12.7042 many other 0.888889 -12.7043 hamlet 0.888889 -12.7043 (richardson 0.888889 -12.7043 release, 0.888889 -12.7043 falsely 0.714286 -12.7044 objective: 0.888889 -12.7047 subparts 0.888889 -12.7047 post-edited 0.818182 -12.7048 section 6 presents 0.517241 -12.7048 ’7 0.888889 -12.705 (cucerzan, 0.888889 -12.705 distance: 0.888889 -12.705 interlocutors 0.888889 -12.705 query logs, 0.888889 -12.705 generate, 0.888889 -12.705 revision, 0.565217 -12.7053 equivalent. 0.888889 -12.7054 (2008)’s 0.888889 -12.7054 0.143 0.888889 -12.7054 pipelines, 0.6 -12.7054 disabilities 0.647059 -12.7057 80.6 0.888889 -12.7058 exchangeability 0.888889 -12.7058 claim, 0.888889 -12.7058 re-annotated 0.888889 -12.7058 advisors 0.714286 -12.706 q(x) 0.631579 -12.706 following. 0.457143 -12.7062 regression, 0.818182 -12.7062 horse 0.888889 -12.7062 82.1 0.888889 -12.7062 auto-generated bilingual 0.888889 -12.7062 weak, 0.565217 -12.7062 women’s 0.435897 -12.7062 entry. 0.418605 -12.7063 synonyms. 0.631579 -12.7065 third column 0.714286 -12.7066 somebody 0.518519 -12.7066 is performed. 0.888889 -12.7066 germanc-gs 0.888889 -12.7066 training data). 0.888889 -12.7066 heilman 0.888889 -12.7066 only) 0.888889 -12.7066 unaffected 0.447368 -12.7067 summarization system 0.590909 -12.7069 12. 0.888889 -12.707 equipment 0.888889 -12.707 alow 0.888889 -12.707 transliteration, 0.888889 -12.707 229 0.888889 -12.707 mwu-forest 0.888889 -12.707 thirds 0.888889 -12.7073 alshawi 0.888889 -12.7073 prolog 0.888889 -12.7077 79.7 0.888889 -12.7077 triger 0.888889 -12.7077 integer-valued 0.888889 -12.7077 shinyama 0.888889 -12.7077 cotraining 0.888889 -12.7077 damages 0.888889 -12.7077 discused 0.888889 -12.7077 40.1 0.888889 -12.7077 zeroing 0.888889 -12.7077 317 0.769231 -12.7078 ipad 0.308642 -12.7078 dashed 0.223684 -12.7079 ∧ 0.6875 -12.708 py 0.714286 -12.7081 8,000 0.888889 -12.7081 34.0 0.888889 -12.7081 0.375 0.6875 -12.7083 adjacent words 0.888889 -12.7085 head movement 0.888889 -12.7085 comprehensively 0.888889 -12.7085 misaligned 0.888889 -12.7085 clinician 0.888889 -12.7085 one-dimensional 0.888889 -12.7085 unnormalized 0.21875 -12.7086 improvements in 0.714286 -12.7087 asertion 0.888889 -12.7089 extractable 0.888889 -12.7089 leaves us 0.888889 -12.7089 31.7 0.229167 -12.709 results obtained 0.647059 -12.709 most significant 0.6 -12.709 right: 0.0684597 -12.7091 this paper 0.565217 -12.7092 slightly higher 0.888889 -12.7093 33.33 0.888889 -12.7093 re-trained 0.888889 -12.7093 cluster id’s 0.888889 -12.7093 m-th 0.888889 -12.7093 209), 0.888889 -12.7093 noun), 0.888889 -12.7093 13) 0.6 -12.7093 69.5 0.818182 -12.7094 bilingual treebanks 0.888889 -12.7096 92.5 0.888889 -12.7096 scu 0.888889 -12.7096 subregular 0.888889 -12.7096 δθ 0.304878 -12.7097 section 2. 0.769231 -12.7098 gregory 0.0881459 -12.71 (see 0.888889 -12.71 0.368 0.888889 -12.71 tolerate 0.888889 -12.71 ghkm algorithm 0.888889 -12.7104 (partial) 0.888889 -12.7104 quasi-newton 0.888889 -12.7104 reestimation 0.565217 -12.7104 henderson, 0.769231 -12.7108 voicing 0.888889 -12.7108 80.4 0.888889 -12.7108 chinese) 0.888889 -12.7108 (core 0.888889 -12.7108 levenberg 0.888889 -12.7108 61.7 0.888889 -12.7112 fairly small 0.888889 -12.7112 39.1 0.888889 -12.7112 method; 0.888889 -12.7112 ←− 0.435897 -12.7113 (toutanova 0.714286 -12.7115 andreas 0.6875 -12.7115 it starts 0.147309 -12.7115 institute 0.888889 -12.7116 liang, 0.888889 -12.7116 webbased 0.538462 -12.7117 have taken 0.818182 -12.7118 hills 0.888889 -12.7119 0.784 0.888889 -12.7119 isolating 0.888889 -12.7119 labelings 0.818182 -12.7122 training phase, 0.888889 -12.7123 finished, 0.888889 -12.7123 two-fold: 0.888889 -12.7123 opinion-labeled 0.888889 -12.7127 incorrectly, 0.818182 -12.7129 very sparse 0.647059 -12.7129 conclusion, 0.6875 -12.713 tea 0.173913 -12.713 experiments on 0.769231 -12.7131 paper). 0.278351 -12.7131 right-hand 0.888889 -12.7131 nouns), 0.888889 -12.7131 pukwac 0.888889 -12.7131 launched 0.888889 -12.7131 fragmentary 0.292135 -12.7133 say that 0.888889 -12.7135 standardization 0.888889 -12.7135 44.3 0.888889 -12.7135 (martin 0.888889 -12.7135 .18 0.888889 -12.7135 tennis 0.888889 -12.7135 co-trained 0.888889 -12.7135 strands 0.457143 -12.7139 23, 0.888889 -12.7139 document) 0.888889 -12.7139 nonparallel 0.888889 -12.7139 above-mentioned 0.888889 -12.7139 (30 0.538462 -12.714 this formulation 0.377358 -12.7142 two annotators 0.714286 -12.7143 we use, 0.888889 -12.7143 97.2 0.470588 -12.7146 slightly more 0.888889 -12.7146 59.2 0.888889 -12.7146 choi, 0.418605 -12.7146 viable 0.714286 -12.7149 rosenfeld 0.888889 -12.715 0.803 0.888889 -12.715 basilisk 0.888889 -12.715 lts 0.888889 -12.715 96.9 0.888889 -12.715 0.8% 0.435897 -12.7151 carrying 0.888889 -12.7154 dags 0.888889 -12.7154 boyd-graber 0.888889 -12.7154 2004); 0.888889 -12.7154 cap 0.888889 -12.7154 practitioners 0.888889 -12.7154 one-against-others 0.888889 -12.7154 intention, 0.714286 -12.7155 km 0.818182 -12.7157 ,f,q 0.6875 -12.7157 significance tests 0.714286 -12.7158 ubiquitous 0.888889 -12.7158 coarse-pass 0.888889 -12.7158 oov rate 0.888889 -12.7162 comprehenders 0.888889 -12.7162 23.0 0.888889 -12.7162 unfolding 0.888889 -12.7162 pap 0.483871 -12.7164 to normalize 0.5 -12.7165 word–colour 0.888889 -12.7166 indefinites, 0.888889 -12.7166 35.9 0.888889 -12.7166 clip 0.888889 -12.7166 array, 0.37037 -12.7168 mentioned above, 0.647059 -12.7169 exciting 0.888889 -12.717 alm, 0.888889 -12.7173 artstein 0.888889 -12.7173 tip 0.888889 -12.7173 compressions, 0.888889 -12.7173 interdependencies 0.888889 -12.7173 welcome 0.888889 -12.7173 arose 0.888889 -12.7173 1982) 0.888889 -12.7173 language pairs: 0.714286 -12.7174 “for 0.647059 -12.7174 turnstileleft 0.888889 -12.7177 dispute 0.888889 -12.7177 4.1% 0.888889 -12.7177 christopher d. manning. 0.888889 -12.7177 doctors 0.6875 -12.7178 unattested 0.888889 -12.7181 fragmentation 0.256637 -12.7182 can have 0.888889 -12.7185 advancement 0.888889 -12.7185 ieer 0.888889 -12.7185 gnc 0.818182 -12.7186 swc 0.714286 -12.7186 bounded by 0.647059 -12.7186 95.6 0.565217 -12.7187 p3 0.888889 -12.7189 clients, 0.888889 -12.7189 bi-text 0.818182 -12.7193 symmetrical 0.818182 -12.7193 (etzioni 0.888889 -12.7193 t-lemmas 0.888889 -12.7193 dorr 0.888889 -12.7193 11.4 0.271845 -12.7194 procedings of 0.818182 -12.7197 lexical item. 0.818182 -12.7197 -を 0.888889 -12.7197 cue-phrase-based 0.888889 -12.7197 abney 0.888889 -12.7197 solicited 0.888889 -12.7197 witnessed 0.888889 -12.7197 14.8 0.888889 -12.7197 moldovan, 0.888889 -12.7197 give, 0.714286 -12.7199 ph 0.818182 -12.72 concrete example 0.818182 -12.72 letter n-gram 0.888889 -12.7201 in proceedings of acl, 0.888889 -12.7201 molecule 0.888889 -12.7201 (30%) 0.888889 -12.7201 agarwal, 0.888889 -12.7201 0.06 0.07 0.888889 -12.7201 (matsuzaki 0.818182 -12.7204 last step 0.888889 -12.7204 plurality 0.888889 -12.7204 measurement, 0.888889 -12.7204 obligation 0.888889 -12.7204 electrons 0.888889 -12.7204 unique, 0.888889 -12.7204 likely, 0.888889 -12.7204 non-fluent 0.518519 -12.7205 impractical 0.888889 -12.7208 differences: 0.888889 -12.7208 13.3 0.447368 -12.7209 misclassified 0.203297 -12.7211 precision. 0.888889 -12.7212 e1, 0.888889 -12.7212 significantly 0.6875 -12.7214 ue 0.418605 -12.7214 we have proposed 0.387755 -12.7216 transforming 0.888889 -12.7216 btg 0.888889 -12.722 com 0.888889 -12.722 maletti 0.888889 -12.722 take, 0.888889 -12.722 graphemic 0.888889 -12.722 asts 0.144444 -12.7221 figure 2 0.888889 -12.7224 (evert, 0.888889 -12.7224 16.2 0.888889 -12.7224 71.9 0.888889 -12.7224 resemblance 0.888889 -12.7224 vandalizing 0.888889 -12.7224 t-lemma 0.888889 -12.7224 case; 0.647059 -12.7225 separately from 0.565217 -12.7227 street, 0.888889 -12.7228 assumed, 0.888889 -12.7228 tractability 0.888889 -12.7228 moens 0.888889 -12.7228 english-language 0.647059 -12.7228 cars 0.714286 -12.723 77.8 0.888889 -12.7232 narrative discourse 0.888889 -12.7232 essential, 0.888889 -12.7232 following subsections 0.888889 -12.7232 errors) 0.247934 -12.7232 this study 0.888889 -12.7235 (in)correctness 0.888889 -12.7235 entrel 0.888889 -12.7235 overestimate 0.6875 -12.7235 though not 0.288889 -12.7237 allocation 0.888889 -12.7239 meeting, 0.888889 -12.7239 erjavec 0.888889 -12.7239 filer 0.888889 -12.7239 67.6 0.888889 -12.7239 motion, 0.888889 -12.7239 constrained, 0.888889 -12.7239 shapes, 0.647059 -12.7239 tightening 0.5 -12.7239 from wikipedia. 0.818182 -12.724 change over 0.470588 -12.7241 independent. 0.631579 -12.7241 manchester, 0.888889 -12.7243 objectives, 0.888889 -12.7243 presume 0.6875 -12.7244 here, we use 0.518519 -12.7246 despite its 0.888889 -12.7247 edges) 0.888889 -12.7247 ncs, 0.888889 -12.7247 lived 0.888889 -12.7247 “oracle” 0.6 -12.7247 an extremely 0.888889 -12.7251 (levy 0.888889 -12.7251 (human 0.888889 -12.7251 ts, 0.888889 -12.7251 7.8 0.888889 -12.7251 bilinear 0.888889 -12.7251 features” 0.0324865 -12.7251 such 0.714286 -12.7252 pda 0.6875 -12.7253 this demonstrates that 0.538462 -12.7253 of. 0.888889 -12.7255 non-polar 0.888889 -12.7255 inspection, 0.0600858 -12.7256 found 0.170543 -12.7256 which has 0.818182 -12.7257 (2007) found 0.435897 -12.7257 andrew 0.132251 -12.7258 enough 0.888889 -12.7259 93.1 0.888889 -12.7259 535 0.888889 -12.7259 lexico-semantic 0.888889 -12.7259 (again, 0.888889 -12.7259 accuracy) 0.888889 -12.7259 classifier-based 0.565217 -12.7259 can incorporate 0.5 -12.7259 to preserve 0.631579 -12.726 mention-entity 0.538462 -12.726 (these 0.888889 -12.7263 licensing 0.888889 -12.7263 63.4 0.888889 -12.7263 bid 0.888889 -12.7263 accented 0.647059 -12.7265 y., 0.418605 -12.7265 are grouped 0.888889 -12.7267 indo-european languages 0.888889 -12.7267 24.3 0.888889 -12.7267 levy, 0.538462 -12.7267 computed over 0.818182 -12.7268 (f1 0.888889 -12.7271 298 0.888889 -12.7271 estimated, 0.888889 -12.7271 driven learning 0.888889 -12.7271 causally 0.888889 -12.7271 possess 0.818182 -12.7272 02-21 0.888889 -12.7274 (almost) 0.888889 -12.7274 indispensable 0.631579 -12.7274 administered 0.888889 -12.7278 co-occurrence, 0.288889 -12.7279 levels. 0.888889 -12.7282 flickinger, 0.888889 -12.7282 into, 0.888889 -12.7282 torgo 0.888889 -12.7282 (s, 0.888889 -12.7282 deprived 0.888889 -12.7282 completes 0.818182 -12.7283 morse 0.818182 -12.7286 supplied with 0.888889 -12.7286 ace, 0.888889 -12.7286 accidents, 0.888889 -12.7286 complexity: 0.888889 -12.7286 carvalho 0.888889 -12.7286 subrequests 0.6875 -12.7286 dificulty 0.647059 -12.7288 i=0 0.888889 -12.729 2.1% 0.888889 -12.729 keller 0.714286 -12.7293 s(a) 0.888889 -12.7294 0.633 0.888889 -12.7294 deficiency 0.888889 -12.7294 self-supervised 0.888889 -12.7294 redundancies 0.888889 -12.7294 sends 0.888889 -12.7298 69.1 0.888889 -12.7298 ovs 0.888889 -12.7298 cal 0.714286 -12.7302 cbhp 0.888889 -12.7302 delimit 0.888889 -12.7302 overrepresented 0.888889 -12.7302 tri-gram 0.301205 -12.7304 bionlp’09 0.6875 -12.7304 tempeval-2010 0.888889 -12.7306 anchoring 0.888889 -12.7306 rouge, 0.888889 -12.7306 automatically-generated 0.888889 -12.7306 rv 0.888889 -12.7306 amani’s 0.888889 -12.7306 ren 0.818182 -12.7308 68-form 0.888889 -12.731 358 0.888889 -12.731 irc 0.541667 -12.7312 so much 0.6 -12.7313 2009); 0.888889 -12.7314 stated, 0.888889 -12.7314 data-oriented 0.888889 -12.7318 allan, 0.888889 -12.7318 definiteness 0.888889 -12.7318 contexts: 0.5 -12.732 (bannard 0.888889 -12.7321 25.2 0.888889 -12.7321 gestural 0.888889 -12.7321 lm) 0.888889 -12.7321 reinforcing 0.888889 -12.7321 elapsed 0.483871 -12.7323 these numbers 0.888889 -12.7325 βi 0.888889 -12.7325 imposed by 0.888889 -12.7325 cplsa 0.888889 -12.7325 preferentially 0.888889 -12.7325 label; 0.888889 -12.7325 redirects 0.714286 -12.7327 landauer 0.888889 -12.7329 reorganization 0.888889 -12.7329 eo-jeols 0.714286 -12.7331 barbara 0.457143 -12.7333 jth 0.888889 -12.7333 charges 0.888889 -12.7333 fat 0.888889 -12.7333 hahn 0.888889 -12.7333 additional, 0.888889 -12.7333 platt 0.888889 -12.7333 msr-nlp 0.888889 -12.7333 (λ 0.888889 -12.7333 userprof 0.714286 -12.7334 configuration 0.177215 -12.7334 toolkit 0.818182 -12.7337 i.e.: 0.818182 -12.7337 beneficiary 0.888889 -12.7337 referent, 0.888889 -12.7337 skills, 0.888889 -12.7337 pertains 0.888889 -12.7337 text-mined 0.888889 -12.7337 0.437 0.35 -12.7337 its performance 0.888889 -12.7341 koc 0.888889 -12.7341 number) 0.888889 -12.7341 (dashed 0.888889 -12.7341 tanaka 0.818182 -12.7344 time: 0.6 -12.7344 (those 0.888889 -12.7345 arabicenglish 0.888889 -12.7345 affixation 0.888889 -12.7345 add, 0.888889 -12.7345 hierarchies, 0.888889 -12.7345 comfortable 0.818182 -12.7348 (w) 0.647059 -12.7348 attribution, 0.0447165 -12.7349 while 0.888889 -12.7349 bioyatea 0.888889 -12.7349 argument expectations 0.888889 -12.7349 frustrated 0.888889 -12.7353 quantification, 0.888889 -12.7353 ans, 0.888889 -12.7353 concatenative 0.888889 -12.7353 rip.v 0.888889 -12.7353 kuhlmann 0.888889 -12.7353 cj, 0.888889 -12.7353 lexicalizations 0.409091 -12.7353 0.64 0.409091 -12.7353 used during 0.714286 -12.7356 historians 0.888889 -12.7357 containment 0.888889 -12.7357 542 0.888889 -12.7357 47.7 0.216561 -12.7358 candidates. 0.714286 -12.7359 pyramid score 0.714286 -12.7359 angeles, 0.888889 -12.7361 cerebral 0.888889 -12.7361 very different. 0.888889 -12.7361 ranker, 0.888889 -12.7361 eastern 0.22449 -12.7363 length. 0.888889 -12.7365 gaze, 0.888889 -12.7365 hard, 0.888889 -12.7365 appliances 0.409091 -12.7367 cambridge, 0.818182 -12.7369 (translation 0.888889 -12.7369 identity, 0.888889 -12.7369 geography, 0.888889 -12.7373 nu 0.888889 -12.7373 length) 0.888889 -12.7373 reinforced 0.888889 -12.7373 mccc-i 0.888889 -12.7373 (h) 0.888889 -12.7373 vps 0.888889 -12.7373 siblings, 0.888889 -12.7373 new paradigm 0.888889 -12.7377 52.9 0.888889 -12.7377 manual alignments 0.888889 -12.7377 mwu-aware 0.888889 -12.7377 girju, 0.888889 -12.7377 phylogeny 0.541667 -12.7377 straightforward, 0.6 -12.7379 ample 0.888889 -12.738 (ramisch 0.5 -12.738 these metrics 0.647059 -12.7382 argument type 0.888889 -12.7384 12.0 0.888889 -12.7384 16.3 0.888889 -12.7384 conforming 0.27957 -12.7385 a great 0.818182 -12.7388 l-match 0.518519 -12.7388 3.1). 0.888889 -12.7388 aged 0.888889 -12.7388 godis dialogue 0.888889 -12.7388 59.4 0.888889 -12.7388 randomness 0.888889 -12.7388 pers 0.888889 -12.7388 grapheme-to-phoneme 0.75 -12.7391 definition questions 0.888889 -12.7392 13.4 0.888889 -12.7392 enters 0.888889 -12.7392 result: 0.888889 -12.7392 modular, 0.888889 -12.7392 mt09 0.20339 -12.7392 table, 0.75 -12.7394 daumé 0.888889 -12.7396 tissue 0.888889 -12.7396 argm-tmp 0.425 -12.7397 southern 0.75 -12.7397 precise, 0.818182 -12.7398 japanese-english 0.888889 -12.74 michelle 0.888889 -12.74 feature ablation 0.888889 -12.74 [0,1], 0.169231 -12.7403 labels. 0.888889 -12.7404 covariate 0.888889 -12.7404 type; 0.888889 -12.7404 pronouncing 0.888889 -12.7404 webscale 0.647059 -12.7405 ohio 0.435897 -12.7406 traversal 0.75 -12.7407 46% 0.323944 -12.7407 tagger. 0.307692 -12.7408 studied in 0.888889 -12.7408 10027, 0.565217 -12.7409 tagset, 0.27551 -12.741 conclude that 0.888889 -12.7412 0.004 0.888889 -12.7412 karlsruhe 0.888889 -12.7412 fij 0.888889 -12.7412 arranz 0.518519 -12.7413 ja 0.409091 -12.7414 provide more 0.565217 -12.7414 minutes, 0.888889 -12.7416 model1 0.888889 -12.7416 379 0.888889 -12.7416 affirmative 0.888889 -12.7416 greg 0.75 -12.7417 simpler than 0.647059 -12.7417 danger 0.818182 -12.742 a wealth 0.888889 -12.742 histone 0.888889 -12.742 top-ranking 0.888889 -12.742 target-response 0.363636 -12.742 is structured as 0.75 -12.7421 hdlr 0.647059 -12.7422 limsi-cnrs 0.818182 -12.7424 276 0.888889 -12.7424 panel, 0.888889 -12.7424 0.164 0.888889 -12.7424 θne 0.363636 -12.7424 accommodate 0.0885142 -12.7425 the performance of 0.138107 -12.7426 cosine 0.818182 -12.7428 menu, 0.888889 -12.7428 sensetagged 0.435897 -12.7429 learner corpus 0.6 -12.743 melm, 0.888889 -12.7432 slight increase 0.888889 -12.7432 graphically 0.888889 -12.7432 event’s 0.518519 -12.7434 agreement among 0.888889 -12.7436 milliseconds 0.565217 -12.7439 comparisons with 0.888889 -12.744 turktest 0.818182 -12.7442 husband 0.714286 -12.7442 certificates 0.888889 -12.7444 fixations 0.888889 -12.7444 chronological 0.888889 -12.7444 instrumental 0.888889 -12.7444 2,500 0.0645342 -12.7448 form 0.714286 -12.7448 detectors, 0.888889 -12.7448 historically 0.888889 -12.7448 alters 0.888889 -12.7448 xiao 0.25 -12.7449 this result 0.714286 -12.7451 26.7 0.888889 -12.7452 esol 0.888889 -12.7452 294 0.888889 -12.7452 reader, 0.888889 -12.7452 declines 0.888889 -12.7452 activating 0.888889 -12.7452 klementiev 0.888889 -12.7452 margins 0.818182 -12.7453 2v 0.818182 -12.7453 protagonist 0.647059 -12.7454 sebastian 0.888889 -12.7456 goldstein, 0.888889 -12.7456 metonymic 0.888889 -12.7456 an off-the-shelf 0.888889 -12.7456 paralinguistic 0.888889 -12.7456 tandem 0.888889 -12.7456 emulate 0.888889 -12.7456 statistics: 0.888889 -12.7456 punctuation) 0.647059 -12.7457 buckets 0.888889 -12.746 novels, 0.38 -12.746 multinomial distribution 0.409091 -12.746 frank 0.75 -12.7461 87.3 0.818182 -12.7464 praise 0.75 -12.7464 syntactic dependencies. 0.75 -12.7464 syntactic categories. 0.888889 -12.7464 url, 0.888889 -12.7464 levantine 0.888889 -12.7464 lub 0.714286 -12.7467 print 0.888889 -12.7468 nsp 0.888889 -12.7468 expert annotations 0.888889 -12.7468 z., 0.333333 -12.7469 sixth 0.647059 -12.7469 log-linear combination 0.483871 -12.7469 higher scores 0.319444 -12.7471 (2009) and 0.75 -12.7471 bernstein 0.888889 -12.7476 ter−bleu 0.888889 -12.7476 care, 0.888889 -12.7476 potentialy 0.888889 -12.7476 trend, 0.888889 -12.7476 0.1) 0.75 -12.7478 baselines: 0.888889 -12.748 senseval 0.888889 -12.748 internet, 0.888889 -12.748 downward 0.75 -12.7481 87.7 0.38 -12.7481 parallel data. 0.888889 -12.7484 1975), 0.888889 -12.7484 χmax 0.714286 -12.7486 know about 0.888889 -12.7488 nc-pattern 0.888889 -12.7488 85.4 0.888889 -12.7488 tct-style 0.888889 -12.7488 (information 0.888889 -12.7488 auxiliaries, 0.888889 -12.7488 effect: 0.888889 -12.7488 editor, 0.333333 -12.7491 assist 0.888889 -12.7492 lady 0.888889 -12.7492 devault 0.888889 -12.7492 affect, 0.888889 -12.7492 64.0 0.13285 -12.7494 starting 0.425 -12.7496 the latest 0.888889 -12.7496 493 0.888889 -12.7496 chained 0.888889 -12.7496 21.7 0.888889 -12.7496 θk 0.888889 -12.7496 (arun et 0.888889 -12.75 oracle, 0.888889 -12.75 ekman’s 0.888889 -12.75 0.8, 0.888889 -12.75 gamma 0.565217 -12.7502 labeled instances 0.571429 -12.7503 wei 0.888889 -12.7504 acording to 0.888889 -12.7504 205 0.888889 -12.7504 agents’ 0.888889 -12.7504 taught 0.888889 -12.7504 isozaki, 0.888889 -12.7504 board, 0.888889 -12.7504 space; 0.714286 -12.7506 when measuring 0.483871 -12.7506 a fixed number 0.888889 -12.7508 311 0.888889 -12.7508 go” 0.565217 -12.751 we prefer 0.888889 -12.7512 dsm 0.888889 -12.7512 plurk 0.888889 -12.7512 metric r p 0.888889 -12.7512 anticipated 0.888889 -12.7516 screens 0.888889 -12.7516 promotion 0.888889 -12.7516 bonus 0.888889 -12.7516 wmeb 0.888889 -12.7516 ohio, 0.518519 -12.7518 roman 0.184332 -12.7519 semantics of 0.888889 -12.752 mention-pair model 0.888889 -12.752 verbalized 0.090209 -12.752 let 0.571429 -12.7521 to promote 0.5 -12.7521 possibility that 0.888889 -12.7524 s2t 0.571429 -12.7524 linguistic features. 0.75 -12.7525 correction task 0.75 -12.7525 acount 0.75 -12.7528 bach 0.888889 -12.7528 very large. 0.888889 -12.7528 (document 0.888889 -12.7528 match) 0.151899 -12.7529 sections 0.75 -12.7532 eparl 0.888889 -12.7532 basic, 0.126087 -12.7533 own 0.197861 -12.7533 is done 0.647059 -12.7533 yp 0.75 -12.7535 28.9 0.888889 -12.7536 not perfect, 0.888889 -12.7536 (figures 0.888889 -12.7536 flowerpot 0.888889 -12.7536 22, 0.888889 -12.7536 cosine-similarity 0.888889 -12.7536 commutative 0.888889 -12.7536 fetch 0.565217 -12.7538 homotopy 0.647059 -12.7539 pops 0.888889 -12.754 intelligible 0.888889 -12.754 (tn), 0.888889 -12.754 wheel 0.888889 -12.754 heap 0.888889 -12.754 3.1.3 0.888889 -12.754 (h 0.38 -12.7542 relatedness between 0.888889 -12.7544 risks 0.888889 -12.7544 propagates 0.888889 -12.7544 hyperplanes 0.5 -12.7544 past, 0.565217 -12.7546 short sentences 0.483871 -12.7547 cardinality 0.888889 -12.7548 framing 0.225352 -12.755 to indicate 0.647059 -12.755 each hit 0.888889 -12.7552 hindi-english 0.6 -12.7554 by far 0.888889 -12.7556 earthquake, 0.888889 -12.7556 0.325 0.888889 -12.7556 uncertain-input 0.647059 -12.7556 converse 0.818182 -12.756 time period 0.888889 -12.756 skantze 0.6 -12.7562 protein modification 0.357143 -12.7563 this assumption 0.888889 -12.7564 neighbourhood 0.888889 -12.7564 cosines 0.888889 -12.7564 quotes from 0.888889 -12.7564 (mainly 0.409091 -12.7567 acts. 0.391304 -12.7568 adjectives. 0.888889 -12.7568 hashtag topics 0.888889 -12.7568 fairly good 0.888889 -12.7568 authority, 0.1 -12.7568 collection 0.571429 -12.7568 our system outperforms 0.75 -12.7569 biocuration 0.818182 -12.7571 occur frequently 0.818182 -12.7571 .64 0.818182 -12.7571 @, 0.818182 -12.7571 maximization. 0.818182 -12.7571 tile 0.888889 -12.7572 (higher 0.888889 -12.7572 alfonseca 0.888889 -12.7572 rephrasings 0.888889 -12.7572 umls, 0.888889 -12.7572 m-bleu 0.888889 -12.7572 gravano, 0.888889 -12.7572 montoyo 0.1673 -12.7574 texts. 0.518519 -12.7574 we start by 0.75 -12.7576 • line 0.888889 -12.7576 segmentation: 0.888889 -12.7576 intermediary 0.27 -12.7578 r: 0.888889 -12.758 appears, 0.888889 -12.758 “full” 0.888889 -12.758 ai, 0.888889 -12.758 downloadable 0.191919 -12.7581 that uses 0.414634 -12.7582 bf 0.888889 -12.7584 concert 0.888889 -12.7584 indegree 0.6 -12.7584 coresponding 0.714286 -12.7586 (creutz 0.541667 -12.7586 parallel texts 0.888889 -12.7588 biocurators 0.888889 -12.7588 acetylation 0.888889 -12.7588 event-based 0.888889 -12.7588 finance, 0.888889 -12.7588 information-state 0.372549 -12.7591 all pairs of 0.432432 -12.7591 blast 0.888889 -12.7592 (kazama 0.888889 -12.7592 business, 0.75 -12.7596 prince 0.75 -12.7596 18.6 0.888889 -12.7596 backchannel-inviting 0.888889 -12.7596 paper; 0.888889 -12.7596 dynamic, 0.125541 -12.7598 degree 0.75 -12.76 natural question 0.888889 -12.76 our lda+hellinger 0.888889 -12.7604 ian 0.888889 -12.7604 2.2) 0.328358 -12.7605 parenrightbigg 0.714286 -12.7606 refering 0.75 -12.7606 54.0 0.75 -12.7606 document; 0.888889 -12.7608 europe, 0.888889 -12.7612 tine 0.233083 -12.7613 := 0.571429 -12.7614 stolcke 0.647059 -12.7615 is basically 0.888889 -12.7616 1.13 0.888889 -12.7616 wordlda 0.6 -12.7616 kernel methods 0.185714 -12.7616 recall. 0.25 -12.7617 future, 0.414634 -12.7618 extraction algorithm 0.46875 -12.7619 miwa et 0.256881 -12.7622 easier to 0.46875 -12.7623 5the 0.888889 -12.7624 crf-lm 0.888889 -12.7624 manning’s 0.888889 -12.7624 pedagogical 0.888889 -12.7624 devices, 0.888889 -12.7624 inegationslash= 0.888889 -12.7624 prp$ 0.38 -12.7625 explains the 0.571429 -12.7627 conflated 0.290698 -12.7628 degrees 0.888889 -12.7628 2003): 0.888889 -12.7628 utilising 0.888889 -12.7628 (2009)), 0.888889 -12.7628 erors 0.888889 -12.7628 languages) 0.888889 -12.7632 277 0.888889 -12.7636 homographs 0.888889 -12.7636 84.2 0.888889 -12.7636 sub-tasks 0.888889 -12.7636 314 0.218543 -12.7637 [ 0.611111 -12.7637 rewarded 0.666667 -12.764 00.0 0.888889 -12.764 change: 0.888889 -12.764 nitin 0.888889 -12.764 2.00 0.888889 -12.764 sight, 0.888889 -12.764 534 0.888889 -12.764 topk 0.888889 -12.764 decodes 0.714286 -12.7642 refsum 0.666667 -12.7643 speech understanding 0.541667 -12.7643 same entity. 0.444444 -12.7644 latent topic 0.888889 -12.7645 forest) 0.888889 -12.7645 shorter, 0.888889 -12.7645 φi 0.888889 -12.7645 hmeant using 0.666667 -12.7646 sharon 0.75 -12.7647 word/phrase 0.75 -12.7647 2p 0.888889 -12.7649 mary, 0.888889 -12.7649 era 0.888889 -12.7653 perceptron, 0.0337026 -12.7656 but 0.888889 -12.7657 retained, 0.888889 -12.7657 genia, 0.666667 -12.7658 aligner. 0.6 -12.766 timegraph 0.75 -12.7661 u(i) 0.888889 -12.7661 informing 0.888889 -12.7661 acceleration 0.888889 -12.7661 italic 0.888889 -12.7661 ppcs 0.24 -12.7662 expected, 0.27957 -12.7664 models trained 0.888889 -12.7665 equilibrium 0.425 -12.7666 0.36 0.888889 -12.7669 12.4 0.888889 -12.7669 ofthe 0.888889 -12.7669 thresholds, 0.75 -12.7671 39.3 0.46875 -12.7671 nproductdisplay 0.647059 -12.7671 heuristic phrase 0.888889 -12.7673 orthography, 0.0833333 -12.7673 a large 0.75 -12.7675 clause boundaries 0.227941 -12.7677 instead, we 0.714286 -12.7677 ˆθ 0.888889 -12.7677 experts’ 0.3 -12.768 iteration. 0.888889 -12.7681 (cr) 0.888889 -12.7681 free-text 0.75 -12.7682 96.5 0.483871 -12.7684 and mccallum, 0.888889 -12.7685 webpage 0.714286 -12.7687 elicited 0.214286 -12.7687 detection. 0.888889 -12.7689 suggested, 0.714286 -12.769 initial seed 0.611111 -12.769 an intelligent 0.75 -12.7692 sublanguage 0.647059 -12.7692 crf, 0.888889 -12.7693 comma, 0.666667 -12.7694 timex3 0.75 -12.7695 proliferation 0.888889 -12.7697 lexicographer 0.611111 -12.7699 x: 0.647059 -12.77 is preferable 0.647059 -12.77 bins. 0.888889 -12.7701 stereotypical 0.888889 -12.7701 posited 0.888889 -12.7701 span threshold 0.75 -12.7702 redundant, 0.75 -12.7702 (percentage 0.75 -12.7702 factor λ 0.888889 -12.7706 αt 0.888889 -12.7706 ingredients 0.6 -12.7707 proposed method. 0.714286 -12.771 uniprot 0.888889 -12.771 substructure 0.194737 -12.771 interpreted 0.75 -12.7713 (ward 0.888889 -12.7714 14, 0.888889 -12.7714 dollars 0.888889 -12.7714 laplacian 0.888889 -12.7714 425 0.75 -12.7716 reiter, 0.888889 -12.7718 cfg, 0.888889 -12.7718 (three 0.888889 -12.7718 occupied 0.888889 -12.7718 relay 0.647059 -12.7718 hoang, 0.75 -12.7723 30.1 0.75 -12.7723 0.746 0.611111 -12.7724 more expressive 0.647059 -12.7724 original phrase 0.888889 -12.7726 negate 0.888889 -12.7726 maca 0.888889 -12.7726 entity_a 0.888889 -12.7726 elsewhere, 0.888889 -12.7726 601 0.571429 -12.7729 dissociation 0.888889 -12.773 kan 0.888889 -12.773 capabilities, 0.888889 -12.773 apparently, 0.888889 -12.773 -0.09 0.888889 -12.773 knowledgebased 0.888889 -12.7734 2010: 0.714286 -12.7736 prep, 0.888889 -12.7738 amit 0.888889 -12.7738 illustration, 0.888889 -12.7738 caled 0.888889 -12.7738 graham, 0.888889 -12.7738 causal, 0.75 -12.774 tradreg 0.52 -12.774 gapping 0.888889 -12.7742 acoustic-prosodic 0.888889 -12.7742 null), 0.888889 -12.7742 excluded, 0.888889 -12.7742 5.3.2 0.888889 -12.7746 modifiers 0.666667 -12.7747 0.0001 0.571429 -12.7748 constructed. 0.178571 -12.775 categories. 0.714286 -12.775 242 0.888889 -12.775 concisely 0.888889 -12.775 preprocess 0.382979 -12.7753 community, 0.888889 -12.7755 m−1 0.888889 -12.7755 accused 0.888889 -12.7759 assumptions: 0.888889 -12.7759 kishore 0.0771144 -12.776 particular 0.75 -12.7761 dale, 0.888889 -12.7763 subtitles 0.888889 -12.7763 link-based 0.382979 -12.7766 2we 0.611111 -12.7766 1024 0.888889 -12.7767 phonology 0.888889 -12.7767 (bottom) 0.75 -12.7768 preliminary experiments, 0.75 -12.7771 reiter 0.75 -12.7771 taging 0.611111 -12.7772 top 5 0.611111 -12.7772 several studies 0.541667 -12.7772 web-derived 0.5 -12.7772 pred 0.888889 -12.7775 simplification engine 0.888889 -12.7775 2.2.3 0.888889 -12.7775 (english, 0.46875 -12.7776 hr 0.372549 -12.7782 system performs 0.888889 -12.7787 video, 0.5 -12.779 test documents 0.888889 -12.7792 subtypes, 0.888889 -12.7792 server, 0.888889 -12.7792 elements: 0.3 -12.7796 an example of a 0.888889 -12.7796 hter, 0.888889 -12.7796 follow, 0.888889 -12.7796 suffice 0.647059 -12.7799 adverbs. 0.0985155 -12.7799 addition 0.8 -12.7799 650 0.888889 -12.78 “ed 0.888889 -12.78 equate 0.888889 -12.78 (whose 0.888889 -12.78 route, 0.888889 -12.78 infection 0.482759 -12.78 corpus; 0.8 -12.7802 r(s) 0.8 -12.7802 posted by 0.8 -12.7802 arousal 0.611111 -12.7803 sij 0.888889 -12.7804 garden 0.888889 -12.7804 commonplace 0.888889 -12.7804 construction: 0.888889 -12.7804 giza-strong 0.888889 -12.7804 global maximalization 0.888889 -12.7804 concordia 0.482759 -12.7807 (griffiths 0.666667 -12.7808 33.4 0.888889 -12.7808 glad 0.888889 -12.7808 bio-molecular 0.8 -12.781 (cer 0.888889 -12.7812 after analyzing 0.75 -12.7813 was initially 0.611111 -12.7814 95.1 0.8 -12.7814 correct interpretation 0.8 -12.7814 (represented 0.5 -12.7814 aspect. 0.365385 -12.7815 context vector 0.888889 -12.7816 diseases, 0.888889 -12.7816 larger scale 0.5 -12.7816 (noun 0.404762 -12.7817 edge. 0.8 -12.7817 regions, 0.109797 -12.7821 along 0.8 -12.7821 altogether, 0.454545 -12.7823 to initialize 0.888889 -12.7824 time-to-maturity 0.888889 -12.7824 passives 0.888889 -12.7824 average: 0.888889 -12.7824 “best” 0.888889 -12.7824 navigating 0.75 -12.7827 69.6 0.432432 -12.7827 we utilize 0.611111 -12.7828 first year 0.209877 -12.7831 construct a 0.195652 -12.7832 0.00 0.75 -12.7834 52.4 0.75 -12.7834 multi-view 0.8 -12.7836 θd 0.888889 -12.7837 324 0.888889 -12.7837 structural, 0.888889 -12.7837 constructions: 0.888889 -12.7837 cathedrals 0.888889 -12.7837 edge-flip 0.714286 -12.7839 ngrams, 0.8 -12.784 gross 0.888889 -12.7841 document coding 0.888889 -12.7841 specializes 0.888889 -12.7841 (50 0.482759 -12.7841 dinu and 0.8 -12.7843 bleu[%] 0.888889 -12.7845 marker, 0.888889 -12.7845 forecasts 0.888889 -12.7845 7,000 0.888889 -12.7845 78.90 0.888889 -12.7845 k-dimensional 0.888889 -12.7845 majority class baseline 0.482759 -12.7845 (koo 0.372549 -12.7846 were computed 0.8 -12.7847 410 0.8 -12.7847 po 0.75 -12.7848 47.8 0.75 -12.7848 valencia, 0.888889 -12.7849 cg 0.647059 -12.785 an example of such 0.8 -12.7851 hockenmaier 0.888889 -12.7853 moses-gc 0.888889 -12.7853 hlen 0.52 -12.7853 accuracy scores 0.314286 -12.7854 zero. 0.8 -12.7854 invest 0.75 -12.7855 −2 0.8 -12.7858 competitors 0.8 -12.7858 frequencybased 0.0533439 -12.7862 machine 0.611111 -12.7863 nx 0.611111 -12.7865 naval 0.75 -12.7865 contour 0.75 -12.7865 fm 0.888889 -12.7866 qualification 0.8 -12.7866 compositional, 0.8 -12.7866 melt 0.666667 -12.7867 derivation d 0.611111 -12.7868 cross-covariance 0.541667 -12.7868 agent, 0.365385 -12.7868 these cases 0.647059 -12.7868 promising. 0.8 -12.7869 crf++ 0.888889 -12.787 supplies 0.888889 -12.787 disallow 0.75 -12.7872 hag 0.611111 -12.7874 byrne, 0.888889 -12.7874 autoslog 0.888889 -12.7874 40, 0.714286 -12.7875 games. 0.714286 -12.7875 ideas. 0.714286 -12.7875 cipher. 0.888889 -12.7878 shortage of 0.172996 -12.7879 classes. 0.647059 -12.788 pre-translation 0.8 -12.7881 excited 0.8 -12.7881 the) 0.888889 -12.7882 invoke 0.888889 -12.7882 chile 0.888889 -12.7882 parses: 0.888889 -12.7882 (respectively 0.888889 -12.7882 cedex 0.888889 -12.7882 subject-auxiliary 0.75 -12.7883 “l” 0.8 -12.7884 89.9 0.8 -12.7884 n2, 0.75 -12.7886 (brants, 0.888889 -12.7887 daonly 0.8 -12.7888 universals 0.666667 -12.7889 three characters 0.666667 -12.7889 portugal 0.75 -12.789 52.1 0.888889 -12.7891 306 0.888889 -12.7891 frequency: 0.888889 -12.7891 alternatives: 0.8 -12.7892 dowty 0.888889 -12.7895 (#) 0.245763 -12.7895 street 0.8 -12.7896 earley’s 0.611111 -12.7897 mos 0.75 -12.7897 57.5 0.75 -12.7897 (zhai 0.888889 -12.7899 cross-topic 0.888889 -12.7899 1981) 0.8 -12.7899 yk 0.75 -12.79 biography 0.454545 -12.7902 name mention 0.888889 -12.7903 hierarchical, 0.8 -12.7903 elimination 0.8 -12.7903 metathesaurus 0.8 -12.7903 epistemic 0.888889 -12.7907 discriminative reranking 0.888889 -12.7907 restructuring 0.888889 -12.7907 0.466667 -12.8144 they share 0.8 -12.8146 (source) 0.875 -12.8148 91.1 0.875 -12.8148 ambitious 0.328125 -12.815 which do not 0.875 -12.8152 diminish 0.75 -12.8153 zipf 0.8 -12.8153 machines (svm) 0.545455 -12.8155 pustejovsky 0.75 -12.8156 pennacchiotti 0.875 -12.8156 2006a; 0.875 -12.8156 wizards. 0.875 -12.8156 instruct 0.875 -12.8156 50.5 0.875 -12.8156 .69 0.875 -12.8161 kopp, 0.875 -12.8161 figure 1: illustration 0.75 -12.8163 desc 0.875 -12.8165 lowest-cost 0.875 -12.8165 substitution sites 0.875 -12.8165 spatially 0.875 -12.8165 arities 0.875 -12.8165 (marcu, 0.875 -12.8165 lowbow framework 0.875 -12.8165 against, 0.8 -12.8165 (meaning 0.8 -12.8165 document: 0.875 -12.8169 null-aligned words 0.875 -12.8169 originates 0.875 -12.8169 cross-lingual textual 0.875 -12.8169 51.7 0.666667 -12.8173 merits of 0.875 -12.8173 2005a; 0.875 -12.8173 column indices 0.75 -12.8174 farthest 0.875 -12.8177 patterns) 0.875 -12.8177 backgrounds 0.875 -12.8177 schlesinger, 0.875 -12.8177 reached, 0.875 -12.8177 post-processing step 0.875 -12.8177 1985), 0.571429 -12.8178 algorithm 1: 0.875 -12.8182 bag-of-features 0.875 -12.8182 interoperable 0.875 -12.8182 contextpos 0.875 -12.8182 neglected 0.875 -12.8182 fidelity 0.318182 -12.8183 string, 0.75 -12.8185 discarded, 0.875 -12.8186 tie, 0.875 -12.8186 (pickering 0.875 -12.8186 p-trans (l 0.875 -12.8186 sexually 0.578947 -12.8186 analysis suggests 0.611111 -12.8187 isbn 0.233871 -12.8189 we demonstrate 0.875 -12.819 compiling 0.875 -12.819 0.012 0.875 -12.819 pas¸ca, 0.8 -12.8192 κ, 0.441176 -12.8193 annotation task 0.875 -12.8194 sense-annotated 0.875 -12.8194 incorporated, 0.875 -12.8194 linearization 0.875 -12.8194 lambda-calculus 0.8 -12.8196 connection with 0.875 -12.8198 word-order 0.875 -12.8198 2002)) 0.75 -12.82 purple 0.875 -12.8203 83.1 0.875 -12.8203 table 6 summarizes 0.875 -12.8203 monetary 0.8 -12.8203 263 0.8 -12.8203 wi,j 0.692308 -12.8204 generations 0.75 -12.8207 section 8. 0.875 -12.8207 disentanglement 0.875 -12.8207 0.636 0.875 -12.8207 asymmetry 0.875 -12.8207 0.847 0.875 -12.8207 koch, 0.26 -12.8208 improved by 0.875 -12.8211 tolerance 0.875 -12.8211 home, 0.875 -12.8211 logical-form 0.875 -12.8211 (ando 0.8 -12.8211 ˚ 0.666667 -12.8212 xiao, 0.482759 -12.8213 table 1 presents 0.571429 -12.8214 eng 0.875 -12.8215 permute 0.875 -12.8215 scratch, 0.875 -12.8215 0.007 0.875 -12.8215 covariates 0.875 -12.8215 full-scale 0.692308 -12.8217 now discuss 0.75 -12.8218 danish, 0.8 -12.8219 i.) 0.8 -12.8219 923 0.875 -12.822 uptk 0.875 -12.822 34.4 0.875 -12.822 thesaurus, 0.875 -12.822 samplers 0.875 -12.822 = 0.777778 -12.9326 s=1 0.777778 -12.9326 (specifically, 0.777778 -12.9326 mcrae 0.264368 -12.9327 (2007). 0.875 -12.9327 protocols 0.875 -12.9327 anything, 0.875 -12.9327 dictionary mining 0.588235 -12.9329 description: 0.1375 -12.9329 this, 0.727273 -12.9332 echo 0.875 -12.9332 corroborates 0.875 -12.9332 utf-8 0.875 -12.9332 (20 0.25 -12.9335 (chen 0.875 -12.9336 (through 0.875 -12.9336 consortium, 0.875 -12.9341 bilateral 0.875 -12.9341 (σ 0.875 -12.9341 0.472 0.777778 -12.9343 jk 0.777778 -12.9343 (208) 0.727273 -12.9344 cprime 0.167442 -12.9346 is likely 0.875 -12.9346 decision summarization 0.875 -12.9346 punctuated 0.875 -12.9346 microsoft, 0.875 -12.9346 (amt) 0.777778 -12.9347 dialogs, 0.777778 -12.9347 school, 0.777778 -12.9347 c(p) 0.875 -12.935 colected 0.875 -12.935 socially 0.875 -12.935 justice 0.875 -12.935 query-entity 0.234234 -12.9351 depicted 0.727273 -12.9352 symptoms 0.727273 -12.9352 ontonotes annotation 0.777778 -12.9355 items: 0.777778 -12.9355 ⊕ 0.777778 -12.9355 pioneering 0.777778 -12.9355 undirected-item 0.875 -12.9355 table3: 0.777778 -12.936 domingos 0.875 -12.936 (c, 0.727273 -12.9364 monotrans2 0.777778 -12.9364 condition: 0.777778 -12.9364 (mi,mj) 0.4375 -12.9365 and david 0.875 -12.9365 (dp) 0.875 -12.9365 looser 0.875 -12.9365 6,000 0.11022 -12.9367 table 2 0.777778 -12.9368 meant, 0.777778 -12.9368 right-branching 0.777778 -12.9368 674 0.642857 -12.9368 chest 0.875 -12.9369 (across 0.875 -12.9369 intonational 0.727273 -12.9372 sink 0.727273 -12.9372 3 experiments 3.1 0.777778 -12.9372 (svm), 0.777778 -12.9372 ‘outlier’ 0.777778 -12.9372 s-factor . 1 -12.9619 point). 1 -12.9619 (collobert 1 -12.9619 membership. 1 -12.9619 bi m-l 1 -12.9619 nc. 1 -12.9619 (reg). 1 -12.9619 jeffrey jordan) 1 -12.9619 (alabau 1 -12.9619 callers. 1 -12.9619 post editing 1 -12.9619 respect. 1 -12.9619 precisely. 1 -12.9619 fj. 1 -12.9619 2.4). 1 -12.9619 00. 1 -12.9619 route. 1 -12.9619 oc. 1 -12.9619 4-grams. 1 -12.9619 (2b). 1 -12.9619 ai. 1 -12.9619 zones. 1 -12.9619 sim. 1 -12.9619 afects 1 -12.9619 vsm. 1 -12.9619 spam. 1 -12.9619 conceptnet. 1 -12.9619 (0.0) 1 -12.9619 commonality 1 -12.9619 inspection. 1 -12.9619 debate-side 1 -12.9619 node). 1 -12.9619 normal. 1 -12.9619 equivalence. 1 -12.9619 clark’s (2000) 1 -12.9619 indirect. 1 -12.9619 e-step. 1 -12.9619 (newman 1 -12.9619 op. 1 -12.9619 null0. 1 -12.9619 lives. 1 -12.9619 rt. 1 -12.9619 circumstances. 1 -12.9619 senses), 1 -12.9619 bindings. 1 -12.9619 mention’s 1 -12.9619 predicated 1 -12.9619 borrows 1 -12.9619 maximally ambiguous 1 -12.9619 phrase? 1 -12.9619 clicks. 1 -12.9619 adjuncts. 1 -12.9619 tokenized. 1 -12.9619 (ward, 1 -12.9619 2007b), 1 -12.9619 (204). 1 -12.9619 paroubek, 1 -12.9619 phrasesu 1 -12.9619 merging, 1 -12.9619 robots. 1 -12.9619 icslp, 1 -12.9619 eng. 1 -12.9619 better). 1 -12.9619 realizer. 1 -12.9619 (m`arquez 1 -12.9619 creativity. 1 -12.9619 1930 1 -12.9619 maxent. 1 -12.9619 absolute. 1 -12.9619 motorcycle. 1 -12.9619 termino-ontology. 1 -12.9619 contention. 1 -12.9619 radio station 1 -12.9619 takamura, 1 -12.9619 h max 1 -12.9619 non-projective. 1 -12.9619 fails, 1 -12.9619 eθg-rl∗,d(cη ∗→cηι 1 -12.9619 july. 1 -12.9619 attack. 1 -12.9619 this? 1 -12.9619 asia. 1 -12.9619 distinguished. 1 -12.9619 hyperparameters. 1 -12.9619 generalize. 1 -12.9619 iob-style 1 -12.9619 (blue, 1 -12.9619 predictable. 1 -12.9619 multifaceted 1 -12.9619 binomials. 1 -12.9619 generative. 1 -12.9619 accessible. 1 -12.9619 subject/object 1 -12.9619 highlevel 1 -12.9619 0.000 lsa 1 -12.9619 english/chinese 1 -12.9619 intelligently 1 -12.9619 privacy. 1 -12.9619 life science 1 -12.9619 (. 1 -12.9619 correspondence. 1 -12.9619 : 118,presentval :8.7× 1 -12.9619 movie. 1 -12.9619 andrew mccallum 1 -12.9619 extended. 1 -12.9619 display. 1 -12.9619 (202). 1 -12.9619 (0.2) 1 -12.9619 ?). 1 -12.9619 acyclic. 1 -12.9619 journal. 1 -12.9619 idioms. 1 -12.9619 christof 1 -12.9619 sh. 1 -12.9619 word-tree syntactic 1 -12.9619 lmbot. 1 -12.9619 fern´andez 1 -12.9619 generaldomain 1 -12.9619 firstorder 1 -12.9619 melamed 1 -12.9619 pre-selected 1 -12.9619 size), 1 -12.9619 1 -12.9619 work-place 1 -12.9619 plan. 1 -12.9619 opponents. 1 -12.9619 (p. 1 -12.9619 chai, 1 -12.9619 bcubed ceafe 1 -12.9619 subtilis. 1 -12.9619 journalism 1 -12.9619 state.” 1 -12.9619 output). 1 -12.9619 published. 1 -12.9619 justified. 1 -12.9619 hour. 1 -12.9619 wn. 1 -12.9619 landmarks. 1 -12.9619 cps. 1 -12.9619 roll-outs 1 -12.9619 paris. 1 -12.9619 pathways. 1 -12.9619 developped 1 -12.9619 two-fold. 1 -12.9619 merging. 1 -12.9619 also. 1 -12.9619 ones). 1 -12.9619 ctb5. 1 -12.9619 pos-tag, 1 -12.9619 strong.” 1 -12.9619 suggested. 1 -12.9619 pageviews. 1 -12.9619 shalmaneser 1 -12.9619 flips 1 -12.9619 (niehues 1 -12.9619 compound splitting. 1 -12.9619 nist. 1 -12.9619 entailment). 1 -12.9619 utilized. 1 -12.9619 completions. 1 -12.9619 graehl 1 -12.9619 suits 1 -12.9619 trainee 1 -12.9619 ninomiya 1 -12.9619 attending 1 -12.9619 (heilman 1 -12.9619 τ. 1 -12.9619 voices. 1 -12.9619 suficient 1 -12.9619 ptb-style 1 -12.9619 exploited. 1 -12.9619 wk. 1 -12.9619 word? 1 -12.9619 alternations. 1 -12.9619 corpora). 1 -12.9619 overlapping. 1 -12.9619 ne. 1 -12.9619 harabagiu, 1 -12.9619 posterior. 1 -12.9619 indicated. 1 -12.9619 theoretical linguistics 1 -12.9619 addition. 1 -12.9619 niu 1 -12.9619 corefers 1 -12.9619 coordinates. 1 -12.9619 terminal. 1 -12.9619 hosting 1 -12.9619 (curran 1 -12.9619 wordlda. 1 -12.9619 tutoring episode 1 -12.9619 retrained. 1 -12.9619 know? 1 -12.9619 selectional restrictions. 1 -12.9619 closure. 1 -12.9619 ood. 1 -12.9619 reconcile. 1 -12.9619 (2008)). 1 -12.9619 subsections. 1 -12.9619 sa. 1 -12.9619 oriented. 1 -12.9619 char. 1 -12.9619 slows 1 -12.9619 cambridge: 1 -12.9619 symposium. 1 -12.9619 discrimination. 1 -12.9619 (webber, 1 -12.9619 parliament. 1 -12.9619 ni. 1 -12.9619 pos-tags. 1 -12.9619 ordered. 1 -12.9619 error-correction 1 -12.9619 polysemous. 1 -12.9619 laboratories. 1 -12.9619 errors). 1 -12.9619 106 107 1 -12.9619 ω. 1 -12.9619 thought. 1 -12.9619 segmenters. 1 -12.9619 86.74 1 -12.9619 1997, 1 -12.9619 ”‘bilingual corpora”’ 1 -12.9619 applies. 1 -12.9619 (nelder 1 -12.9619 ap. 1 -12.9619 fluent. 1 -12.9619 1999. 1 -12.9619 episodes. 1 -12.9619 consensus. 1 -12.9619 lehal g.s., 1 -12.9619 induced. 1 -12.9619 transliteration. 1 -12.9619 grads 1 -12.9619 disambiguated. 1 -12.9619 al’s 1 -12.9619 (condition 1 -12.9619 90135 5000 le 1 -12.9619 incorporated. 1 -12.9619 (hastie 1 -12.9619 50-best 1 -12.9619 transcribers, 1 -12.9619 motivated, 1 -12.9619 collins. 1 -12.9619 beijing 100084, 1 -12.9619 bleu). 1 -12.9619 cohesion. 1 -12.9619 disease. 1 -12.9619 mar. 1 -12.9619 v2. 1 -12.9619 13). 1 -12.9619 iphone 3gs 1 -12.9619 category) 1 -12.9619 (lao 1 -12.9619 places. 1 -12.9619 srilm, 1 -12.9619 maximalization. 1 -12.9619 corefer. 1 -12.9619 row). 1 -12.9619 inner loop 1 -12.9619 epps 1 -12.9619 boch’s 1 -12.9619 conclude. 1 -12.9619 widely believed 1 -12.9619 (woz) 1 -12.9619 k). 1 -12.9619 rankings, 1 -12.9619 inside. 1 -12.9619 rewards. 1 -12.9619 a/the 1 -12.9619 pre-computed 1 -12.9619 algorithm). 1 -12.9619 gb. 1 -12.9619 (navigli, 1 -12.9619 nonterminal. 1 -12.9619 instrument. 1 -12.9619 cities. 1 -12.9619 extract. 1 -12.9619 (pdtb) 1 -12.9619 hendrickx 1 -12.9619 ambati 1 -12.9619 dct. 1 -12.9619 tried. 1 -12.9619 2.1). 1 -12.9619 clitics. 1 -12.9619 fλx.f 1 -12.9619 2%. 1 -12.9619 1994. 1 -12.9619 targets) 1 -12.9619 marking. 1 -12.9619 [ilp seg] 1 -12.9619 slower. 1 -12.9619 verbunden. 1 -12.9619 ./. 1 -12.9619 humanities. 1 -12.9619 shamma 1 -12.9619 75.00 1 -12.9619 pictures. 1 -12.9619 sub-parts 1 -12.9619 cross-covariance matrix. 1 -12.9619 κ. 1 -12.9619 red green 1 -12.9619 g2. 1 -12.9619 headline. 1 -12.9619 66.78 1 -12.9619 300. 1 -12.9619 economic growth 1 -12.9619 scores), 1 -12.9619 vorbei:by früh:morning 1 -12.9619 addressee. 1 -12.9619 sentence-internal 1 -12.9619 bracket. 1 -12.9619 call. 1 -12.9619 equation. 1 -12.9619 (yazdani 1 -12.9619 unlabeled. 1 -12.9619 sprime. 1 -12.9619 impossible, 1 -12.9619 belief. 1 -12.9619 france. 1 -12.9619 unpair-t pair-t wilc. 1 -12.9619 objects). 1 -12.9619   1 -12.9619 stated. 1 -12.9619 λx.f 1 -12.9619 sdg 1 -12.9619 lines). 1 -12.9619 mass. 1 -12.9619 concatenation. 1 -12.9619 ny. 1 -12.9619 interrelation 1 -12.9619 (% impr) 1 -12.9619 (negri 1 -12.9619 sub-trees. 1 -12.9619 voluntary 1 -12.9619 n=1 n=2 1 -12.9619 incorrectly. 1 -12.9619 posing 1 -12.9619 sequentially. 1 -12.9619 insignificant. 1 -12.9619 count.pl 1 -12.9619 coalescing 1 -12.9619 reconstruction error. 1 -12.9619 p@1 p@5 1 -12.9619 mark. 1 -12.9619 role-annotated 1 -12.9619 newspaper. 1 -12.9619 archive. 1 -12.9619 attempt, 1 -12.9619 coercion. 1 -12.9619 typed. 1 -12.9619 union. 1 -12.9619 fr. 1 -12.9619 an. 1 -12.9619 epoch. 1 -12.9619 sub-words. 1 -12.9619 interruptions, 1 -12.9619 planer. 1 -12.9619 sense-specific 1 -12.9619 endto-end 1 -12.9619 co-reference. 1 -12.9619 decade. 1 -12.9619 small; 1 -12.9619 npk) 1 -12.9619 blue. 1 -12.9619 2.3). 1 -12.9619 1989. 1 -12.9619 colors. 1 -12.9619 emissions. 1 -12.9619 panel. 1 -12.9619 pause. 1 -12.9619 impasse states 1 -12.9619 ottawa, ontario, 1 -12.9619 stemmer. 1 -12.9619 coordhypermeroattrieventran.nran.jran.v 1 -12.9619 91.4% 1 -12.9619 trigram. 1 -12.9619 i.i.d. 1 -12.9619 affords 1 -12.9619 abstraction orig. 1 -12.9619 .sh/ 1 -12.9619 xu. 1 -12.9619 table1. 1 -12.9619 nonexpert 1 -12.9619 form; 1 -12.9619 prompting 1 -12.9619 candidate’s 1 -12.9619 ccg. 1 -12.9619 multioutput 1 -12.9619 multi-sentence 1 -12.9619 h1. 1 -12.9619 the. 1 -12.9619 instantiations. 1 -12.9619 infoboxes. 1 -12.9619 :: ℓt 1 -12.9619 comp. 1 -12.9619 lower lobe 1 -12.9619 arg. 1 -12.9619 concatenated. 1 -12.9619 predicted/gold 1 -12.9619 lexrank (erkan 1 -12.9619 taskoriented 1 -12.9619 nsf. 1 -12.9619 uniform. 1 -12.9619 (s) prop. 1 -12.9619 linear-time 1 -12.9619 oov. 1 -12.9619 eskenazi. 1 -12.9619 type). 1 -12.9619 (ravichandran 1 -12.9619 confinement. 1 -12.9619 per. 1 -12.9619 tutors, 1 -12.9619 hypergraphs. 1 -12.9619 understandable. 1 -12.9619 cases). 1 -12.9619 x). 1 -12.9619 potentials. 1 -12.9619 little. 1 -12.9619 destination. 1 -12.9619 ψ. 1 -12.9619 portion. 1 -12.9619 unfavorable 1 -12.9619 threshold-based 1 -12.9619 smoothly 1 -12.9619 gazetteers. 1 -12.9619 broadening 1 -12.9619 30%. 1 -12.9619 martn-wanton 1 -12.9619 constructs. 1 -12.9619 tres. 1 -12.9619 confluence. 1 -12.9619 artiles 1 -12.9619 pre-processing. 1 -12.9619 institutions. 1 -12.9619 single-best 1 -12.9619 child-father 1 -12.9619 “much 1 -12.9619 >14. 1 -12.9619 (ams) 1 -12.9619 attractive. 1 -12.9619 non-merge 1 -12.9619 jdmwe. 1 -12.9619 investigations. 1 -12.9619 collocates 1 -12.9619 delimitation 1 -12.9619 (91.9) 1 -12.9619 n2. 1 -12.9619 r1. 1 -12.9619 judgement. 1 -12.9619 1...n. 1 -12.9619 urdu. 1 -12.9619 lrscore. 1 -12.9619 nell. 1 -12.9619 zeros. 1 -12.9619 causes. 1 -12.9619 chih-jen 1 -12.9619 unification, 1 -12.9619 stephen clark 1 -12.9619 λ2, j, 1 -12.9619 sj. 1 -12.9619 sagot, 1 -12.9619 supports. 1 -12.9619 (vs. 1 -12.9619 citation. 1 -12.9619 negative). 1 -12.9619 bp,q 1 -12.9619 phrases). 1 -12.9619 parents. 1 -12.9619 94%. 1 -12.9619 eventually, 1 -12.9619 generality. 1 -12.9619 mance 1 -12.9619 genzel 1 -12.9619 timelines. 1 -12.9619 adjunct. 1 -12.9619 augmentation. 1 -12.9619 uninformative extractions 1 -12.9619 submodels 1 -12.9619 boschee 1 -12.9619 (patwardhan 1 -12.9619 complementary. 1 -12.9619 caution. 1 -12.9619 lexicography. 1 -12.9619 form). 1 -12.9619 predication. 1 -12.9619 posts). 1 -12.9619 turney (2002) 1 -12.9619 exactly. 1 -12.9619 norm. 1 -12.9619 vary. 1 -12.9619 indistinguishable. 1 -12.9619 collins’ 1 -12.9619 lexicon). 1 -12.9619 susceptibility 1 -12.9619 1980. 1 -12.9619 thatcher. 1 -12.9619 interpretable. 1 -12.9619 charts. 1 -12.9619 vidal, 1 -12.9619 pubmed. 1 -12.9619 breath. 1 -12.9619 biases. 1 -12.9619 encoded. 1 -12.9619 non-terminal. 1 -12.9619 nouns). 1 -12.9619 fusion. 1 -12.9619 (§3). 1 -12.9619 obo 1 -12.9619 skills. 1 -12.9619 29.73 1 -12.9619 england. 1 -12.9619 (mohri 1 -12.9619 unreliable. 1 -12.9619 highdimensional 1 -12.9619 intention. 1 -12.9619 compensation 1 -12.9619 wine. 1 -12.9619 windowbased 1 -12.9619 ace. 1 -12.9619 anything. 1 -12.9619 70%. 1 -12.9619 poesio. 1 -12.9619 funny_day 1 -12.9619 f-structure 1 -12.9619 salient. 1 -12.9619 subj. 1 -12.9619 lexeme-based 1 -12.9619 (203). 1 -12.9619 nse t 1 -12.9619 anyway. 1 -12.9619 clusters; 1 -12.9619 refinement. 1 -12.9619 saldo. 1 -12.9619 optimized. 1 -12.9619 descriptors. 1 -12.9619 playing. 1 -12.9619 performance? 1 -12.9619 predictive. 1 -12.9619 paroubek 1 -12.9619 e. boschee, 1 -12.9619 2004)). 1 -12.9619 override 1 -12.9619 non-contentful 1 -12.9619 modelling. 1 -12.9619 humanhuman 1 -12.9619 acquired. 1 -12.9619 mentions). 1 -12.9619 assists 1 -12.9619 statisticaly 1 -12.9619 disabilities, 1 -12.9619 133, 91403 1 -12.9619 5.2). 1 -12.9619 meeting. 1 -12.9619 check. 1 -12.9619 np-complete. 0.777778 -12.962 ρ2 0.777778 -12.962 4.4.1 0.433333 -12.962 similarity values 0.5 -12.9621 exclamation 0.642857 -12.9622 st, 0.857143 -12.9624 phone. 0.857143 -12.9624 integer. 0.857143 -12.9624 hyperedge. 0.857143 -12.9624 ilp. 0.857143 -12.9624 audio. 0.857143 -12.9624 checked. 0.857143 -12.9624 populations. 0.857143 -12.9624 domain). 0.857143 -12.9624 .05). 0.857143 -12.9624 structured sparsity. 0.857143 -12.9624 uttered. 0.857143 -12.9624 extractors, 0.857143 -12.9624 log-likelihood. 0.857143 -12.9624 lower-level 0.857143 -12.9624 entity linking. 0.857143 -12.9624 later). 0.857143 -12.9624 started. 0.857143 -12.9624 diverges 0.857143 -12.9624 lu. 0.857143 -12.9624 time consuming, 0.857143 -12.9624 28.08 0.857143 -12.9624 crowds. 0.857143 -12.9624 phonetic transcription 0.857143 -12.9624 intl. 0.857143 -12.9624 competence. 0.857143 -12.9624 sub-structures 0.857143 -12.9624 graph; 0.857143 -12.9624 size). 0.857143 -12.9624 triggering policies 0.857143 -12.9624 loss functions. 0.857143 -12.9624 run-time, 0.857143 -12.9624 brown. 0.857143 -12.9624 p/r 0.857143 -12.9624 sc (all 0.857143 -12.9624 log-linear interpolation 0.857143 -12.9624 advice. 0.857143 -12.9624 recognized. 0.857143 -12.9624 strong. 0.857143 -12.9624 methods). 0.857143 -12.9624 hmmbased 0.857143 -12.9624 hickl 0.461538 -12.9624 built. 0.777778 -12.9624 estonian 0.6 -12.9625 to escape 0.777778 -12.9628 coach 0.857143 -12.9629 tags). 0.857143 -12.9629 standard practice, 0.857143 -12.9629 metaphorically 0.857143 -12.9629 downhill 0.857143 -12.9629 twets. 0.857143 -12.9629 (1-a) 0.857143 -12.9629 80% 20% 0.857143 -12.9629 1]. 0.857143 -12.9629 4.2) 0.857143 -12.9633 (miller, 1995). 0.857143 -12.9633 reduce transitions 0.777778 -12.9633 qualitative differences 0.777778 -12.9637 ik 0.857143 -12.9638 dbr 0.857143 -12.9638 mu li, 0.857143 -12.9638 ng+lex+par 0.666667 -12.9638 oxford university 0.666667 -12.9641 the chinese-side 0.666667 -12.9641 regressions 0.857143 -12.9643 visual stimuli 0.857143 -12.9643 radev, 2004) 0.857143 -12.9643 geographical names 0.857143 -12.9643 multiword unit 0.727273 -12.9643 nf 0.285714 -12.9643 semantic relation 0.857143 -12.9648 deepwater horizon 0.857143 -12.9648 light verb. 0.857143 -12.9648 (pos) tags, 0.857143 -12.9648 ciphers, 0.857143 -12.9648 gradient-free 0.857143 -12.9648 noord 0.857143 -12.9648 social networks, 0.857143 -12.9648 (kobdani 0.857143 -12.9648 venugopal (2006) 0.247423 -12.9648 tagset 0.666667 -12.9649 complicated. 0.777778 -12.965 tweet text 0.777778 -12.965 landing 0.666667 -12.9652 aligns with 0.857143 -12.9653 ilp solver 0.642857 -12.9654 switchtest 0.6 -12.9655 will address 0.6 -12.9655 dance 0.170854 -12.9655 values. 0.857143 -12.9657 specially designed 0.857143 -12.9657 open re 0.857143 -12.9657 speakingrate 0.777778 -12.9659 externally 0.777778 -12.9659 identically to 0.666667 -12.966 wedding 0.666667 -12.966 89.6 0.857143 -12.9662 43.7 0.857143 -12.9662 0.936 0.857143 -12.9662 6.2) 0.555556 -12.9665 scbr 0.857143 -12.9667 aligner (liang 0.857143 -12.9667 jose, ca 0.857143 -12.9667 pay attention 0.857143 -12.9667 43.9 0.857143 -12.9667 improve, 0.857143 -12.9667 won, 0.857143 -12.9667 third stage 0.857143 -12.9667 publically available 0.424242 -12.9668 10000 0.666667 -12.9671 397 0.857143 -12.9672 androutsopoulos 0.857143 -12.9672 ksummationdisplay i=1 0.727273 -12.9672 da sequences 0.727273 -12.9672 this dataset contains 0.777778 -12.9672 quantifier, 0.6 -12.9672 brian 0.857143 -12.9676 apt 0.857143 -12.9676 agichtein 0.857143 -12.9676 (riloff, 0.777778 -12.9676 requirement, 0.6 -12.9679 ming 0.727273 -12.968 very complex 0.777778 -12.968 1980 0.777778 -12.968 testset 0.777778 -12.968 spike 0.777778 -12.968 genia track 0.857143 -12.9681 abend 0.857143 -12.9681 rbf kernel 0.666667 -12.9682 (shi 0.233645 -12.9683 from multiple 0.727273 -12.9684 stefan 0.857143 -12.9686 (§3) 0.857143 -12.9686 never-ending language 0.857143 -12.9686 schler, 0.857143 -12.9686 russian-english 0.857143 -12.9686 tesla-b, 0.857143 -12.9686 particularly suited 0.857143 -12.9686 officially 0.857143 -12.9686 spin 0.777778 -12.9689 jon 0.777778 -12.9689 (product 0.777778 -12.9689 486 0.857143 -12.9691 (or negative) 0.857143 -12.9691 uniform initialization 0.857143 -12.9696 transliteration module 0.857143 -12.9696 77.8% 0.857143 -12.9696 hwa et al. 0.857143 -12.9696 density values 0.555556 -12.9696 woz-07 0.555556 -12.9696 word cloud 0.461538 -12.9696 it follows 0.857143 -12.97 firstand 0.857143 -12.97 0.838 0.857143 -12.97 [a1 → 0.857143 -12.97 it-timeml 0.666667 -12.9701 (26) 0.52381 -12.9701 uppsala 0.642857 -12.9701 preferred by 0.363636 -12.9702 assumption, 0.777778 -12.9702 (entity 0.777778 -12.9702 owners 0.6 -12.9702 joint query 0.857143 -12.9705 (yates 0.857143 -12.9705 favourable 0.857143 -12.9705 bve, 0.857143 -12.9705 general safety 0.777778 -12.9707 concern, 0.777778 -12.9707 21st 0.727273 -12.9708 (word) 0.857143 -12.971 knowledge; 0.857143 -12.971 evaluation purposes. 0.857143 -12.971 monz 0.857143 -12.971 sammons 0.857143 -12.9715 available) 0.857143 -12.9715 lattice) 0.857143 -12.9715 special links 0.777778 -12.9715 p0, 0.175532 -12.9715 are able 0.6 -12.9716 takamura 0.857143 -12.972 gamon et al. 0.857143 -12.972 negotiating 0.857143 -12.972 49.2 0.857143 -12.972 indepth 0.857143 -12.972 inventories, 0.666667 -12.972 unless it 0.727273 -12.9721 r-su4 0.6 -12.9722 finkel et 0.289855 -12.9723 similar, 0.857143 -12.9724 genlex 0.857143 -12.9724 shivakumar vaithyanathan. 0.857143 -12.9724 hypotheses: 0.857143 -12.9724 adept 0.311475 -12.9724 kim and 0.727273 -12.9725 three steps: 0.777778 -12.9728 a127a98a38a114a31a133a97a134a107 0.857143 -12.9729 100) 0.244898 -12.9732 but only 0.433333 -12.9732 “as 0.777778 -12.9733 french: 0.857143 -12.9734 38.2 0.857143 -12.9734 second-level 0.857143 -12.9734 (xue, 0.857143 -12.9734 0.782 0.727273 -12.9737 29, 0.642857 -12.9737 disfluencies, 0.857143 -12.9739 interrogation 0.857143 -12.9739 graduate center 0.857143 -12.9739 decoupling 0.857143 -12.9739 1991 0.857143 -12.9739 residing 0.727273 -12.9741 mgs 0.727273 -12.9741 pennebaker, 0.666667 -12.9742 theoretic 0.394737 -12.9742 blunsom 0.448276 -12.9742 richness 0.555556 -12.9743 diacritics 0.857143 -12.9744 ...”, 0.857143 -12.9744 cross-sentence 0.857143 -12.9744 ddtk 0.642857 -12.9744 females 0.26506 -12.9745 argument. 0.52381 -12.9745 scores for all 0.777778 -12.9746 uni 0.6 -12.9746 universit´e 0.588235 -12.9747 we associate 0.5 -12.9747 by varying 0.857143 -12.9749 relation) 0.857143 -12.9749 non-return 0.857143 -12.9749 sourcecontext 0.857143 -12.9749 validity, 0.14 -12.975 policy 0.375 -12.9751 nullnullnullnull 0.555556 -12.9752 .36 0.857143 -12.9753 radiation 0.857143 -12.9753 details), 0.857143 -12.9753 points) 0.857143 -12.9753 unacceptable 0.857143 -12.9753 1082 0.857143 -12.9753 to; 0.727273 -12.9753 disambiguations 0.222222 -12.9753 semantic information 0.555556 -12.9756 keep only 0.727273 -12.9757 waste 0.727273 -12.9757 z(x) 0.857143 -12.9758 2004a; 0.857143 -12.9758 annotated) 0.857143 -12.9758 duced 0.384615 -12.9758 spite of 0.777778 -12.9759 for each category, 0.48 -12.9761 they require 0.5 -12.9761 s0 and 0.727273 -12.9762 metric: 0.727273 -12.9762 longer dependencies 0.857143 -12.9763 0.023 0.857143 -12.9763 ⊕r 0.857143 -12.9763 spanish-english, 0.857143 -12.9763 sentenceand 0.857143 -12.9763 welldefined 0.315789 -12.9764 we suggest 0.727273 -12.9766 twitter data, 0.136508 -12.9767 develop 0.857143 -12.9768 depended 0.857143 -12.9768 unimportant 0.857143 -12.9768 markert 0.777778 -12.9768 v) 0.411765 -12.9769 phrase alignment 0.52381 -12.9769 illustrate this 0.52381 -12.9769 navigli, 0.0926471 -12.9771 science 0.555556 -12.9771 shieber 0.857143 -12.9773 1995 0.857143 -12.9773 two-class 0.857143 -12.9773 client, 0.857143 -12.9778 labels; 0.857143 -12.9778 collapsed sampling 0.857143 -12.9778 tables) 0.857143 -12.9778 hand-built 0.138158 -12.978 wide 0.6 -12.978 term recognition 0.777778 -12.9781 dl 0.857143 -12.9782 unexpected, 0.857143 -12.9782 language impairment 0.857143 -12.9782 dependency relationships. 0.857143 -12.9782 0.464 0.857143 -12.9782 localmax 0.24 -12.9783 this problem, 0.857143 -12.9787 42.3 0.857143 -12.9787 key/value 0.857143 -12.9787 distorted 0.857143 -12.9787 .12 0.727273 -12.979 inline 0.777778 -12.979 nully 0.857143 -12.9792 malakasiotis, 0.857143 -12.9792 0.6% 0.857143 -12.9792 side) 0.857143 -12.9792 rftagger 0.857143 -12.9792 coverages 0.857143 -12.9792 0.738 0.588235 -12.9796 twofold: 0.857143 -12.9797 sults 0.857143 -12.9797 argamon, 0.857143 -12.9797 morphologically-rich 0.857143 -12.9797 prospective 0.857143 -12.9797 0.565 0.857143 -12.9797 (juola, 0.857143 -12.9797 .89 0.857143 -12.9797 stent, 0.857143 -12.9797 versley 0.24 -12.98 x = 0.857143 -12.9802 reset 0.857143 -12.9802 28.5 0.857143 -12.9802 (mirkin 0.857143 -12.9802 lexratio 0.666667 -12.9802 39.6 0.727273 -12.9803 arnold 0.727273 -12.9803 an inverted 0.727273 -12.9803 2004): 0.5 -12.9806 to improve performance 0.857143 -12.9807 (schwartz 0.857143 -12.9807 prefixed 0.857143 -12.9807 runtime, 0.857143 -12.9807 desiderata 0.857143 -12.9807 bosch, 0.857143 -12.9807 big, 0.857143 -12.9807 alike 0.777778 -12.9807 ,e 0.777778 -12.9807 ie, 0.6 -12.9807 0.002 0.857143 -12.9811 trigrams) 0.857143 -12.9811 wsj10 0.857143 -12.9811 entailment pairs, 0.857143 -12.9811 language-independent, 0.26506 -12.9815 is clearly 0.857143 -12.9816 friendship 0.857143 -12.9816 21.9 0.857143 -12.9816 experiments) 0.857143 -12.9816 additional challenges 0.642857 -12.9817 yi, 0.48 -12.9818 field, 0.321429 -12.9819 right. 0.52381 -12.9819 annotation cost 0.6 -12.982 94.4 0.857143 -12.9821 source cept 0.857143 -12.9821 “click” 0.857143 -12.9821 lignos 0.857143 -12.9821 9.4% 0.857143 -12.9821 ped30-c 0.857143 -12.9821 319 0.777778 -12.9821 calculated, 0.166667 -12.9822 inspired 0.237624 -12.9822 4 experiments 0.857143 -12.9826 culture-specific 0.857143 -12.9826 uncertain, 0.857143 -12.9826 end-of-sentence 0.857143 -12.9826 extraction (re) 0.666667 -12.9829 large pool 0.306452 -12.983 one such 0.857143 -12.9831 definition questions. 0.857143 -12.9831 wagner 0.857143 -12.9831 attaining 0.857143 -12.9831 biemann 0.857143 -12.9831 ped30-p 0.857143 -12.9831 crime 0.857143 -12.9831 40.52 0.727273 -12.9831 dmoz 0.276316 -12.9832 following two 0.777778 -12.9834 “there 0.777778 -12.9834 99.1 0.375 -12.9835 0.96 0.857143 -12.9836 (zanzotto 0.857143 -12.9836 near future. 0.857143 -12.9836 nominals, 0.857143 -12.9836 rewards, 0.857143 -12.9836 40.5 0.857143 -12.9836 rehbein 0.34 -12.9836 full model 0.666667 -12.984 canada. 0.52381 -12.984 argumentation and 0.857143 -12.9841 truncating 0.857143 -12.9841 lexical: 0.857143 -12.9841 barzilay, 2006; 0.857143 -12.9841 unexplored 0.857143 -12.9841 favorite 0.857143 -12.9841 fscores 0.857143 -12.9841 77.0 0.857143 -12.9841 congress, 0.857143 -12.9841 one-sided 0.857143 -12.9841 term-label pair 0.857143 -12.9846 unsolved 0.857143 -12.9846 concession 0.857143 -12.9846 videos, 0.857143 -12.9846 59.0 0.857143 -12.9846 correspondingly, 0.777778 -12.9847 negativity 0.727273 -12.9848 validated by 0.857143 -12.985 (zesch 0.857143 -12.985 clarifications 0.777778 -12.9851 employed, 0.777778 -12.9851 position: 0.642857 -12.9853 reduced, 0.857143 -12.9855 win4 0.857143 -12.9855 locus 0.857143 -12.9855 epochs, 0.857143 -12.9855 jiampojamarn 0.857143 -12.9855 words] 0.857143 -12.9855 plates 0.857143 -12.9855 news10 0.666667 -12.9855 favre, 0.6 -12.9858 loi 0.666667 -12.9859 sign languages 0.857143 -12.986 (grenager 0.857143 -12.986 basili 0.857143 -12.986 rus, 0.857143 -12.986 recognizers, 0.857143 -12.986 46.1 0.857143 -12.986 recency 0.4 -12.9864 clothing 0.857143 -12.9865 (marneffe 0.857143 -12.9865 grouping-based 0.857143 -12.9865 tag sequences, 0.777778 -12.9865 100-best 0.727273 -12.9869 1-2 0.857143 -12.987 centrality. 0.857143 -12.987 7.5% 0.857143 -12.987 0.639 0.857143 -12.987 intricate 0.857143 -12.987 0.724 0.857143 -12.987 mds 0.857143 -12.987 (feng 0.411765 -12.987 table 6 shows the 0.461538 -12.9871 exploit this 0.296875 -12.9872 discriminate 0.857143 -12.9875 (ie), 0.857143 -12.9875 protocol, 0.857143 -12.9875 (lms) 0.857143 -12.988 lintean 0.857143 -12.988 accelerates 0.857143 -12.988 ratnaparkhi, 0.857143 -12.988 resolution: 0.857143 -12.988 cleared 0.857143 -12.988 outof-vocabulary 0.172775 -12.9882 an approach 0.857143 -12.9885 callan, 0.857143 -12.9885 competitive, 0.857143 -12.9885 0.603 0.857143 -12.9885 (white 0.857143 -12.9885 philadelphia, pa 0.6 -12.9885 /aba/ 0.642857 -12.9886 crammer 0.289855 -12.9887 (np 0.857143 -12.9889 ated 0.857143 -12.9889 task-based evaluation 0.857143 -12.9889 portuguese: 0.857143 -12.9889 15.5 0.666667 -12.9889 grades 0.857143 -12.9894 items) 0.857143 -12.9894 credible 0.857143 -12.9894 signi 0.857143 -12.9894 data-to-text 0.857143 -12.9894 adaptability 0.857143 -12.9894 enlarging 0.857143 -12.9894 (mdl) 0.727273 -12.9898 claims. 0.857143 -12.9899 85.96 0.857143 -12.9899 distributing 0.857143 -12.9899 complementary, 0.857143 -12.9899 j-prf, 0.857143 -12.9899 ji, 0.857143 -12.9899 37.3 0.857143 -12.9899 investigation, 0.857143 -12.9904 flattened 0.857143 -12.9904 0.381 0.857143 -12.9904 0.804 0.857143 -12.9904 decompounding 0.857143 -12.9904 typebased 0.857143 -12.9904 35.3 0.195946 -12.9904 strength 0.727273 -12.9906 center. 0.857143 -12.9909 (local 0.857143 -12.9909 successors 0.857143 -12.9909 ullman, 0.857143 -12.9909 mannem 0.857143 -12.9909 /b/, 0.857143 -12.9909 landmarks 0.857143 -12.9909 44.2 0.365854 -12.9911 symbols. 0.375 -12.9912 weaknesses 0.857143 -12.9914 0.015 0.857143 -12.9914 centers 0.857143 -12.9914 entirety 0.857143 -12.9914 (baldwin 0.857143 -12.9914 georgila 0.857143 -12.9914 operationalize 0.857143 -12.9914 triple-gold 0.857143 -12.9914 jindal 0.857143 -12.9914 evolution, 0.857143 -12.9914 injured 0.857143 -12.9914 jaeger’s 0.857143 -12.9919 toolkits 0.857143 -12.9919 fairer 0.6 -12.992 “all 0.727273 -12.9923 positive feedback 0.727273 -12.9923 ground, 0.857143 -12.9924 naist 0.857143 -12.9924 (federico 0.857143 -12.9924 potentially, 0.857143 -12.9924 dbpedia relation 0.857143 -12.9924 novice 0.857143 -12.9924 5.1) 0.666667 -12.9924 (columns 0.727273 -12.9927 t1. 0.777778 -12.9927 subjectivity, 0.52381 -12.9927 transitioning 0.857143 -12.9929 (tateisi 0.857143 -12.9929 juman 0.857143 -12.9929 node’s 0.857143 -12.9929 phosphorylation, 0.857143 -12.9929 2.3.1 0.857143 -12.9929 quarreling 0.857143 -12.9929 19.3 0.857143 -12.9929 wouldn’t 0.857143 -12.9929 apriori 0.6 -12.9933 prediction: 0.857143 -12.9934 fetched 0.857143 -12.9934 scenarios: 0.857143 -12.9934 nadya 0.857143 -12.9934 encyclopedias 0.857143 -12.9934 13.0 0.857143 -12.9934 fixed 0.857143 -12.9938 cent 0.857143 -12.9938 user-friendly 0.857143 -12.9938 liberman, 0.857143 -12.9938 coresponds to 0.857143 -12.9938 0.398 0.857143 -12.9938 encoded, 0.857143 -12.9938 pos2 0.857143 -12.9938 φk 0.857143 -12.9938 (1991), 0.857143 -12.9938 en-detection 0.857143 -12.9938 substring alignment 0.857143 -12.9938 breck 0.777778 -12.994 one option 0.857143 -12.9943 sundance 0.857143 -12.9943 (boyd-graber 0.857143 -12.9943 prior; 0.857143 -12.9943 0.411 0.857143 -12.9943 4.5% 0.857143 -12.9943 (mohri, 0.857143 -12.9943 loop, 0.857143 -12.9943 logic-based 0.857143 -12.9943 shorten 0.310345 -12.9943 itself is 0.777778 -12.9945 voudrais 0.857143 -12.9948 histograms, 0.857143 -12.9948 callan 0.857143 -12.9948 top-100 0.857143 -12.9948 noun-verb 0.857143 -12.9948 diameter 0.857143 -12.9948 achievements 0.857143 -12.9948 sinha 0.0641483 -12.9949 need 0.857143 -12.9953 ibm-mac 0.857143 -12.9953 0.773 0.857143 -12.9953 (pan 0.857143 -12.9953 their applicability 0.857143 -12.9953 two steps, 0.857143 -12.9953 32.4 0.857143 -12.9953 jerry 0.857143 -12.9953 elliptical 0.777778 -12.9954 for) 0.6 -12.9954 mention boundaries 0.857143 -12.9958 mammals 0.857143 -12.9958 em; 0.857143 -12.9958 wcj 0.857143 -12.9958 deploying 0.857143 -12.9958 velldal 0.384615 -12.9958 by extracting 0.110169 -12.996 parts 0.205882 -12.9961 relying 0.666667 -12.9962 scientific research 0.857143 -12.9963 after: 0.857143 -12.9963 table 2 illustrates 0.857143 -12.9963 2.4% 0.857143 -12.9963 35.5 0.857143 -12.9963 non-random 0.857143 -12.9963 inject 0.666667 -12.9966 41.9 0.777778 -12.9967 review helpfulness 0.857143 -12.9968 teufel, 0.857143 -12.9968 rwr 0.857143 -12.9968 96.7 0.857143 -12.9968 50.9 0.857143 -12.9968 single-class 0.857143 -12.9968 sco 0.333333 -12.9969 action. 0.6 -12.9971 emerges 0.777778 -12.9972 keith 0.857143 -12.9973 0.640 0.857143 -12.9973 storage, 0.857143 -12.9973 cross-lingual/interlingual 0.857143 -12.9973 nullable 0.857143 -12.9973 isues 0.857143 -12.9973 graphemes 0.857143 -12.9973 mcmc-based 0.857143 -12.9973 alignment: 0.857143 -12.9973 blend 0.857143 -12.9973 (2008)) 0.857143 -12.9973 they’re 0.857143 -12.9973 decoder’s 0.388889 -12.9974 corpus, but 0.52381 -12.9975 [conceived 0.857143 -12.9978 rp, 0.857143 -12.9978 c′ 0.857143 -12.9978 0.761 0.857143 -12.9978 repetitions, 0.857143 -12.9978 fusions 0.857143 -12.9978 beamsearch 0.857143 -12.9978 crosscultural 0.857143 -12.9978 oepen, 0.857143 -12.9978 0.389 0.857143 -12.9978 ap-closure 0.25 -12.9978 literature. 0.433333 -12.9978 an underlying 0.461538 -12.9979 nissim’s 0.857143 -12.9983 simplified: 0.857143 -12.9983 line), 0.857143 -12.9983 non-bottom-up 0.857143 -12.9983 prod-rule 0.857143 -12.9983 you mean 0.857143 -12.9983 hypothesised 0.857143 -12.9983 608 0.857143 -12.9983 36.9 0.857143 -12.9983 holders, 0.857143 -12.9983 campbell, 0.333333 -12.9983 tells 0.857143 -12.9988 frame-to-frame 0.857143 -12.9988 caution 0.857143 -12.9988 katakana, 0.857143 -12.9988 hewlett 0.857143 -12.9988 collaborate 0.857143 -12.9988 instance’s 0.857143 -12.9988 dataskdelay adjpair 0.857143 -12.9988 duplicating 0.258824 -12.999 annotations are 0.857143 -12.9993 general) 0.857143 -12.9993 better) 0.857143 -12.9993 depart 0.857143 -12.9993 skewed, 0.156118 -12.9994 type. 0.526316 -12.9995 gs 0.5 -12.9995 interlingual 0.857143 -12.9998 term-document 0.857143 -12.9998 purposefully 0.857143 -12.9998 clause-level 0.857143 -13.0003 suspicion 0.857143 -13.0003 exponent 0.857143 -13.0003 agreement: 0.857143 -13.0003 atm 0.857143 -13.0003 0.016 0.857143 -13.0003 58.9 0.777778 -13.0003 king, 0.777778 -13.0003 str(a,i) 0.526316 -13.0004 determined using 0.555556 -13.0006 herein 0.6 -13.0006 tokens; 0.857143 -13.0008 acute 0.857143 -13.0008 wire 0.857143 -13.0008 deleted, 0.857143 -13.0008 refines 0.857143 -13.0008 apear 0.857143 -13.0008 non-filtered 0.857143 -13.0008 voting, 0.857143 -13.0008 (gordon 0.857143 -13.0008 help alleviate 0.857143 -13.0008 subsquare 0.857143 -13.0008 put, 0.461538 -13.0008 distinct from 0.857143 -13.0013 24.9 0.857143 -13.0013 highlighting, 0.857143 -13.0013 548 0.857143 -13.0013 component; 0.4 -13.0015 strube, 0.310345 -13.0016 two. 0.857143 -13.0017 mistaken 0.857143 -13.0017 disciplines 0.857143 -13.0017 revising 0.857143 -13.0017 (ravi 0.857143 -13.0017 primary advantage 0.555556 -13.0019 be traced 0.296875 -13.0019 a document, 0.355556 -13.0021 parameter. 0.444444 -13.0021 nullnullnullnullnullnullnull 0.857143 -13.0022 calendar 0.857143 -13.0022 61.5% 0.857143 -13.0022 ool 0.526316 -13.0022 quite similar 0.52381 -13.0023 they represent 0.555556 -13.0025 runtime. 0.777778 -13.0025 316 0.857143 -13.0027 uralic 0.857143 -13.0027 ff 0.857143 -13.0027 1983), 0.857143 -13.0027 rebholz-schuhmann 0.857143 -13.0027 wa, 0.365854 -13.0029 figure, 0.419355 -13.003 it performs 0.777778 -13.0034 50.6 0.777778 -13.0034 j¸@ 0.857143 -13.0037 saccades 0.857143 -13.0037 rate) 0.857143 -13.0037 reconstructs 0.857143 -13.0037 paraphraser 0.857143 -13.0037 pinochet 0.857143 -13.0037 singletons, 0.857143 -13.0037 cross-entropy, 0.0745402 -13.0038 least 0.433333 -13.0038 when combined 0.777778 -13.0039 (n\n)/np 0.857143 -13.0042 corpus’ 0.857143 -13.0042 fine-grained, 0.857143 -13.0042 47.9 0.857143 -13.0042 1.25 0.857143 -13.0042 prohibits 0.388889 -13.0042 ontology, 0.777778 -13.0043 it falls 0.777778 -13.0043 raced 0.857143 -13.0047 0.041 0.857143 -13.0047 implicit, 0.857143 -13.0047 16.9 0.857143 -13.0047 kidnap 0.857143 -13.0047 0.452 0.857143 -13.0047 nominalized 0.198582 -13.0047 laboratory 0.296875 -13.0048 advances 0.6 -13.0051 rel-lda 0.857143 -13.0052 time-out 0.857143 -13.0052 (me) 0.857143 -13.0052 samples) 0.777778 -13.0053 ability, 0.777778 -13.0053 anthony 0.857143 -13.0057 pch 0.857143 -13.0057 coercive 0.857143 -13.0057 syntacticosemantic 0.857143 -13.0057 single-reference 0.857143 -13.0057 yule+match 0.857143 -13.0057 1954) 0.857143 -13.0057 0.755 0.857143 -13.0057 4.1.3 0.857143 -13.0057 bpcs 0.857143 -13.0057 multiple-choice 0.857143 -13.0057 wn’s 0.777778 -13.0057 367 0.526316 -13.0058 easy–to–read 0.388889 -13.006 unchanged 0.857143 -13.0062 relex 0.857143 -13.0062 one: 0.857143 -13.0062 age group. 0.857143 -13.0062 subdialogue 0.857143 -13.0062 define, 0.857143 -13.0062 (carletta, 0.857143 -13.0062 words (those 0.857143 -13.0062 silent 0.116945 -13.0063 column 0.666667 -13.0066 berkeleyparser 0.857143 -13.0067 morante 0.857143 -13.0067 (linear 0.857143 -13.0067 heafield 0.857143 -13.0072 self-contained 0.857143 -13.0072 role) 0.857143 -13.0072 mis 0.857143 -13.0072 (local) 0.857143 -13.0072 (krippendorff, 0.857143 -13.0072 earnings 0.857143 -13.0072 449 0.857143 -13.0072 bcb-s 0.777778 -13.0075 moo 0.4 -13.0076 york times 0.857143 -13.0077 conected 0.857143 -13.0077 (demir 0.857143 -13.0077 wade 0.857143 -13.0077 annotators: 0.857143 -13.0077 68.4% 0.857143 -13.0077 overgeneration 0.315789 -13.008 not found 0.777778 -13.008 (b) select 0.666667 -13.0081 rule out 0.857143 -13.0082 sourcelanguage 0.857143 -13.0082 0.557 0.857143 -13.0082 diagonal, 0.857143 -13.0082 negatives: 0.857143 -13.0082 word-internal 0.857143 -13.0082 (noisy) 0.857143 -13.0082 filtering techniques 0.857143 -13.0082 lift 0.5 -13.0084 temporal expression 0.666667 -13.0085 ubl 0.305085 -13.0086 for future work 0.857143 -13.0087 first-stage 0.857143 -13.0087 zh-en 0.857143 -13.0087 17.1 0.857143 -13.0087 sequences) 0.857143 -13.0087 trees; 0.0689369 -13.0088 described in 0.777778 -13.0089 of  0.277778 -13.0089 are connected 0.388889 -13.0091 we then use 0.857143 -13.0092 consecutively 0.857143 -13.0092 0.729 0.857143 -13.0092 accumulating 0.857143 -13.0092 nodes; 0.857143 -13.0092 2.8% 0.857143 -13.0092 tales have 0.857143 -13.0092 accuracy; 0.666667 -13.0093 (go 0.777778 -13.0093 verbalizer 0.6 -13.0093 labs 0.857143 -13.0097 word’s end 0.857143 -13.0097 problem) 0.857143 -13.0097 trunk, 0.857143 -13.0097 avatars 0.0863874 -13.01 appear 0.857143 -13.0102 (ig) 0.857143 -13.0102 2.3, 0.857143 -13.0102 ‘a’ 0.857143 -13.0102 8-way 0.857143 -13.0102 sizeable 0.857143 -13.0102 18.7 0.857143 -13.0102 taboada 0.857143 -13.0102 anchor, 0.857143 -13.0102 selection criteria 0.857143 -13.0102 383 0.857143 -13.0102 2.09 0.777778 -13.0102 conceptually similar 0.258824 -13.0103 were extracted 0.555556 -13.0106 oi 0.857143 -13.0107 bridges 0.857143 -13.0107 clinical texts 0.857143 -13.0107 detection; 0.526316 -13.0107 a formalism 0.305085 -13.0107 effective at 0.461538 -13.0107 also contain 0.777778 -13.0107 0.75 0.8 0.526316 -13.0111 的 0.433333 -13.0111 martha 0.857143 -13.0112 frozen 0.857143 -13.0112 48.1 0.857143 -13.0112 tlinks 0.857143 -13.0112 functioning 0.857143 -13.0112 barry 0.857143 -13.0112 36.4 0.857143 -13.0112 terminated 0.320755 -13.0112 cube 0.857143 -13.0117 neded 0.857143 -13.0117 64.7 0.857143 -13.0117 pathway representations 0.857143 -13.0117 sub-phrases 0.857143 -13.0117 backtrack 0.857143 -13.0117 slot values 0.252874 -13.0118 probabilities are 0.777778 -13.012 hk 0.857143 -13.0122 (von 0.857143 -13.0122 55.13 0.857143 -13.0122 umls semantic 0.857143 -13.0122 romanian, 0.310345 -13.0123 alignments are 0.666667 -13.0124 letter transformation 0.777778 -13.0125 confirming that 0.777778 -13.0125 documents retrieved 0.857143 -13.0127 reversibility 0.857143 -13.0127 splits, 0.857143 -13.0127 32.41 0.857143 -13.0127 0.592 0.777778 -13.013 (argamon et 0.444444 -13.013 to solve this 0.277778 -13.013 by looking 0.857143 -13.0132 goals: 0.857143 -13.0132 rhythm 0.857143 -13.0132 vv+uc 0.857143 -13.0132 elaborations 0.857143 -13.0132 mixed, 0.857143 -13.0132 21.8 0.857143 -13.0132 wrap 0.857143 -13.0137 incremented 0.857143 -13.0137 “common 0.857143 -13.0137 extrinsically 0.857143 -13.0137 corrupted 0.857143 -13.0142 2one 0.857143 -13.0142 postings 0.857143 -13.0142 54.5% 0.857143 -13.0142 and) 0.526316 -13.0144 euromatrixplus 0.666667 -13.0144 ∗this 0.6 -13.0145 pennsylvania 0.857143 -13.0147 lexical-syntactic 0.857143 -13.0147 algorithm; 0.857143 -13.0147 ramshaw, 0.857143 -13.0147 26.4% 0.857143 -13.0147 strictest 0.857143 -13.0147 retweet 0.857143 -13.0147 collision 0.666667 -13.0148 11.9 0.777778 -13.0148 post-process 0.666667 -13.0151 w* 0.666667 -13.0151 sun, 0.857143 -13.0152 texts) 0.857143 -13.0152 standard” 0.857143 -13.0152 rl, 0.857143 -13.0152 components) 0.857143 -13.0152 stop, 0.857143 -13.0152 polarized 0.857143 -13.0152 performances, 0.112108 -13.0155 latter 0.857143 -13.0157 shallow-semantic 0.857143 -13.0157 purposely 0.857143 -13.0157 “xylitol 0.857143 -13.0157 mj, 0.857143 -13.0157 hasan, 0.857143 -13.0157 chunkiness 0.857143 -13.0157 ‘good’ 0.357143 -13.0157 it makes 0.857143 -13.0162 workforce 0.857143 -13.0162 uneven 0.857143 -13.0162 reisinger 0.857143 -13.0162 intercepts 0.857143 -13.0162 no) 0.857143 -13.0162 most) 0.116788 -13.0162 cases, 0.0714932 -13.0163 introduction 0.235294 -13.0165 use. 0.419355 -13.0165 these tasks 0.419355 -13.0165 park, 0.433333 -13.0165 beijing, 0.777778 -13.0166 916 0.857143 -13.0167 wordwise 0.857143 -13.0167 scbr model 0.857143 -13.0167 match” 0.857143 -13.0167 ifrs 0.857143 -13.0167 intra-sentence 0.857143 -13.0167 q2, 0.857143 -13.0167 trnka 0.666667 -13.0167 mcp 0.555556 -13.0171 cow 0.857143 -13.0172 kirchhoff 0.857143 -13.0172 vilar 0.857143 -13.0172 sprawl 0.857143 -13.0172 nets 0.857143 -13.0172 (dev 0.5 -13.0172 split. 0.857143 -13.0177 henceforth) 0.857143 -13.0177 fuly 0.857143 -13.0177 phenotypic 0.857143 -13.0177 ablated 0.857143 -13.0177 1976), 0.180233 -13.018 dictionary. 0.777778 -13.018 by discarding 0.857143 -13.0182 direct correspondence 0.857143 -13.0182 uniformity 0.857143 -13.0187 it) 0.857143 -13.0187 eskenazi 0.857143 -13.0187 missing, 0.857143 -13.0187 gracefully 0.857143 -13.0187 broadly, 0.857143 -13.0192 0.137 0.857143 -13.0192 pφ 0.857143 -13.0192 nakano 0.857143 -13.0192 esential 0.857143 -13.0192 stacks, 0.857143 -13.0192 english-to-german 0.857143 -13.0192 (1992), 0.857143 -13.0192 0.634 0.857143 -13.0192 1-5 0.857143 -13.0192 management: 0.857143 -13.0192 binomials, 0.857143 -13.0192 randomhigh 0.461538 -13.0194  0.857143 -13.0197 mturkmpqa 0.857143 -13.0197 (melville et 0.857143 -13.0197 (pauls 0.857143 -13.0197 (androutsopoulos 0.857143 -13.0197 protein-location 0.857143 -13.0197 h1, 0.857143 -13.0197 1.57 0.857143 -13.0197 animacy 0.555556 -13.0197 nelken 0.666667 -13.0198 w∗ 0.6 -13.0198 “semantic 0.857143 -13.0203 p4 0.857143 -13.0203 lexicals 0.857143 -13.0203 1980; 0.857143 -13.0203 zc07 0.777778 -13.0203 house, 0.40625 -13.0204 these problems 0.4 -13.0207 throughout this 0.857143 -13.0208 mxl 0.857143 -13.0208 sophistication 0.857143 -13.0208 discontinuities 0.857143 -13.0208 proxies 0.857143 -13.0208 full-length 0.857143 -13.0213 0.234 0.857143 -13.0213 kn06 0.857143 -13.0213 slowed 0.857143 -13.0213 overwhelm 0.857143 -13.0213 (corley 0.857143 -13.0213 siddharthan 0.857143 -13.0213 i-computation 0.857143 -13.0218 generators 0.857143 -13.0218 binary: 0.857143 -13.0218 language-based 0.857143 -13.0218 reverb, 0.857143 -13.0218 sparked 0.857143 -13.0218 location) 0.857143 -13.0218 mater 0.857143 -13.0218 comprehend 0.6 -13.0219 workers were 0.526316 -13.0222 null-aligned 0.857143 -13.0223 lu’s 0.857143 -13.0223 surname 0.857143 -13.0223 70.22 0.857143 -13.0223 topics; 0.0998217 -13.0223 4.2 0.378378 -13.0224 translated. 0.185185 -13.0225 argue 0.857143 -13.0228 malay, 0.857143 -13.0228 sproat, 0.857143 -13.0228 correspondent 0.857143 -13.0228 woeparse 0.857143 -13.0228 0.777 0.857143 -13.0228 976 0.857143 -13.0228 lord, 0.857143 -13.0228 ethical 0.857143 -13.0228 phrase-table, 0.857143 -13.0228 individual’s 0.320755 -13.0232 document pairs 0.857143 -13.0233 idiomaticity 0.857143 -13.0233 arg2) 0.310345 -13.0234 indicators of 0.526316 -13.0237 cuny 0.857143 -13.0238 metric’s 0.857143 -13.0238 okay 0.857143 -13.0238 fi, 0.857143 -13.0238 16.1 0.857143 -13.0238 fertilities 0.777778 -13.0239 sub-sequences 0.777778 -13.0239 automobile 0.347826 -13.0241 tooth 0.666667 -13.0242 keyword ranking 0.857143 -13.0243 botom 0.857143 -13.0243 headings 0.857143 -13.0243 albrecht 0.255814 -13.0247 words per 0.857143 -13.0248 curently 0.857143 -13.0248 wiki, 0.857143 -13.0248 twitter users. 0.857143 -13.0248 mitchell’s 0.857143 -13.0248 jr. 0.857143 -13.0248 yearly 0.857143 -13.0248 saccade 0.857143 -13.0248 mismatch, 0.857143 -13.0248 queries) 0.857143 -13.0248 methods; 0.857143 -13.0248 remembrance 0.555556 -13.0249 nullnullnullnullnullnullnullnullnull 0.857143 -13.0253 segmentors 0.857143 -13.0253 (198) 0.25 -13.0257 were then 0.857143 -13.0258 disorder, 0.857143 -13.0258 cowan, 0.857143 -13.0258 denser 0.857143 -13.0258 length: 0.857143 -13.0258 boils 0.777778 -13.0258 velocity 0.5 -13.0258 4.1.1 0.666667 -13.0261 child nodes 0.526316 -13.0262 table 1: example 0.777778 -13.0262 wt+1 0.777778 -13.0262  0.857143 -13.0263 magazine, 0.857143 -13.0263 reorderings, 0.857143 -13.0263 placeholders 0.857143 -13.0263 into-english 0.857143 -13.0268 vertex, 0.857143 -13.0268 eforts 0.857143 -13.0268 2008)) 0.857143 -13.0273 sugested 0.857143 -13.0273 non-alphanumeric 0.857143 -13.0273 noises 0.857143 -13.0273 displayed, 0.857143 -13.0273 mini 0.857143 -13.0273 discordant 0.857143 -13.0273 (gillenwater 0.666667 -13.0273 month, 0.526316 -13.0278 best results, 0.857143 -13.0279 substantially, 0.857143 -13.0279 elpr 0.857143 -13.0279 disappointing 0.857143 -13.0279 bhattacharyya 0.857143 -13.0279 17.5 0.092006 -13.0283 analysis of 0.857143 -13.0284 “content 0.857143 -13.0284 hakkani-tur, 0.857143 -13.0284 4.4.2 0.857143 -13.0284 ecd 0.857143 -13.0284 λ-expressions 0.857143 -13.0284 75.5% 0.204545 -13.0285 independent of 0.777778 -13.0285 10x 0.777778 -13.0285 a b+-tree 0.777778 -13.0285 ken 0.201493 -13.0286 speech. 0.461538 -13.0287 translation performance. 0.6 -13.0287 b-matching 0.857143 -13.0289 concession, 0.857143 -13.0289 slaves 0.857143 -13.0289 toolbox 0.857143 -13.0289 1984) 0.857143 -13.0289 36.7 0.857143 -13.0289 ueffing 0.0917293 -13.029 case, 0.777778 -13.029 points). 0.666667 -13.0293 asymptotic 0.857143 -13.0294 web data, 0.857143 -13.0294 transcribed, 0.857143 -13.0294 a0, 0.857143 -13.0294 schuler, 0.857143 -13.0294 mturk’s 0.857143 -13.0294 [ma10] 0.777778 -13.0295 292 0.195804 -13.0295 applications, 0.857143 -13.0299 rss 0.857143 -13.0299 my computer 0.857143 -13.0299 german-english, 0.857143 -13.0299 sassi 0.857143 -13.0299 (copestake, 0.857143 -13.0299 (al) 0.666667 -13.0301 section 5) 0.857143 -13.0304 dificulties 0.857143 -13.0304 crash 0.857143 -13.0304 359 0.857143 -13.0304 1996, 0.857143 -13.0304 19.6 0.857143 -13.0304 evaluation; 0.857143 -13.0304 setzer 0.857143 -13.0304 consonantal 0.857143 -13.0304 contradicts 0.857143 -13.0304 (manandhar 0.857143 -13.0304 totrtale 0.666667 -13.0305 available resources 0.3 -13.0307 dimensional 0.777778 -13.0308 acl-ijcnlp 0.857143 -13.0309 1.07 0.857143 -13.0309 benchmarks, 0.857143 -13.0309 ball, 0.857143 -13.0309 reduplication 0.857143 -13.0309 harry 0.857143 -13.0309 1400 0.419355 -13.0311 works. 0.777778 -13.0313 (2009) presented 0.857143 -13.0314 skilled 0.857143 -13.0314 parsing strategies 0.857143 -13.0314 98.5 0.857143 -13.0314 tight, 0.857143 -13.0314 515 0.857143 -13.0314 svmtool 0.857143 -13.0314 spatiotemporal 0.388889 -13.0314 date, 0.2 -13.0317 this work is 0.777778 -13.0318 to reestimate 0.777778 -13.0318 20: 0.857143 -13.0319 premise: 0.857143 -13.0319 languagespecific 0.857143 -13.0319 proceedings, 0.857143 -13.0319 ex(n 0.857143 -13.0319 scrutiny 0.857143 -13.0319 infiltrates 0.857143 -13.0319 function; 0.857143 -13.0319 convote 0.777778 -13.0322 654 0.378378 -13.0323 subtilis 0.857143 -13.0324 nist’s 0.857143 -13.0324 klebanov 0.857143 -13.0324 category’s 0.857143 -13.0324 undertaken 0.857143 -13.0324 coherently 0.857143 -13.0324 auto-labeled 0.857143 -13.0324 structure-based 0.7 -13.0326 0.628 0.777778 -13.0327  0.314815 -13.0329 richard 0.857143 -13.033 unfeasible 0.857143 -13.033 quite different. 0.857143 -13.033 s(i, 0.857143 -13.033 nephrectomy 0.857143 -13.033 eventive 0.857143 -13.033 type-logical 0.7 -13.0334 variable y 0.7 -13.0334 hybrid algorithm 0.857143 -13.0335 names) 0.857143 -13.0335 re-weighting 0.857143 -13.0335 0.273 0.857143 -13.0335 incapable 0.857143 -13.0335 tic 0.857143 -13.0335 wmt2011 0.857143 -13.0335 hinton, 0.857143 -13.0335 ai(x) 0.857143 -13.0335 comonsense 0.857143 -13.0335 hi, 0.857143 -13.0335 nonsense 0.857143 -13.0335 rouge-1, 0.857143 -13.0335 linguistics: 0.857143 -13.0335 repeatedly, 0.857143 -13.0335 contextual probability 0.857143 -13.0335 j; 0.388889 -13.0337 semantic category 0.0977702 -13.0338 such a 0.857143 -13.034 0.598 0.857143 -13.034 cases), 0.857143 -13.034 (j 0.857143 -13.034 when opt 0.857143 -13.034 pg(σ) 0.857143 -13.034 speculated 0.857143 -13.034 re-implementation 0.857143 -13.034 perform, 0.197183 -13.034 annotators. 0.5625 -13.0341 enju parser 0.7 -13.0342 predicate labeling 0.857143 -13.0345 configurations: 0.857143 -13.0345 ‘stem’ 0.857143 -13.0345 reservation 0.857143 -13.0345 presupposes 0.857143 -13.0345 0.487 0.555556 -13.0345 ]. 0.7 -13.0346 forexample, 0.217391 -13.0347 research has 0.666667 -13.0349 we defer 0.857143 -13.035 blocks, 0.857143 -13.035 36.27 0.857143 -13.035 \" 0.857143 -13.035 following properties: 0.857143 -13.035 higherorder 0.857143 -13.035 1972) 0.857143 -13.035 egypt 0.857143 -13.035 plant, 0.777778 -13.035 cross-lingual similarity 0.348837 -13.0354 efficient. 0.857143 -13.0355 bigrams) 0.857143 -13.0355 brill, 0.857143 -13.0355 isr-wn 0.857143 -13.0355 paralel 0.419355 -13.0356 testing data 0.7 -13.0359 fs based 0.857143 -13.036 segtagl 0.857143 -13.036 delays 0.857143 -13.036 contradiction, 0.857143 -13.036 re-written 0.857143 -13.036 underlines 0.857143 -13.036 long-simmering 0.857143 -13.036 stability, 0.857143 -13.036 worldwide 0.857143 -13.036 bci 0.857143 -13.036 (28) 0.857143 -13.036 wbest 0.857143 -13.036 0.300 0.857143 -13.036 one-on-one 0.7 -13.0363 paraphrase patterns 0.857143 -13.0365 cursory 0.857143 -13.0365 “can 0.857143 -13.0365 sub-result 0.857143 -13.0365 attraction 0.857143 -13.0365 polarities, 0.857143 -13.0365 1.09 0.666667 -13.0365 spelling errors, 0.7 -13.0367 78.7 0.666667 -13.0369 affine 0.777778 -13.0369 relation distributions 0.857143 -13.0371 43.0 0.857143 -13.0371 (end 0.857143 -13.0371 71.8% 0.857143 -13.0371 :λ 0.857143 -13.0371 realiser 0.857143 -13.0371 (leuski 0.7 -13.0372 semi-supervised methods 0.777778 -13.0373 two sides 0.857143 -13.0376 graph-theoretic 0.857143 -13.0376 57.2 0.857143 -13.0376 single-type 0.857143 -13.0376 {x 0.857143 -13.0376 q4 0.234694 -13.0379 dialogue system 0.857143 -13.0381 experiment 2: 0.857143 -13.0381 hs 0.857143 -13.0381 submision 0.666667 -13.0381 subcellular 0.666667 -13.0381 integrative 0.777778 -13.0383 money, 0.5 -13.0383 sessions. 0.7 -13.0384 linear classifiers 0.666667 -13.0385 there are various 0.857143 -13.0386 p(t→h) 0.857143 -13.0386 f-measure) 0.857143 -13.0386 gene/gene 0.857143 -13.0386 duo 0.857143 -13.0386 condition) 0.857143 -13.0386 zi, 0.857143 -13.0386 voutilainen 0.857143 -13.0386 dependency relation, 0.857143 -13.0386 52.88 0.857143 -13.0386 2008)), 0.857143 -13.0386 weischedel, 0.7 -13.0388 an important part 0.555556 -13.0388 squaresolid squaresolid 0.428571 -13.0388 fk 0.857143 -13.0391 dit 0.857143 -13.0391 14) 0.857143 -13.0391 tillmann 0.857143 -13.0391 set” 0.857143 -13.0391 repairs 0.857143 -13.0391 greedy, 0.857143 -13.0391 aphasia, 0.857143 -13.0391 attainable 0.857143 -13.0391 uem 0.857143 -13.0396 58.4 0.857143 -13.0396 humor 0.857143 -13.0396 produce, 0.857143 -13.0396 p′ 0.857143 -13.0396 mapping probabilities 0.857143 -13.0396 object; 0.857143 -13.0396 experimentation, 0.857143 -13.0396 1.29 0.857143 -13.0396 instances), 0.428571 -13.0396 an email 0.40625 -13.0397 callison-burch et 0.176136 -13.0397 input. 0.208 -13.0397 our experimental 0.857143 -13.0401 collapses 0.857143 -13.0401 synchronize 0.857143 -13.0401 other: 0.857143 -13.0401 locates 0.857143 -13.0401 responsive 0.857143 -13.0401 к 0.857143 -13.0401 (duan 0.666667 -13.0405 coindexation 0.154185 -13.0405 assigns 0.857143 -13.0407 yield, 0.857143 -13.0407 j-e 0.857143 -13.0407 conll, 0.237113 -13.0408 p: 0.857143 -13.0412 mainstream 0.857143 -13.0412 dmrs 0.857143 -13.0412 (ponzetto 0.857143 -13.0412 “car” 0.857143 -13.0412 (directed 0.6 -13.0412 finally, we conclude 0.348837 -13.0412 directions, 0.666667 -13.0413 an investigation 0.10917 -13.0414 sets. 0.777778 -13.0416 “new 0.5 -13.0416 0.01. 0.857143 -13.0417 quartile 0.857143 -13.0417 anymore 0.857143 -13.0417 mantic 0.857143 -13.0417 pronouns; 0.857143 -13.0417 innovations, 0.857143 -13.0417 examiners 0.857143 -13.0417 terminal, 0.857143 -13.0417 queries: 0.857143 -13.0417 460 0.857143 -13.0417 markables, 0.857143 -13.0417 mert’s 0.857143 -13.0417 impersonal 0.378378 -13.0418 spitkovsky et 0.461538 -13.0419 tag set 0.857143 -13.0422 kawahara 0.857143 -13.0422 0.6, 0.857143 -13.0422 concreteness, 0.857143 -13.0422 797 0.7 -13.0426 “do 0.084088 -13.0426 this work 0.857143 -13.0427 million, 0.857143 -13.0427 corelex 0.857143 -13.0427 difference: 0.857143 -13.0427 ggp 0.857143 -13.0427 (f), 0.857143 -13.0427 feature’s 0.857143 -13.0427 yprime, 0.857143 -13.0427 0.418 0.857143 -13.0427 r(e) 0.6 -13.0427 layers. 0.777778 -13.043 which encodes 0.857143 -13.0432 (3.1) 0.857143 -13.0432 givenness 0.857143 -13.0432 st’11 0.2625 -13.0434 no significant 0.857143 -13.0437 arranging 0.857143 -13.0437 2.12 0.857143 -13.0437 hpsg, 0.857143 -13.0437 bless, 0.857143 -13.0437 (192) 0.219298 -13.0437 were able 0.7 -13.0439 format: 0.314815 -13.044 0.05) 0.857143 -13.0443 (lord 0.857143 -13.0443 lance 0.857143 -13.0443 view: 0.857143 -13.0443 (cheng 0.857143 -13.0443 advised 0.857143 -13.0443 m) 0.6 -13.0445 ) ; 0.857143 -13.0448 algebra, 0.857143 -13.0448 non-coerced 0.857143 -13.0448 collapsed, 0.857143 -13.0448 (msa) 0.857143 -13.0448 text-to-semantics 0.857143 -13.0453 discriminates 0.857143 -13.0453 correct; 0.857143 -13.0453 correspondences, 0.857143 -13.0453 interlingua 0.857143 -13.0453 26.20 0.857143 -13.0453 question-queries 0.857143 -13.0453 hawaiian 0.101695 -13.0456 meeting 0.857143 -13.0458 accumulation 0.857143 -13.0458 being processed 0.857143 -13.0458 sec, 0.857143 -13.0458 transparently 0.857143 -13.0458 333 0.857143 -13.0458 pag 0.857143 -13.0458 slopes 0.857143 -13.0458 0.0051 0.857143 -13.0458 hyperlinked 0.777778 -13.0458 (v,e) 0.7 -13.046 32, 0.5 -13.0461 null3 0.777778 -13.0462 jacob 0.857143 -13.0463 inflect 0.857143 -13.0463 reachability 0.857143 -13.0463 finish, 0.7 -13.0464 these issues. 0.7 -13.0464 84.6% 0.777778 -13.0467 id: 0.108932 -13.0467 corpora. 0.857143 -13.0469 marconi 0.857143 -13.0469 86.7% 0.857143 -13.0469 timing, 0.857143 -13.0469 kilgarriff, 0.857143 -13.0469 directives 0.857143 -13.0469 0.983 0.857143 -13.0469 0.665 0.326531 -13.0469 are always 0.295082 -13.047 al., 2001). 0.777778 -13.0472 omits 0.857143 -13.0474 setzer, 0.857143 -13.0474 pair; 0.857143 -13.0474 parametrize 0.857143 -13.0474 accurately, 0.857143 -13.0474 structures; 0.857143 -13.0474 426 0.857143 -13.0474 phrase; 0.857143 -13.0474 embeds 0.857143 -13.0474 (normalized) 0.615385 -13.0475 disagreed 0.393939 -13.0476 imbalance 0.666667 -13.0477 tenses, 0.7 -13.0477 (is 0.857143 -13.0479 591 0.857143 -13.0479 tukey 0.857143 -13.0479 milne 0.857143 -13.0479 subtitle 0.857143 -13.0479 bioscience 0.857143 -13.0479 constituency-based 0.295082 -13.0479 format. 0.5625 -13.048 codeword 0.857143 -13.0484 customers, 0.857143 -13.0484 sion 0.857143 -13.0484 (mturk) 0.857143 -13.0484 unfair 0.305085 -13.0484 their approach 0.615385 -13.0486 lem 0.857143 -13.0489 labels), 0.857143 -13.0489 27.34 0.857143 -13.0489 “true” 0.478261 -13.0489 theme, 0.7 -13.049 .51 0.857143 -13.0494 sentence-initial 0.857143 -13.0494 diving 0.857143 -13.0494 s5 0.857143 -13.0494 weaklysupervised 0.478261 -13.0495 candidate selection 0.857143 -13.05 (abney 0.857143 -13.05 aan 0.857143 -13.05 translation b: 0.857143 -13.05 reviews: 0.615385 -13.0501 swts 0.7 -13.0502 bannard, 0.857143 -13.0505 article: 0.857143 -13.0505 cue, 0.857143 -13.0505 frontier node 0.857143 -13.0505 anomaly 0.7 -13.0507 95.7 0.357143 -13.0508 model). 0.666667 -13.0509 (red 0.066129 -13.0509 likely 0.857143 -13.051 (singular 0.857143 -13.051 tops 0.857143 -13.051 strong indicator 0.857143 -13.051 .43 0.857143 -13.051 occ(t) 0.857143 -13.051 (kb) 0.857143 -13.051 extended, 0.857143 -13.051 stevens, 0.857143 -13.051 trees), 0.5 -13.051 elderly 0.777778 -13.0514 single pass 0.777778 -13.0514 despite using 0.857143 -13.0515 2it 0.357143 -13.0515 see figure 0.314815 -13.0516 is repeated 0.615385 -13.0516 five. 0.348837 -13.0518 outcome of 0.777778 -13.0519 =< 0.857143 -13.0521 resist 0.857143 -13.0521 deliberate 0.857143 -13.0521 “hear” 0.857143 -13.0521 1974; 0.857143 -13.0521 false, 0.857143 -13.0521 initiates 0.857143 -13.0521 simile 0.857143 -13.0521 wishing 0.526316 -13.0524 our framework. 0.7 -13.0524 .42 0.857143 -13.0526 leg 0.857143 -13.0526 9), 0.857143 -13.0526 (lex) 0.857143 -13.0526 sys0 0.777778 -13.0528 ∂l2 0.27027 -13.0528 this makes 0.857143 -13.0531 indicator variables 0.615385 -13.0531 bioinformatics 0.7 -13.0532 ccp 0.388889 -13.0534 a user’s 0.857143 -13.0536 holmes 0.857143 -13.0536 chapman 0.857143 -13.0536 dmv, 0.857143 -13.0536 admits 0.7 -13.0536 entity-mention 0.6 -13.0536 query q 0.183544 -13.0539 reflects 0.857143 -13.0541 (1987), 0.857143 -13.0541 m-step: 0.857143 -13.0541 jitter 0.857143 -13.0541 [7] 0.857143 -13.0541 accept/reject 0.7 -13.0541 desrt 0.357143 -13.0541 lexical information 0.134228 -13.0542 = 1 0.666667 -13.0546 structured data 0.615385 -13.0546 zechner 0.857143 -13.0547 2m 0.857143 -13.0547 “from” 0.857143 -13.0547 infrastructure, 0.857143 -13.0547 0.257 0.7 -13.0549 may involve 0.2 -13.055 middle 0.857143 -13.0552 portrayed 0.857143 -13.0552 instances: 0.857143 -13.0552 accuracies, 0.857143 -13.0552 (al-onaizan 0.857143 -13.0552 (yeh, 0.7 -13.0553 351 0.857143 -13.0557 behaviors, 0.857143 -13.0557 (run 0.857143 -13.0557 deviation, 0.857143 -13.0557 through, 0.857143 -13.0557 0.706 0.857143 -13.0557 0.813 0.857143 -13.0557 2department 0.857143 -13.0557 lao 0.0150404 -13.0557 by 0.0682594 -13.0558 we will 0.5 -13.0558 richness of 0.478261 -13.0559 o: 0.857143 -13.0562 2-grams, 0.857143 -13.0562 lse, 0.857143 -13.0562 θj 0.857143 -13.0562 ˆl = 0.857143 -13.0562 unimodal model 0.857143 -13.0562 919 0.857143 -13.0562 preview 0.526316 -13.0563 between them, 0.241758 -13.0565 scenario, 0.5625 -13.0566 quite well 0.666667 -13.0566 certain type 0.7 -13.0566 (mext, 0.857143 -13.0568 myriad 0.857143 -13.0568 neds 0.857143 -13.0568 µi 0.857143 -13.0568 dialogue-style speech 0.857143 -13.0568 interfere 0.857143 -13.0568 distance) 0.857143 -13.0568 522 0.6 -13.0569 µ(ν) 0.7 -13.0571 typographical 0.5625 -13.0572 serena 0.2 -13.0572 non-terminal 0.857143 -13.0573 393 0.857143 -13.0573 set-theoretic 0.857143 -13.0573 n} 0.857143 -13.0573 1.33 0.5 -13.0574 that, given 0.857143 -13.0578 zb 0.857143 -13.0578 (c&c 0.857143 -13.0578 machine-learned 0.203125 -13.0579 would not 0.857143 -13.0583 safety, 0.857143 -13.0583 englishgerman 0.857143 -13.0583 470 0.857143 -13.0583 newsgroup 0.857143 -13.0583 semantics: 0.857143 -13.0583 ([v1,v2]) 0.857143 -13.0583 hydrogen 0.857143 -13.0583 negatives, 0.6 -13.0584 giza++. 0.857143 -13.0588 sl-full 0.857143 -13.0588 (coltheart, 0.857143 -13.0588 facebook, 0.857143 -13.0588 s-neg 0.857143 -13.0588 chartbased 0.857143 -13.0588 y’ 0.7 -13.0588 βj 0.777778 -13.059 series. 0.777778 -13.059 bosch 0.777778 -13.059 0.1). 0.777778 -13.059 distractors. 0.777778 -13.059 (lo 0.777778 -13.059 – sad,” 0.777778 -13.059 acronym. 0.777778 -13.059 arg1. 0.111111 -13.0591 ability 0.857143 -13.0594 ppc 0.857143 -13.0594 inter-anotator 0.857143 -13.0594 numeracy 0.857143 -13.0594 accidental 0.857143 -13.0594 3.3.3 0.857143 -13.0594 imediately 0.857143 -13.0594 table5: 0.857143 -13.0594 dependency-to-string 0.857143 -13.0594 dangerous 0.7 -13.0596 94.5% 0.5625 -13.0597 insensitive 0.857143 -13.0599 grapheme 0.857143 -13.0599 (k¨ubler 0.857143 -13.0599 relaxations 0.458333 -13.0601 ces 0.857143 -13.0604 128, 0.857143 -13.0604 anaphoricity features 0.857143 -13.0604 jenny 0.857143 -13.0604 wood 0.857143 -13.0604 (depending 0.857143 -13.0604 plotting 0.857143 -13.0604 table) 0.857143 -13.0604 constructs, 0.857143 -13.0604 toolbar 0.666667 -13.0607 already present 0.857143 -13.0609 -coref 0.857143 -13.0609 label bias 0.857143 -13.0609 (manine 0.857143 -13.0609 irony, 0.857143 -13.0609 robotics 0.615385 -13.061 texter 0.526316 -13.0614 a trigram language 0.253012 -13.0614 itself, 0.857143 -13.0615 irreversible 0.857143 -13.0615 everywhere 0.857143 -13.0615 china) 0.857143 -13.0615 ‘new 0.857143 -13.062 economics, 0.857143 -13.062 subwords 0.857143 -13.062 straight-forward 0.857143 -13.062 rogers 0.857143 -13.062 tac_kbp 0.857143 -13.062 phrases; 0.5625 -13.0621 views, 0.615385 -13.0621 linguistica 0.615385 -13.0621 developments in 0.7 -13.0622 an ie 0.857143 -13.0625 binding trigger 0.857143 -13.0625 glossing 0.857143 -13.0625 semantics; 0.857143 -13.0625 93.04 0.857143 -13.0625 lying 0.857143 -13.0625 hua 0.857143 -13.0625 checking, 0.857143 -13.0625 query-level 0.857143 -13.0625 mentions: 0.857143 -13.0625 injection 0.458333 -13.0629 stretchy 0.5 -13.0629 touch 0.857143 -13.0631 concerns, 0.857143 -13.0631 meronymic 0.857143 -13.0631 recscore 0.857143 -13.0631 wpartial 0.857143 -13.0631 d-n 0.7 -13.0631 20.00 0.478261 -13.0632 (2010) propose 0.615385 -13.0632 araucaria 0.857143 -13.0636 82.9 0.857143 -13.0636 b(k,m,lscript) 0.857143 -13.0636 lens 0.857143 -13.0636 spanish-to-english 0.7 -13.0639 -1, 0.857143 -13.0641 grass, 0.857143 -13.0641 longitudinal 0.857143 -13.0641 (pollard and 0.857143 -13.0641 query; 0.857143 -13.0641 〈e,t〉 0.857143 -13.0641 s4: 0.857143 -13.0641 0.266 0.23 -13.0642 procedure. 0.6 -13.0642 trial, 0.857143 -13.0646 413 0.857143 -13.0646 seated 0.857143 -13.0646 0.781 0.857143 -13.0646 manageable, 0.857143 -13.0646 narrative, 0.857143 -13.0646 preferably 0.857143 -13.0646 (17 0.857143 -13.0646 ork 0.175439 -13.0647 funded by 0.208 -13.065 are computed 0.857143 -13.0652 interdependent 0.857143 -13.0652 dept 0.857143 -13.0652 sentiwordnet: 0.857143 -13.0652 cheng, 0.857143 -13.0652 tolerable 0.857143 -13.0652 (pos), 0.5625 -13.0655 decision-related 0.5 -13.0656 untrained 0.857143 -13.0657 ˆp 0.857143 -13.0657 jobs, 0.857143 -13.0657 jjr 0.857143 -13.0657 implies, 0.857143 -13.0657 spoken, 0.857143 -13.0657 cascading 0.6 -13.0657 training iterations 0.190476 -13.0658 content. 0.215517 -13.066 represented in 0.6 -13.0661 are ignored 0.857143 -13.0662 scale: 0.857143 -13.0662 localisation 0.857143 -13.0662 benefits, 0.857143 -13.0662 instantiated arguments 0.857143 -13.0662 copredicative 0.857143 -13.0662 subs 0.615385 -13.0663 fillers, 0.7 -13.0665 cancer, 0.6 -13.0665 lct 0.857143 -13.0667 surviving 0.857143 -13.0667 sequenced 0.857143 -13.0667 150,000 0.857143 -13.0667 aida 0.857143 -13.0667 non-collaborative 0.857143 -13.0667 1.4) 0.164103 -13.0668 numbers of 0.215517 -13.0668 steps. 0.526316 -13.0669 noticeably 0.7 -13.0669 it satisfies 0.140741 -13.067 we now 0.857143 -13.0673 1.37 0.857143 -13.0673 v-measure, 0.857143 -13.0673 #2: 0.857143 -13.0673 satisfiability 0.857143 -13.0673 possesses 0.857143 -13.0673 deserving 0.857143 -13.0673 inevitable 0.5 -13.0675 rambow, 0.857143 -13.0678 baroni and lenci 0.857143 -13.0678 subdialogues 0.857143 -13.0678 expert scorers 0.857143 -13.0678 gk 0.857143 -13.0678 suzanne 0.857143 -13.0678 dep2str 0.137809 -13.0678 improvement in 0.7 -13.0682 wordnet::similarity 0.857143 -13.0683 nonlexical 0.857143 -13.0683 stupid 0.857143 -13.0683 n-th 0.857143 -13.0683 head-argument 0.857143 -13.0683 364 0.857143 -13.0689 v-o 0.857143 -13.0689 regeneration 0.857143 -13.0689 categories; 0.857143 -13.0689 fd 0.857143 -13.0689 wikipedia: 0.7 -13.0691 austin, 0.857143 -13.0694 char 0.7 -13.0695 price, 0.109865 -13.0696 examples of 0.615385 -13.0697 16th 0.857143 -13.0699 results) 0.857143 -13.0699 582 0.857143 -13.0699 somewhat, 0.857143 -13.0699 webkb 0.857143 -13.0699 naturalistic 0.458333 -13.07 proposed methods 0.478261 -13.07 opinion words. 0.7 -13.0704 92.9 0.7 -13.0708 end-to-end system 0.857143 -13.071 component systems, 0.857143 -13.071 title) 0.857143 -13.071 crowdflower’s 0.857143 -13.071 factual information 0.857143 -13.071 desired output 0.5625 -13.0711 universit¨at 0.428571 -13.0712 and cardie 0.12766 -13.0713 analysis. 0.857143 -13.0715 sid 0.857143 -13.0715 testable 0.857143 -13.0715 ol 0.857143 -13.0715 recognizable 0.857143 -13.0715 hanks 0.615385 -13.0716 ghz 0.478261 -13.0717 syntactic context 0.458333 -13.072 word association 0.615385 -13.072 cap-macro 0.857143 -13.0721 interpolation: 0.857143 -13.0721 0.9% 0.857143 -13.0721 inactive 0.857143 -13.0721 office, 0.857143 -13.0721 artistic 0.204724 -13.0723 turk 0.857143 -13.0726 taged 0.857143 -13.0726 crowd-sourcing 0.857143 -13.0726 seeded with 0.857143 -13.0726 agreement) 0.857143 -13.0726 14we 0.857143 -13.0726 tween 0.857143 -13.0726 cricket 0.666667 -13.073 lll 0.857143 -13.0731 davis, 0.857143 -13.0731 gains, 0.7 -13.0734 79.1 0.7 -13.0734 308 0.666667 -13.0735 a nice 0.857143 -13.0736 n1, 0.857143 -13.0736 (recall) 0.857143 -13.0736 uas, 0.857143 -13.0736 countermeasures 0.857143 -13.0736 latin, 0.857143 -13.0736 (bishop, 0.857143 -13.0736 dropped, 0.5 -13.0737 (this is 0.615385 -13.0739 mixture models 0.857143 -13.0742 jeff 0.7 -13.0743 discretize 0.526316 -13.0744 sentiment detection 0.333333 -13.0744 for tuning 0.5625 -13.0746 query-comp 0.119241 -13.0746 occurrences 0.857143 -13.0747 m1, 0.857143 -13.0747 right-hand-side 0.857143 -13.0747 h-ter 0.615385 -13.0747 (2010) report 0.7 -13.0748 parameters α 0.413793 -13.0749 also performed 0.615385 -13.075 subordinating 0.666667 -13.0751 bigram model, 0.666667 -13.0751 325 0.309091 -13.0751 just as 0.857143 -13.0752 non-content 0.857143 -13.0752 nominative, 0.857143 -13.0752 rosenfeld, 0.857143 -13.0752 0.310 0.857143 -13.0752 magazines, 0.7 -13.0752 up” 0.7 -13.0752 msa, 0.615385 -13.0754 “go 0.444444 -13.0755 collected. 0.5625 -13.0756 )] 0.7 -13.0756 43.6 0.7 -13.0756 -0.02 0.348837 -13.0756 it includes 0.857143 -13.0758 soundex 0.857143 -13.0758 798 0.857143 -13.0758 heidelberg 0.458333 -13.0759 sentences contain 0.526316 -13.0761 contains many 0.368421 -13.0761 nps. 0.857143 -13.0763 text+ 0.857143 -13.0763 corpus-wide 0.857143 -13.0763 apples 0.857143 -13.0763 expressiveness 0.857143 -13.0763 1 motivation 0.5625 -13.0763 .25 0.666667 -13.0764 adequacy. 0.7 -13.0765 functors 0.478261 -13.0765 empty. 0.857143 -13.0768 hastie, 0.857143 -13.0768 606 0.857143 -13.0768 roth, 2010) 0.857143 -13.0768 pp attachment 0.666667 -13.0768 hiden 0.368421 -13.0768 for developing 0.857143 -13.0774 vf 0.857143 -13.0774 560 0.857143 -13.0774 weeks, 0.285714 -13.0775 as mentioned in 0.295082 -13.0775 vbd 0.857143 -13.0779 pakistan 0.857143 -13.0779 subtasks: 0.857143 -13.0779 411 0.857143 -13.0779 study: 0.615385 -13.0781 59.8 0.382353 -13.0781 expensive. 0.7 -13.0782 (sproat 0.368421 -13.0782 each component 0.263158 -13.0783 alignment between 0.25641 -13.0784 power of 0.128931 -13.0784 5.2 0.857143 -13.0785 appear, 0.857143 -13.0785 372 0.857143 -13.0785 polynomial-time 0.444444 -13.0785 holder 0.444444 -13.0785 which combines 0.7 -13.0787 this result, 0.7 -13.0787 stop words. 0.7 -13.0787 hinrich 0.142308 -13.0789 dirichlet 0.857143 -13.079 subtraction 0.857143 -13.079 stuttz 0.857143 -13.079 tie-breaking 0.857143 -13.079 causes, 0.857143 -13.079 462 0.857143 -13.0795 default moses 0.857143 -13.0795 pilot task 0.615385 -13.0796 (tables 0.615385 -13.0796 what kinds of 0.666667 -13.0797 iters 0.666667 -13.0797 translation”. 0.378378 -13.0798 reduces to 0.857143 -13.0801 (sl) 0.857143 -13.0801 mines 0.857143 -13.0801 catalysis events 0.857143 -13.0801 (perfect 0.857143 -13.0806 337 0.857143 -13.0806 1100 0.857143 -13.0806 mf 0.857143 -13.0806 fame 0.227723 -13.0808 indication 0.857143 -13.0811 ze 0.857143 -13.0811 16.0 0.857143 -13.0811 post-nominal 0.857143 -13.0811 base: 0.857143 -13.0811 reparandum, 0.615385 -13.0812 matt 0.7 -13.0813 pradhan 0.857143 -13.0817 infused 0.857143 -13.0817 corefer 0.857143 -13.0817 http 0.7 -13.0817 (druck 0.382353 -13.0817 experiments described 0.148936 -13.0817 journal 0.615385 -13.082 ohta 0.857143 -13.0822 laws, 0.857143 -13.0822 zero auxiliary 0.857143 -13.0822 adj, 0.857143 -13.0822 “social 0.7 -13.0822 instantiates 0.615385 -13.0823 modeling framework 0.857143 -13.0827 speeches 0.615385 -13.0827 approx 0.478261 -13.0828 three. 0.857143 -13.0833 a80a80a80 0.857143 -13.0833 author-recipient 0.857143 -13.0833 u(s) 0.857143 -13.0833 r1,2 0.857143 -13.0833 summand 0.857143 -13.0833 current token. 0.857143 -13.0833 shuffling 0.32 -13.0834 we explored 0.857143 -13.0838 732 0.857143 -13.0838 pausal 0.857143 -13.0838 fly, 0.615385 -13.0839 15.1 0.615385 -13.0843 nflteam 0.857143 -13.0844 kept, 0.857143 -13.0844 instantaneous 0.857143 -13.0844 770 0.857143 -13.0844 owner 0.857143 -13.0844 word), 0.857143 -13.0844 above), 0.857143 -13.0844 0.260 0.7 -13.0844 next describe 0.7 -13.0844 trip 0.7 -13.0848 24.4 0.857143 -13.0849 itself: 0.857143 -13.0849 spacing 0.857143 -13.0849 relation-based 0.857143 -13.0849 modality, 0.32 -13.0852 bilingual lexicon 0.7 -13.0853 (np), 0.857143 -13.0854 excessively 0.857143 -13.0854 lprime 0.857143 -13.0854 assurance for 0.857143 -13.0854 53, 0.857143 -13.0854 t4 0.857143 -13.0854 reflexive, 0.857143 -13.086 mismatches between 0.857143 -13.086 tractability, 0.857143 -13.086 wished 0.857143 -13.086 4.2), 0.857143 -13.086 liwc+_f 0.857143 -13.086 yx 0.40625 -13.0861 ralph 0.0480559 -13.0861 does 0.857143 -13.0865 totals 0.133562 -13.0865 provides a 0.7 -13.0866 downtown 0.413793 -13.0868 learning problem 0.857143 -13.087 modal verbs 0.857143 -13.087 agreement), 0.857143 -13.087 paper examines 0.857143 -13.087 regulated 0.526316 -13.087 asbestos 0.7 -13.087 carried by 0.615385 -13.087 these reasons, 0.5 -13.0874 fear, 0.857143 -13.0876 de–en 0.857143 -13.0876 45.43 0.857143 -13.0876 23.77 0.857143 -13.0876 qr 0.857143 -13.0876 must, 0.857143 -13.0876 postulating 0.857143 -13.0876 zq 0.7 -13.0879 ⇐⇒ 0.857143 -13.0881 0.320 0.857143 -13.0881 5see 0.0431095 -13.0881 shows 0.5 -13.0881 particular type 0.348837 -13.0883 epoch 0.029053 -13.0884 if 0.313725 -13.0885 translation task. 0.857143 -13.0887 stake 0.857143 -13.0887 bod 0.857143 -13.0887 (cucerzan 0.5 -13.089 demographics 0.102616 -13.0891 advantage 0.857143 -13.0892 politics, 0.857143 -13.0892 documentlevel 0.857143 -13.0892 martins, 0.857143 -13.0892 chloride 0.0978261 -13.0892 workshop on 0.187919 -13.0892 our algorithm 0.7 -13.0897 0.700 0.615385 -13.0897 average f-score 0.857143 -13.0898 (use 0.857143 -13.0898 disallowing 0.857143 -13.0898 set’s 0.857143 -13.0898 term-vectors 0.393939 -13.09 computational model 0.857143 -13.0903 corespond 0.144 -13.0905 output. 0.857143 -13.0908 onlyseed 0.857143 -13.0908 0.055 0.428571 -13.0908 colour. 0.208333 -13.0909 apart 0.393939 -13.091 gao 0.298246 -13.091 daum´e 0.857143 -13.0914 0.914 0.478261 -13.0915 we adapt 0.615385 -13.0916 11th 0.857143 -13.0919 configurations 0.857143 -13.0919 look, 0.857143 -13.0919 timestamps 0.5625 -13.0919 precondition 0.857143 -13.0925 destination, 0.857143 -13.0925 hosted 0.857143 -13.0925 uniform, 0.857143 -13.0925 subgroups 0.857143 -13.0925 cruz 0.5 -13.0925 yo 0.114504 -13.0926 a specific 0.7 -13.0928 macr 0.7 -13.0928 42.8 0.615385 -13.0928 slower, 0.526316 -13.0929 (resnik 0.857143 -13.093 sugest 0.857143 -13.093 q(y) 0.7 -13.0932 eigenvector 0.7 -13.0932 women, 0.0810127 -13.0933 highest 0.857143 -13.0935 forbes-riley, 0.857143 -13.0935 ↑ 0.857143 -13.0935 poem 0.857143 -13.0935 prospects 0.857143 -13.0935 pace 0.7 -13.0936 bombs 0.5625 -13.094 teams, 0.857143 -13.0941 wikulu’s 0.857143 -13.0941 1.0) 0.857143 -13.0941 26, 0.857143 -13.0941 (while 0.7 -13.0941 event mention, 0.234043 -13.0942 are marked 0.857143 -13.0946 negex 0.615385 -13.0948 lemmatization, 0.428571 -13.0949 reliable. 0.368421 -13.0949 also shown 0.529412 -13.0951 mutual information, 0.857143 -13.0952 fixedness 0.857143 -13.0952 governor, 0.7 -13.0954 concept mapping 0.458333 -13.0955 scientists 0.857143 -13.0957 lr-kb4 0.857143 -13.0957 desired, 0.857143 -13.0957 (cf) 0.857143 -13.0957 reconcile’s 0.428571 -13.0957 mitigate 0.75 -13.0959 coresponding to 0.231579 -13.096 subject to 0.276923 -13.0962 results indicate 0.857143 -13.0963 including, 0.857143 -13.0963 kdd-t 0.857143 -13.0963 english-czech 0.857143 -13.0963 “bin-2” 0.7 -13.0963 3k 0.75 -13.0964 880 0.393939 -13.0965 information sciences 0.5625 -13.0965 implement. 0.857143 -13.0968 manipulations 0.857143 -13.0968 chromosome 0.857143 -13.0968 col 0.75 -13.0968 out) 0.529412 -13.0968 2009a), 0.7 -13.0972 method) 0.168478 -13.0972 asked to 0.75 -13.0973 enclosed 0.333333 -13.0973 koo et 0.857143 -13.0974 filestore 0.857143 -13.0974 satoshi 0.857143 -13.0974 testwsj 0.857143 -13.0974 r2, 0.857143 -13.0974 decorated 0.25641 -13.0975 linguistics. 0.458333 -13.0976 lessons 0.7 -13.0976 evening 0.7 -13.0976 annotators agree 0.131757 -13.0976 overview 0.75 -13.0978 spain, 0.75 -13.0978 probabilistic, 0.857143 -13.0979 changed, 0.857143 -13.0979 wmt2010 0.857143 -13.0979 other; 0.857143 -13.0979 qualitatively, 0.857143 -13.0979 2.4.2 0.857143 -13.0979 specificaly, 0.47619 -13.0979 hvs 0.214286 -13.0981 are reported 0.5 -13.0982 completeness, 0.75 -13.0983 saturation 0.75 -13.0983 stephanie 0.75 -13.0983 tri-nmf 0.75 -13.0983 98052 0.75 -13.0983 jjf 0.75 -13.0983 maptask 0.857143 -13.0984 mouth 0.857143 -13.0984 abbreviation, 0.857143 -13.0984 spanish/english 0.857143 -13.0984 801 0.857143 -13.0984 common parent 0.857143 -13.0984 1274 0.458333 -13.0988 1991), 0.857143 -13.099 ast, 0.857143 -13.099 (dark 0.857143 -13.099 sing 0.857143 -13.099 duc-04 0.857143 -13.099 gravity 0.75 -13.0992 prohibited 0.75 -13.0992 divide-and-conquer 0.75 -13.0992 2.0, 0.115979 -13.0992 some of the 0.7 -13.0994 gaizauskas, 0.857143 -13.0995 bernstein, 0.857143 -13.0995 thick 0.857143 -13.0995 delta, 0.857143 -13.0995 ssummationdisplay 0.857143 -13.0995 container, 0.857143 -13.0995 1.5m 0.857143 -13.0995 αx 0.857143 -13.0995 48.88 0.44 -13.0995 clarity, 0.529412 -13.0995 were excluded 0.75 -13.0997 axioms: 0.75 -13.0997 [lda] 0.0712891 -13.1 framework 0.857143 -13.1001 adjective) 0.5625 -13.1001 pu 0.7 -13.1003 (ap) 0.7 -13.1003 giant 0.0504407 -13.1004 should 0.857143 -13.1006 computers, 0.857143 -13.1006 αc 0.857143 -13.1006 tabulated 0.857143 -13.1006 find, 0.857143 -13.1006 witten 0.857143 -13.1006 12.7 0.857143 -13.1006 (ed), 0.75 -13.1007 lall 0.75 -13.1007 1.67 0.7 -13.1008 depth values 0.75 -13.1011 overlaps, 0.857143 -13.1012 blatz 0.857143 -13.1012 sparseptron 0.857143 -13.1012 brendan 0.7 -13.1012 bank, 0.7 -13.1012 543 0.529412 -13.1015 into english, 0.263889 -13.1015 cs 0.75 -13.1016 primitives, 0.75 -13.1016 chat, 0.75 -13.1016 has gained 0.75 -13.1016 conflict with 0.857143 -13.1017 28, 0.857143 -13.1017 04 0.857143 -13.1017 61.54 0.857143 -13.1017 541 0.7 -13.1017 varieties, 0.7 -13.1021 bot 0.7 -13.1021 2available 0.7 -13.1021 1983). 0.75 -13.1021 additional condition 0.857143 -13.1023 3par 0.857143 -13.1023 copiale 0.241379 -13.1025 whether or not 0.75 -13.1026 happiness 0.75 -13.1026 (binary 0.75 -13.1026 punctuations, 0.0692521 -13.1027 propose 0.857143 -13.1028 married 0.857143 -13.1028 installation 0.857143 -13.1028 φ(vi,v 0.857143 -13.1028 sdrt 0.0570533 -13.1029 1: 0.0712195 -13.1029 allows 0.7 -13.103 equivalence between 0.75 -13.1031 91.9 0.75 -13.1031 (e,f) 0.75 -13.1031 generalizers 0.75 -13.1031 anotator 0.75 -13.1035 deu 0.75 -13.1035 numerator 0.75 -13.1035 beings 0.857143 -13.1039 retired 0.857143 -13.1039 tiers 0.7 -13.1039 smoking 0.75 -13.104 vote. 0.75 -13.104 ”‘machine 0.75 -13.104 innate 0.75 -13.104 update: 0.25974 -13.1042 tool for 0.188811 -13.1044 to be able 0.857143 -13.1045 huffman codes 0.857143 -13.1045 go term 0.75 -13.1045 phrase’s 0.75 -13.1045 greeting 0.75 -13.1045 gleaned 0.75 -13.1045 synthesized speech 0.857143 -13.105 tion, 0.857143 -13.105 0.862 0.857143 -13.105 ungrammatical, 0.615385 -13.105 word positions 0.75 -13.1055 restriction, 0.75 -13.1055 pupa 0.75 -13.1055 includes: 0.75 -13.1055 multext-east 0.75 -13.1055 tired 0.529412 -13.1055 a notable 0.857143 -13.1056 [6] 0.857143 -13.1056 planes 0.4 -13.1057 mert. 0.413793 -13.106 be: 0.857143 -13.1061 characterizations 0.75 -13.1064 loan 0.75 -13.1064 opened 0.75 -13.1064 africa 0.7 -13.1066 1.23 0.7 -13.1066 ee 0.857143 -13.1067 premises, 0.857143 -13.1067 b2, 0.857143 -13.1067 surfaces 0.857143 -13.1067 (ef) 0.857143 -13.1067 (4)) 0.5625 -13.1069 among them 0.75 -13.1069 bands 0.857143 -13.1072 indicate, 0.857143 -13.1072 c’s 0.857143 -13.1072 spirit, 0.857143 -13.1072 dic 0.857143 -13.1072 ipod 0.857143 -13.1072 confusion-related facial 0.857143 -13.1072 sign, 0.857143 -13.1072 “make 0.188811 -13.1074 difference is 0.7 -13.1075 out–of–domain 0.857143 -13.1078 hyper 0.7 -13.1079 responds with 0.75 -13.1079 bs 0.75 -13.1079 (i), 0.857143 -13.1083 genes, 0.246914 -13.1083 each type 0.7 -13.1084 combined score 0.75 -13.1084 illegal 0.75 -13.1084 1–4 0.313725 -13.1085 sprime 0.7 -13.1088 not considered. 0.75 -13.1088 (k) 0.75 -13.1088 coarser-grained 0.75 -13.1088 fashion: 0.857143 -13.1089 “the”, 0.857143 -13.1089 cosine, 0.857143 -13.1089 penultimate 0.529412 -13.1089 were actually 0.529412 -13.1089 rbf 0.615385 -13.1089 universally 0.313725 -13.1091 model outperforms 0.7 -13.1093 i believe 0.75 -13.1093 client-side 0.857143 -13.1094 rdfs:subclassof 0.857143 -13.1094 very small. 0.44 -13.1094 π 0.5625 -13.1095 n-gram precision 0.7 -13.1097 vocal 0.7 -13.1097 sweep 0.192593 -13.1099 participating 0.857143 -13.11 t-he 0.857143 -13.11 erotica 0.857143 -13.11 sj1 0.857143 -13.11 ‘*’ 0.857143 -13.11 vulnerable 0.857143 -13.11 att 0.75 -13.1103 i1 0.75 -13.1103 recast 0.75 -13.1103 advertisements 0.75 -13.1103 ga-extr 0.857143 -13.1105 kernel: 0.857143 -13.1105 craig, 0.145833 -13.1105 denoted 0.75 -13.1108 304 0.75 -13.1108 nmfdep 0.857143 -13.1111 campaigns, 0.857143 -13.1111 optimal, 0.313725 -13.1111 new york, 0.75 -13.1113 elsevier 0.75 -13.1113 combo 0.75 -13.1113 alignment move 0.615385 -13.1113 speling 0.857143 -13.1116 observation: 0.857143 -13.1116 in haiti, 0.857143 -13.1116 german: 0.857143 -13.1116 bestperforming 0.857143 -13.1116 u3 0.857143 -13.1116 daughters 0.857143 -13.1116 either: 0.5625 -13.1116 negative example 0.7 -13.1116 all instances. 0.75 -13.1117 distributed inference 0.75 -13.1117 liwc+_p 0.75 -13.1117 13.8% 0.393939 -13.1119 two additional 0.246914 -13.1119 are quite 0.615385 -13.1121 10] 0.857143 -13.1122 arranz’s 0.857143 -13.1122 keneth 0.75 -13.1122 analysis revealed 0.75 -13.1122 scovnpres,dcl e 0.833333 -13.1453 (see sag 0.833333 -13.1453 exceptions, 0.833333 -13.1453 sparseness, 0.833333 -13.1453 within-sentence 0.636364 -13.1455 nathan 0.183673 -13.1455 values are 0.145923 -13.1457 relevant to 0.571429 -13.1457 (ruppenhofer 0.833333 -13.1458 semeval’07 0.833333 -13.1458 performs competitively 0.833333 -13.1458 average linkage 0.833333 -13.1458 paired fscore 0.833333 -13.1458 visualization, 0.833333 -13.1458 arriving at 0.833333 -13.1458 headdependents 0.833333 -13.1458 hobbs et al. 0.5 -13.1458 workshops 0.636364 -13.146 pen 0.454545 -13.1461 they find 0.75 -13.1462 otherwise). 0.75 -13.1462 16.7% 0.833333 -13.1464 alows 0.833333 -13.1464 anders søgaard 0.833333 -13.1464 2003a; 0.833333 -13.1464 backoff property 0.833333 -13.1464 program u. 0.7 -13.1465 +d 0.7 -13.1465 20000 0.571429 -13.1465 acknowledges 0.75 -13.1467 orderings, 0.413793 -13.1467 are optimized 0.636364 -13.1468 unstated 0.615385 -13.1469 minimal rules 0.833333 -13.1469 exceptional 0.833333 -13.1469 keyphrase generation 0.833333 -13.1469 0.717 0.833333 -13.1469 very nice 0.833333 -13.1469 precedesequal ∗ 0.7 -13.147 nflgame 0.25 -13.1474 phenomena. 0.243902 -13.1475 these three 0.833333 -13.1475 iglesias et al. 0.833333 -13.1475 (cao 0.833333 -13.1475 phrasestructure 0.75 -13.1477 “y 0.454545 -13.1479 length n 0.529412 -13.1479 mwa 0.529412 -13.1479 sd 0.529412 -13.1479 unsuitable 0.833333 -13.1481 we call this method 0.833333 -13.1481 dienes, 0.833333 -13.1481 0.421 0.833333 -13.1481 subtreelets 0.833333 -13.1481 ≥ 0, 0.833333 -13.1481 branching entropy 0.833333 -13.1481 random initialization. 0.75 -13.1482 [ko10] 0.25 -13.1482 1996). 0.222222 -13.1484 patterns are 0.833333 -13.1486 error reduction) 0.833333 -13.1486 57.4% 0.833333 -13.1486 no nst 0.75 -13.1487 (xi,yi) 0.636364 -13.1489 mount 0.615385 -13.1489 ˜null 0.75 -13.1492 maximising 0.833333 -13.1492 tier-based strictly 0.833333 -13.1492 discourages 0.833333 -13.1492 100 iterations, 0.833333 -13.1492 cosine similarities 0.833333 -13.1492 common denominator 0.833333 -13.1492 hidden semi-markov 0.615385 -13.1493 compressions. 0.75 -13.1497 gl 0.833333 -13.1497 0.901 0.833333 -13.1497 tesla-f, 0.615385 -13.1501 ogi 0.75 -13.1502 word’s context 0.833333 -13.1503 execution time 0.833333 -13.1503 part-of-speech tag, 0.833333 -13.1503 0.435 0.833333 -13.1503 ds2 0.833333 -13.1503 lwlm+wordnet 0.529412 -13.1507 def≡ 0.833333 -13.1509 predicate logic 0.833333 -13.1509 architectures, 0.833333 -13.1509 pseudo-relevant 0.833333 -13.1509 (tan 0.833333 -13.1509 10.4% 0.833333 -13.1509 vertex vi 0.833333 -13.1509 truecased 0.75 -13.1512 fel 0.529412 -13.1514 mid 0.615385 -13.1514 co-citation. 0.615385 -13.1514 linguistic patterns 0.615385 -13.1514 γ1 0.833333 -13.1514 significant progress 0.833333 -13.1514 at runtime. 0.833333 -13.1514 argue that, 0.833333 -13.1514 user demographics 0.75 -13.1517 simpler, 0.75 -13.1517 discretized 0.75 -13.1517 enlarge the 0.833333 -13.152 pool, 0.833333 -13.152 row), 0.833333 -13.152 log odds 0.833333 -13.152 n-gram overlaps 0.833333 -13.152 ntcir-7 0.833333 -13.152 prepositional phrase. 0.75 -13.1522 xerox 0.423077 -13.1522 task, but 0.5 -13.1522 promoter 0.529412 -13.1525 not) 0.833333 -13.1525 percentage points, 0.5 -13.1525 hat 0.75 -13.1527 fast approximate 0.636364 -13.1528 non-initial 0.47619 -13.1528 main events 0.615385 -13.153 walk, 0.833333 -13.1531 multiple-word 0.833333 -13.1531 6.3% 0.833333 -13.1531 shift reduce 0.833333 -13.1531 1130 0.833333 -13.1531 out” 0.75 -13.1532 (goldstein et 0.571429 -13.1534 gn 0.228261 -13.1535 " 0.833333 -13.1537 dist – 0.833333 -13.1537 tasks; 0.833333 -13.1537 applicability, 0.833333 -13.1537 95.3 0.833333 -13.1537 multilevel 0.75 -13.1538 axis, 0.4 -13.1538 variance in 0.326087 -13.154 purpose. 0.108434 -13.154 2009), 0.636364 -13.1541 vol 0.833333 -13.1542 about 15% 0.833333 -13.1542 independent views 0.833333 -13.1542 vertically composed 0.833333 -13.1542 (lin, 2004). 0.833333 -13.1542 〈e,t〉1 0.75 -13.1548 (bigram 0.833333 -13.1548 annotation guideline 0.833333 -13.1548 cky algorithm. 0.636364 -13.1549 5gm 0.75 -13.1553 56.8 0.833333 -13.1553 posterior distribution. 0.833333 -13.1559 (plural 0.833333 -13.1559 state item, 0.833333 -13.1559 41.5 0.833333 -13.1559 bresnan 0.833333 -13.1559 ontogene 0.182432 -13.1561 is: 0.26087 -13.1561 are treated 0.636364 -13.1562 twsss 0.75 -13.1563 query-oriented 0.833333 -13.1565 concepts; 0.833333 -13.1565 discussion fora 0.833333 -13.1565 html tags 0.833333 -13.1565 0.293 0.833333 -13.1565 humancomputer 0.319149 -13.1566 predictions are 0.47619 -13.1566 an author 0.115702 -13.1568 5.1 0.833333 -13.157 chicken, 0.833333 -13.157 anaphoric, 0.833333 -13.157 36.5 0.636364 -13.1571 sup 0.290909 -13.1571 independently. 0.423077 -13.1574 step 1 0.423077 -13.1574 3.3, 0.636364 -13.1575 poor, 0.833333 -13.1576 parada 0.833333 -13.1576 geman, 0.833333 -13.1576 strives 0.833333 -13.1576 negation scope 0.833333 -13.1576 whitelaw 0.423077 -13.1577 these methods are 0.75 -13.1578 each entry. 0.636364 -13.1579 dans 0.7 -13.1582 ptb. 0.296296 -13.1582 is indeed 0.833333 -13.1582 movements, 0.833333 -13.1582 human s-patterns 0.833333 -13.1582 boards, 0.833333 -13.1582 (window 0.833333 -13.1582 code-specific 0.5 -13.1582 results, but 0.75 -13.1583 dai 0.75 -13.1583 959 0.47619 -13.1585 quirk et 0.5 -13.1586 event mentions 0.833333 -13.1587 tremendously 0.833333 -13.1587 pseudo-code 0.833333 -13.1587 rijke, 0.833333 -13.1587 tending 0.833333 -13.1587 angeli 0.833333 -13.1587 events” 0.833333 -13.1587 triplets, 0.833333 -13.1587 intensity, 0.296296 -13.1588 0 5 10 0.7 -13.1592 introducing new 0.7 -13.1592 pesa 0.371429 -13.1593 ece 0.833333 -13.1593 graphs) 0.833333 -13.1593 wilkins, 0.833333 -13.1593 core12 + 0.833333 -13.1593 elucidate 0.833333 -13.1593 near future, 0.833333 -13.1593 it] 0.211009 -13.1596 we propose to 0.529412 -13.1596 cos 0.5 -13.1596 green, 0.75 -13.1598 lsvm 0.833333 -13.1598 treebank-based 0.833333 -13.1598 expecting 0.833333 -13.1598 4we use 0.833333 -13.1598 many-to-1 0.833333 -13.1598 n0, 0.833333 -13.1598 deteriorate 0.833333 -13.1598 cutoff, 0.833333 -13.1598 anchor detection 0.44 -13.1599 under these 0.5 -13.1599 this assumption is 0.75 -13.1603 wilkins 0.571429 -13.1603 colin 0.833333 -13.1604 ence 0.833333 -13.1604 murakami 0.833333 -13.1604 loads 0.833333 -13.1604 pruned, 0.833333 -13.1604 f• 0.7 -13.1606 formation. 0.75 -13.1608 by letting 0.5 -13.1609 p1, 0.529412 -13.161 popescu and 0.833333 -13.161 dialog cohesion 0.833333 -13.161 domain experts, 0.833333 -13.161 causal relations. 0.833333 -13.161 lf patterns 0.131673 -13.161 edinburgh, scotland, 0.454545 -13.1613 by reducing 0.5 -13.1613 (2007), which 0.387097 -13.1614 can get 0.75 -13.1614 a16 0.529412 -13.1614 summarization task. 0.833333 -13.1615 exploitable 0.833333 -13.1615 accumulates 0.833333 -13.1615 utilities 0.833333 -13.1615 (cook 0.833333 -13.1615 0.730 0.833333 -13.1615 1990s 0.75 -13.1619 reranking algorithm 0.833333 -13.1621 (xu, 0.833333 -13.1621 42.7% 0.833333 -13.1621 follows immediately 0.833333 -13.1621 0.313 0.833333 -13.1621 rodriguez 0.833333 -13.1621 shepherd, 0.833333 -13.1621 indirectly, 0.47619 -13.1621 singer, 0.141667 -13.1627 beginning 0.833333 -13.1627 79.9 0.833333 -13.1627 lowest-scoring 0.833333 -13.1627 na, 0.833333 -13.1627 reportable-event 0.833333 -13.1627 antonymy 0.833333 -13.1627 lecture, 0.833333 -13.1627 an n-dimensional 0.833333 -13.1627 (sporleder 0.75 -13.1629 “1” 0.75 -13.1629 dont 0.75 -13.1629 y: 0.068599 -13.1631 a new 0.833333 -13.1632 (carreras, 0.833333 -13.1632 6.2.2 0.833333 -13.1632 fsm 0.833333 -13.1632 hamper 0.833333 -13.1632 ocean 0.833333 -13.1632 burges, 0.833333 -13.1632 0.669 0.833333 -13.1632 d2, 0.833333 -13.1632 (ea) 0.833333 -13.1632 (velldal 0.833333 -13.1632 generalizability 0.7 -13.1635 crossvalidation. 0.636364 -13.1635 obtain: 0.833333 -13.1638 phonemes, 0.833333 -13.1638 mitkov 0.833333 -13.1638 ranking) 0.833333 -13.1638 patients, 0.833333 -13.1638 fairly easy 0.833333 -13.1638 0.845 0.833333 -13.1638 sparsification 0.833333 -13.1638 min zhang, 0.833333 -13.1638 0.946 0.111111 -13.1638 if it 0.5 -13.1643 section 00 0.7 -13.1644 (girju 0.7 -13.1644 spelling. 0.7 -13.1644 heng 0.833333 -13.1644 uninteresting 0.833333 -13.1644 all-pairs 0.833333 -13.1644 kappa, 0.571429 -13.1646 communicates 0.454545 -13.1647 logs. 0.75 -13.1649 average ratings 0.75 -13.1649 schema, 0.833333 -13.1649 pseudo-labeled 0.833333 -13.1649 m ’@ 0.833333 -13.1649 (jones 0.833333 -13.1649 vsm, 0.833333 -13.1649 poly-co 0.833333 -13.1649 ¯m 0.833333 -13.1649 ,v) 0.833333 -13.1649 corn, 0.833333 -13.1649 operative 0.833333 -13.1649 rip.v, 0.833333 -13.1649 b termino-ontology 0.833333 -13.1649 stoyan 0.833333 -13.1649 jmlm 0.296296 -13.1651 far more 0.75 -13.1654 483 0.833333 -13.1655 22.3 0.833333 -13.1655 assessments, 0.833333 -13.1655 spanish), 0.833333 -13.1655 frame) 0.833333 -13.1655 7.6 0.833333 -13.1655 lexicographers, 0.833333 -13.1655 asr system, 0.833333 -13.1655 venn 0.833333 -13.1655 (hara 0.833333 -13.1655 ك 0.833333 -13.1655 pvalue 0.833333 -13.1655 x∈x summationdisplay 0.833333 -13.1655 x; 0.636364 -13.1657 instance-instance 0.833333 -13.1661 stopwords) 0.833333 -13.1661 tem 0.833333 -13.1661 address, 0.833333 -13.1661 (phrase 0.833333 -13.1661 imperfect, 0.44 -13.1665 are you 0.75 -13.1665 package, 0.319149 -13.1666 characters are 0.833333 -13.1666 year) 0.833333 -13.1666 79.8 0.833333 -13.1666 widelyused 0.833333 -13.1666 algorithms; 0.833333 -13.1666 punkte 0.833333 -13.1666 vossen 0.833333 -13.1666 soderland, 0.75 -13.167 r=1 0.274194 -13.167 this method is 0.833333 -13.1672 0.416 0.833333 -13.1672 penstein-ros´e, 0.833333 -13.1672 nov. 0.833333 -13.1672 feature schemas 0.833333 -13.1672 instructions uttered 0.833333 -13.1672 (pardo 0.833333 -13.1672 meteor (lavie 0.833333 -13.1672 compresses 0.833333 -13.1672 oberlander, 0.833333 -13.1678 65.2 0.833333 -13.1678 [felzenszwalb 0.833333 -13.1678 composable 0.833333 -13.1678 similar) 0.833333 -13.1678 variables) 0.75 -13.168 :( 0.75 -13.168 inexpensive 0.636364 -13.1683 hfl 0.833333 -13.1683 inbetween 0.833333 -13.1683 nonconstituent 0.833333 -13.1683 nokia 0.833333 -13.1683 pre-nominal 0.833333 -13.1683 (walker, 0.833333 -13.1683 benefits: 0.833333 -13.1683 1.26 0.290909 -13.1683 is rather 0.75 -13.1685 names: 0.833333 -13.1689 (rizzolo 0.833333 -13.1689 standardisation 0.833333 -13.1689 p,q 0.833333 -13.1689 0.432 0.833333 -13.1689 punish 0.833333 -13.1689 test-taker 0.833333 -13.1689 trillions 0.833333 -13.1689 irbc 0.833333 -13.1689 foreground 0.833333 -13.1689 additive reliability 0.833333 -13.1689 sarkar, 0.47619 -13.1689 primitives. 0.75 -13.169 8-class 0.75 -13.1695 1.) 0.833333 -13.1695 1.66 0.833333 -13.1695 0.796 0.833333 -13.1695 benchmarking 0.151659 -13.1697 move 0.571429 -13.17 bulgarian, 0.833333 -13.1701 and aravind 0.833333 -13.1701 substantive 0.454545 -13.1706 copy of 0.75 -13.1706 genericity 0.75 -13.1706 emea 0.833333 -13.1706 seneff, 0.833333 -13.1706 tutorials 0.833333 -13.1706 0.853 0.833333 -13.1706 cuteforce 0.209091 -13.1712 been applied 0.833333 -13.1712 aply 0.833333 -13.1712 96.1 0.833333 -13.1712 max-product 0.833333 -13.1712 hindi-punjabi 0.833333 -13.1712 10fold 0.833333 -13.1712 (2007)’s 0.636364 -13.1714 gulf 0.75 -13.1716 initializes 0.75 -13.1716 infected 0.636364 -13.1718 pierre 0.423077 -13.1718 chodorow 0.833333 -13.1718 “hallucinated” 0.833333 -13.1718 0.673 0.833333 -13.1718 (library 0.833333 -13.1718 brew, 0.833333 -13.1718 halteren, 0.833333 -13.1718 pressed 0.833333 -13.1718 fluctuations 0.47619 -13.1718 rolesets 0.109453 -13.1721 rest 0.833333 -13.1723 link¨oping 0.833333 -13.1723 25%, 0.833333 -13.1723 0.716 0.833333 -13.1723 (formula 0.833333 -13.1723 successes 0.833333 -13.1723 2-edge 0.833333 -13.1723 (odni), intelligence 0.833333 -13.1723 settles 0.833333 -13.1723 term-sentence 0.833333 -13.1723 first-pass 0.75 -13.1726 7: end 0.529412 -13.1726 conceivable 0.264706 -13.1726 each cluster 0.833333 -13.1729 conversions 0.833333 -13.1729 object-only 0.833333 -13.1729 reitter 0.833333 -13.1729 reversing 0.833333 -13.1729 examples), 0.833333 -13.1729 valences 0.833333 -13.1729 referees, 0.833333 -13.1729 'computer' 0.833333 -13.1729 compensates 0.833333 -13.1729 gonzalez, 0.833333 -13.1735 (automatic 0.833333 -13.1735 37.2 0.833333 -13.1735 12.4% 0.75 -13.1736 implicativity and 0.571429 -13.1736 6-gram 0.5 -13.1739 graphic’s 0.636364 -13.174 kristina 0.833333 -13.1741 2003] 0.833333 -13.1741 empty) 0.833333 -13.1741 gence 0.833333 -13.1741 56.1 0.833333 -13.1741 pbase 0.833333 -13.1741 menus 0.833333 -13.1741 alphabets, 0.833333 -13.1741 phone, 0.833333 -13.1741 (idf) 0.833333 -13.1741 re-orderings 0.361111 -13.1742 (wiebe 0.75 -13.1742 c&j 0.571429 -13.1743 pos-tags, 0.833333 -13.1746 malakasiotis 0.833333 -13.1746 (boyd 0.833333 -13.1746 322 0.833333 -13.1746 (hirschberg 0.833333 -13.1746 task), 0.833333 -13.1746 advocate 0.833333 -13.1746 regular, 0.833333 -13.1746 afterglow 0.833333 -13.1746 0.908 0.833333 -13.1746 criticizing 0.75 -13.1747 y¸@ 0.75 -13.1747 grammarfactorizationbytreedecomposition 0.636364 -13.1749 fronting 0.833333 -13.1752 0.340 0.833333 -13.1752 0.880 0.833333 -13.1752 (taylor 0.833333 -13.1752 778 0.833333 -13.1752 research fund 0.833333 -13.1752 (sebastiani, 0.833333 -13.1752 (2002)) 0.833333 -13.1752 datasets; 0.833333 -13.1752 easily, 0.833333 -13.1752 investigated, 0.3125 -13.1753 decoder, 0.529412 -13.1755 key difference 0.75 -13.1757 cross-lingual information 0.833333 -13.1758 sindhwani 0.833333 -13.1758 whissell’s 0.833333 -13.1758 contextword 0.833333 -13.1758 (bloehdorn 0.833333 -13.1758 inter-date 0.571429 -13.1759 (o 0.75 -13.1762 transport, 0.833333 -13.1763 solicit 0.833333 -13.1763 26th 0.833333 -13.1763 (pearl, 0.833333 -13.1763 tf-idf, 0.833333 -13.1763 (french 0.833333 -13.1763 uttering 0.833333 -13.1763 tpo 0.833333 -13.1763 further subdivided 0.833333 -13.1763 figure 4(a) 0.75 -13.1767 svm-light 0.75 -13.1767 permutations, 0.833333 -13.1769 discarded; 0.833333 -13.1769 ineffective 0.833333 -13.1769 obstacles 0.833333 -13.1769 80.3% 0.833333 -13.1769 target-style 0.833333 -13.1769 (macwhinney, 0.833333 -13.1769 adaption 0.833333 -13.1769 0.630 0.149533 -13.1771 al., 2010; 0.636364 -13.1775 foreach 0.833333 -13.1775 geertzen 0.833333 -13.1775 open/closed 0.833333 -13.1775 bands, 0.833333 -13.1775 college, 0.833333 -13.1775 12.5% 0.833333 -13.1775 jumped 0.75 -13.1778 dialectological 0.571429 -13.1779 extracted paraphrases 0.529412 -13.1781 adaptations 0.833333 -13.1781 2001); 0.833333 -13.1781 chinese-english, 0.833333 -13.1781 92.20 0.833333 -13.1781 proach 0.833333 -13.1781 w1j 0.833333 -13.1781 (leaman 0.833333 -13.1781 downey 0.833333 -13.1781 fleischman 0.75 -13.1783 851 0.636364 -13.1784 unigram distribution 0.833333 -13.1786 incrementing 0.833333 -13.1786 highest, 0.833333 -13.1786 earlier versions 0.833333 -13.1786 supplementing 0.215686 -13.1786 results were 0.423077 -13.179  0.833333 -13.1792 (2006; 0.833333 -13.1792 1017 0.75 -13.1793 prohibit 0.636364 -13.1797 adhoc 0.636364 -13.1797 time t. 0.636364 -13.1797 interplay 0.75 -13.1798 m2, 0.833333 -13.1798 not), 0.833333 -13.1798 (goldsmith, 0.833333 -13.1798 question raised 0.833333 -13.1798 pos-tagger, 0.833333 -13.1798 690 0.833333 -13.1798 person-x 0.833333 -13.1798 +null-aalign 0.0635593 -13.1801 identify 0.636364 -13.1801 (goldberg 0.211538 -13.1802 disambiguation. 0.174194 -13.1803 we evaluated 0.833333 -13.1804 yan, 0.833333 -13.1804 telephone, 0.833333 -13.1804 threshold-20 0.833333 -13.1804 parse) 0.833333 -13.1804 berger 0.423077 -13.1805 automatically extract 0.127586 -13.1807 treated 0.47619 -13.1807 as positive. 0.75 -13.1809 ,a 0.75 -13.1809 and alon 0.833333 -13.1809 northwestern 0.833333 -13.1809 (tanabe 0.833333 -13.1809 kilicoglu 0.833333 -13.1809 anyway, 0.833333 -13.1809 alternation, 0.833333 -13.1809 seneff 0.833333 -13.1809 top-most 0.833333 -13.1809 0.725 0.833333 -13.1809 subgraph, 0.833333 -13.1809 expert users 0.221053 -13.181 national science 0.571429 -13.1814 and llr 0.833333 -13.1815 83.7 0.833333 -13.1815 0.846 0.833333 -13.1815 internship 0.833333 -13.1815 pages) 0.333333 -13.1818 useful information 0.833333 -13.1821 preferred, 0.833333 -13.1821 extracting bacteria 0.833333 -13.1821 discernible 0.833333 -13.1821 qtb-test 0.833333 -13.1821 80.7% 0.833333 -13.1821 (mairesse 0.833333 -13.1821 callisonburch 0.833333 -13.1821 lkb 0.833333 -13.1821 0.542 0.375 -13.1821 assignment. 0.109974 -13.1822 specifically, 0.636364 -13.1823 hunter 0.75 -13.1824  peak 0.253521 -13.1824 for selecting 0.571429 -13.1826 emily 0.833333 -13.1827 (training) 0.833333 -13.1827 olive 0.833333 -13.1827 q-r 0.833333 -13.1827 lam 0.833333 -13.1827 pop-up 0.833333 -13.1827 magazines 0.833333 -13.1827 partner, 0.833333 -13.1827 dlfs 0.833333 -13.1827 rivers 0.636364 -13.1828 ui, 0.75 -13.1829 ford 0.341463 -13.1829 snyder 0.833333 -13.1833 kalish, 0.833333 -13.1833 timbl (daelemans 0.833333 -13.1833 0.561 0.833333 -13.1833 (cs) 0.833333 -13.1833 states) 0.833333 -13.1833 collocation errors. 0.833333 -13.1833 0.728 0.833333 -13.1833 whispers 0.833333 -13.1833 revu-it 0.47619 -13.1833 aux 0.833333 -13.1838 optimizers 0.833333 -13.1838 priming 0.833333 -13.1838 eat, 0.833333 -13.1838 anymalign 0.833333 -13.1838 text-to-scene 0.833333 -13.1838 translations; 0.833333 -13.1838 wellformed 0.833333 -13.1838 venues 0.833333 -13.1838 right), 0.833333 -13.1838 (gambell 0.833333 -13.1838 deciding, 0.833333 -13.1838 capacity, 0.833333 -13.1838 bikes, 0.833333 -13.1838 perform well, 0.833333 -13.1838 blitzer, 0.833333 -13.1838 1it 0.107843 -13.1843 time. 0.833333 -13.1844 (definition 0.833333 -13.1844 normally, 0.833333 -13.1844 case insensitive 0.833333 -13.1844 (komodakis 0.833333 -13.1844 realisation, 0.833333 -13.1844 column), 0.833333 -13.1844 (schuler, 0.833333 -13.1844 memorizes 0.833333 -13.1844 based java 0.75 -13.1845 francis, 0.75 -13.1845 sentence constructions 0.75 -13.1845 typically require 0.636364 -13.1845 c-feel-it: 0.529412 -13.185 (gold 0.833333 -13.185 as” 0.833333 -13.185 textrunner, 0.833333 -13.185 0.590 0.833333 -13.185 zero) 0.833333 -13.185 rankers, 0.833333 -13.185 stephens 0.833333 -13.185 online-behavior + 0.833333 -13.185 happening 0.833333 -13.185 skips 0.833333 -13.185 (breiman, 0.833333 -13.185 0.371 0.833333 -13.185 direct-mapped 0.833333 -13.185 undo 0.833333 -13.185 laso 0.119048 -13.1851 acts 0.636364 -13.1854 main purpose 0.533333 -13.1854 1.00 1.00 0.833333 -13.1856 no-oth 0.833333 -13.1856 tapping 0.833333 -13.1856 aided 0.833333 -13.1856 affixal 0.833333 -13.1856 (cai 0.833333 -13.1856 0.208 0.833333 -13.1856 inheritance, 0.833333 -13.1856 sub-event 0.454545 -13.186 an event. 0.423077 -13.1861 heuristic, 0.833333 -13.1861 3.6% 0.833333 -13.1861 regime 0.833333 -13.1861 riyadh 0.833333 -13.1861 completions, 0.833333 -13.1861 maximalization, 0.833333 -13.1861 newscommentary 0.833333 -13.1861 369 0.253521 -13.1864 matrix, 0.636364 -13.1867 sla 0.833333 -13.1867 allophony 0.833333 -13.1867 multi-threaded 0.833333 -13.1867 completely, 0.833333 -13.1867 labelling, 0.833333 -13.1867 sscrf 0.107317 -13.1869 seems 0.529412 -13.1869 xeon 0.296296 -13.1873 figure 6 0.833333 -13.1873 germann 0.833333 -13.1873 finin, 0.833333 -13.1873 theune 0.833333 -13.1873 shared-task 0.833333 -13.1873 12.6 0.833333 -13.1873 evans, 0.833333 -13.1873 0.132 0.75 -13.1876 ∅, 0.636364 -13.1876 which avoids 0.571429 -13.1878 impute 0.833333 -13.1879 gesmundo 0.833333 -13.1879 sub-structure 0.833333 -13.1879 ir systems 0.833333 -13.1879 atention 0.833333 -13.1879 communicated 0.833333 -13.1879 r−→r) 0.833333 -13.1879 t-pos 0.833333 -13.1879 (third 0.833333 -13.1879 wellestablished 0.833333 -13.1879 (partially 0.833333 -13.1879 (asr), 0.833333 -13.1879 “closed” 0.833333 -13.1879 abcd 0.833333 -13.1879 1977), 0.571429 -13.1882 also explored 0.175325 -13.1885 to increase 0.833333 -13.1885 41.02 0.833333 -13.1885 heidelberg, 0.833333 -13.1885 lucky 0.833333 -13.1885 flickinger 0.833333 -13.1885 time), 0.833333 -13.1885 episodeduration 0.833333 -13.1885 raux, 0.833333 -13.1885 em), 0.833333 -13.1885 computational efficiency 0.113514 -13.1887 - - 0.47619 -13.1889 style. 0.833333 -13.189 “white 0.833333 -13.189 impeded 0.833333 -13.189 accurate, but 0.833333 -13.189 (godfrey 0.833333 -13.189 formalising 0.833333 -13.189 roy, 0.833333 -13.189 perceptions 0.833333 -13.189 preparatory 0.833333 -13.189 qi, 0.833333 -13.189 (mp) 0.833333 -13.1896 organic 0.833333 -13.1896 mairesse 0.833333 -13.1896 sdc 0.833333 -13.1896 4our 0.833333 -13.1896 folow 0.833333 -13.1896 52.1% 0.833333 -13.1896 37.4 0.833333 -13.1896 redefined 0.833333 -13.1896 i buy 0.833333 -13.1896 string-totree 0.833333 -13.1896 different) 0.833333 -13.1896 irstlm, 0.833333 -13.1896 advancing 0.833333 -13.1896 tying 0.833333 -13.1896 pot 0.75 -13.1897 computer-mediated 0.833333 -13.1902 twente 0.833333 -13.1902 etc, 0.833333 -13.1902 lot, 0.833333 -13.1902 2002); 0.833333 -13.1902 (friedman 0.833333 -13.1902 (henderson 0.833333 -13.1908 (biemann, 0.833333 -13.1908 fusional 0.833333 -13.1908 distract 0.833333 -13.1908 advocates 0.571429 -13.191 our crowdsourced 0.269841 -13.1911 agglomerative 0.636364 -13.1912 weights w 0.75 -13.1913 poker 0.75 -13.1913 u2 0.833333 -13.1914 ture 0.833333 -13.1914 3.5, 0.833333 -13.1914 (ashburner 0.833333 -13.1914 18.9 0.833333 -13.1914 declined 0.833333 -13.1914 door, 0.47619 -13.1916 rather than using 0.75 -13.1918 enull 0.75 -13.1918 presentedin 0.264706 -13.1918 retrieval. 0.5 -13.1919 figure 7. 0.833333 -13.192 wn-4f 0.833333 -13.192 qualifications 0.833333 -13.192 50s 0.833333 -13.192 (matusov 0.833333 -13.192 (baeza-yates 0.833333 -13.192 34.3 0.833333 -13.192 extrapolate 0.833333 -13.192 skantze, 0.833333 -13.192 threshold-3 0.342105 -13.1922 counts are 0.75 -13.1923 highly significant 0.833333 -13.1925 (types) 0.833333 -13.1925 totaling 0.833333 -13.1925 crossover 0.833333 -13.1925 crashed 0.833333 -13.1925 set-expansion 0.833333 -13.1925 web-based measures 0.3125 -13.1927 acyclic 0.25 -13.1927 also has 0.833333 -13.1931 input) 0.833333 -13.1931 rewritings 0.833333 -13.1931 (lari 0.833333 -13.1931 arrival 0.833333 -13.1931 each trial, 0.833333 -13.1931 iteratively, 0.833333 -13.1931 right; 0.833333 -13.1931 645 0.833333 -13.1931 markup) 0.75 -13.1934 in field-s 0.833333 -13.1937 2005b) 0.833333 -13.1937 inter-judge 0.833333 -13.1937 inhouse 0.833333 -13.1937 bible, 0.833333 -13.1937 (ferrucci 0.833333 -13.1937 138632 0.833333 -13.1937 hcr-test 0.833333 -13.1937 6this 0.833333 -13.1937 morpha 0.833333 -13.1937 (leusch 0.454545 -13.1939 traced 0.533333 -13.1939 powers 0.636364 -13.1943 compact, 0.833333 -13.1943 ingredient 0.833333 -13.1943 (hdp) 0.833333 -13.1943 non-speech 0.833333 -13.1943 pos-taggers 0.833333 -13.1943 sumita, 0.833333 -13.1949 domain-specific, 0.833333 -13.1949 studied, 0.75 -13.195 combination: 0.533333 -13.1951 impractical to 0.636364 -13.1952 co, 0.26087 -13.1954 been made 0.5 -13.1954 prevalence of 0.75 -13.1955 varies between 0.833333 -13.1955 (sekine, 0.833333 -13.1955 yuan 0.833333 -13.1955 co-reference, 0.318182 -13.1956 center of 0.407407 -13.1959 count. 0.75 -13.196 rudolph 0.833333 -13.196 being released 0.833333 -13.196 weinberger 0.833333 -13.196 lion 0.833333 -13.196 (choudhury 0.833333 -13.196 elements; 0.636364 -13.1961 sim-k 0.833333 -13.1966 force, 0.833333 -13.1966 disliked 0.0943396 -13.1967 moreover, 0.833333 -13.1972 experiments; 0.833333 -13.1972 conceptresolver, 0.833333 -13.1972 eficient 0.833333 -13.1972 n-grams) 0.833333 -13.1972 (t1,t2) 0.636364 -13.1974 sts 0.75 -13.1976 bakeoff 0.833333 -13.1978 governmental 0.833333 -13.1978 7.3% 0.833333 -13.1978 98.7 0.833333 -13.1978 entity-aspect 0.833333 -13.1978 5.8% 0.833333 -13.1978 troublesome 0.833333 -13.1978 appends 0.833333 -13.1978 definites, 0.75 -13.1981 goes up 0.211538 -13.1983 reordering model 0.833333 -13.1984 yeh 0.833333 -13.1984 transpose 0.833333 -13.1984 subsection, 0.833333 -13.1984 promotional 0.833333 -13.1984 nonanaphoric 0.833333 -13.1984 pp) 0.833333 -13.1984 conceptnet, 0.833333 -13.1984 strange 0.833333 -13.1984 0.118 0.306122 -13.1985 however, there are 0.571429 -13.1986 step 3 0.407407 -13.1988 the x-axis 0.833333 -13.199 rewording 0.833333 -13.199 pyramid, 0.833333 -13.199 perhaps, 0.833333 -13.199 0.324 0.361111 -13.1991 (excluding 0.0660465 -13.1991 must 0.533333 -13.1996 felice 0.833333 -13.1996 regimes 0.833333 -13.1996 politicians, 0.833333 -13.1996 drilling 0.833333 -13.1996 posits 0.351351 -13.1997 emotions in 0.351351 -13.1997 erk 0.5 -13.1997 estimation, 0.5 -13.1997 line shows 0.833333 -13.2002 meng, 0.833333 -13.2002 (williams, 0.833333 -13.2002 real world. 0.833333 -13.2002 50.8 0.833333 -13.2002 2.20 0.833333 -13.2002 swapped 0.833333 -13.2007 cleanly 0.833333 -13.2007 win3 0.833333 -13.2007 paramount 0.833333 -13.2007 sketches 0.833333 -13.2007 sinv 0.636364 -13.201 passion 0.407407 -13.2011 th e 0.75 -13.2013 t3, 0.833333 -13.2013 genetics 0.833333 -13.2013 hyponymy, 0.833333 -13.2013 spellings, 0.833333 -13.2013 763 0.833333 -13.2013 metrical 0.833333 -13.2013 closed, 0.0910714 -13.2014 calculated 0.145455 -13.2015 2003), 0.636364 -13.2019 p(%) 0.833333 -13.2019 a lot. 0.833333 -13.2019 banks 0.833333 -13.2019 dose 0.833333 -13.2019 49.12 0.833333 -13.2019 lefff 0.833333 -13.2019 (mylonakis 0.833333 -13.2019 leaking 0.130435 -13.2023 level, 0.636364 -13.2023 approximately one 0.223404 -13.2024 manning 0.75 -13.2024 query: 0.75 -13.2024 under grants 0.174194 -13.2024 this way, 0.833333 -13.2025 bohus, 0.833333 -13.2025 7note that 0.833333 -13.2025 outputting 0.833333 -13.2025 interviewer 0.833333 -13.2025 58.7 0.833333 -13.2025 (srl), 0.833333 -13.2025 ext100 0.833333 -13.2025 met, 0.833333 -13.2025 0.934 0.833333 -13.2025 pozo 0.571429 -13.2026 rare words 0.533333 -13.2029 hobbs 0.253521 -13.2031 parse. 0.833333 -13.2031 98.0 0.833333 -13.2031 chelba 0.833333 -13.2031 military, 0.833333 -13.2031 intractability 0.833333 -13.2031 finch 0.833333 -13.2031 detectable 0.833333 -13.2031 (u) 0.833333 -13.2031 (soricut 0.636364 -13.2032 metonymies 0.473684 -13.2033 we summarize the 0.75 -13.2034 αv 0.75 -13.2034 phrasebased machine 0.146789 -13.2035 (ii) 0.833333 -13.2037 0.445 0.833333 -13.2037 competition, 0.833333 -13.2037 lan 0.833333 -13.2037 pre-trained 0.434783 -13.2039 software. 0.571429 -13.2042 op 0.833333 -13.2043 expertise, 0.833333 -13.2043 tsarfaty 0.833333 -13.2043 0.400 0.833333 -13.2043 relation’s 0.833333 -13.2043 marx 0.833333 -13.2043 uploaded 0.833333 -13.2043 ξ(not 0.269841 -13.2044 nnp 0.75 -13.2045 come into 0.181159 -13.2048 expressions. 0.833333 -13.2049 ¯d(i) 0.833333 -13.2049 immersive 0.833333 -13.2049 fiction, 0.833333 -13.2049 substances 0.833333 -13.2049 (rs) 0.833333 -13.2049 bw 0.833333 -13.2049 0.792 0.833333 -13.2049 17.6 0.833333 -13.2049 iida 0.833333 -13.2049 depicting 0.636364 -13.205 guodong 0.333333 -13.2051 experiment was 0.232558 -13.2054 together. 0.571429 -13.2054  0.833333 -13.2055 phonologically 0.833333 -13.2055 persist 0.833333 -13.2055 lexical-prosodic 0.833333 -13.2055 weight) 0.75 -13.2056 wordnet-based semantic 0.18797 -13.2057 availability 0.434783 -13.2057 iphone 0.636364 -13.206 cs-en 0.833333 -13.206 matrax 0.833333 -13.206 manshadi 0.833333 -13.206 lemma-tag 0.833333 -13.206 flexibility, 0.833333 -13.206 classification; 0.833333 -13.206 0.907 0.833333 -13.206 38.7 0.833333 -13.206 insure 0.833333 -13.206 numbers) 0.75 -13.2061 lieu 0.15122 -13.2062 ratio of 0.571429 -13.2063 these results. 0.833333 -13.2066 699 0.833333 -13.2066 decays 0.833333 -13.2066 bucketed 0.833333 -13.2066 away, 0.833333 -13.2066 biomedicine 0.833333 -13.2066 j(x,y,z) 0.833333 -13.2066 identifier, 0.833333 -13.2066 28.34 0.833333 -13.2066 massively 0.833333 -13.2066 long) 0.833333 -13.2066 predicate: 0.833333 -13.2066 β2 0.833333 -13.2066 syntbip-monods 0.833333 -13.2066 740 0.833333 -13.2066 has gone 0.833333 -13.2072 recording, 0.833333 -13.2072 garera 0.833333 -13.2072 formatted 0.833333 -13.2072 purandare 0.833333 -13.2072 clinicians 0.833333 -13.2072 psychologically 0.833333 -13.2072 rose-reg 0.833333 -13.2072 ease, 0.833333 -13.2072 (r 2 0.636364 -13.2073 standards, 0.5 -13.2074 he-nom 0.636364 -13.2078 [1, 0.833333 -13.2078 0.9) 0.833333 -13.2078 crises, 0.833333 -13.2078 spontaneous, 0.833333 -13.2078 barrier 0.833333 -13.2078 hajiˇc 0.833333 -13.2078 markovization 0.833333 -13.2078 priori, 0.833333 -13.2084 commonly-used 0.833333 -13.2084 lexeme, 0.833333 -13.2084 word– 0.833333 -13.2084 (correct) 0.833333 -13.2084 “some” 0.636364 -13.2087 ̸= 0.0708915 -13.209 fact 0.833333 -13.209 (mwe) 0.833333 -13.209 patterns). 0.833333 -13.209 987 0.833333 -13.209 90k 0.833333 -13.209 σjt 0.833333 -13.209 generalizations, 0.833333 -13.209 2.3.2 0.636364 -13.2091 geneva, 0.636364 -13.2091 chambre 0.75 -13.2093 vds-7 (a) 0.75 -13.2093 work well. 0.533333 -13.2094 these kinds 0.636364 -13.2096 hyperedges. 0.636364 -13.2096 grammar: 0.363636 -13.2096 oxford 0.833333 -13.2096 match; 0.833333 -13.2096 likelihod 0.833333 -13.2096 assistance, 0.833333 -13.2096 frameworks, 0.833333 -13.2096 html, 0.833333 -13.2096 corpus1 0.833333 -13.2096 text-mining 0.833333 -13.2096 jiang-conrath 0.5 -13.2099 2011a). 0.209524 -13.21 be interpreted 0.0951456 -13.21 mentioned 0.387097 -13.2101 using wordnet 0.833333 -13.2102 0.812 0.833333 -13.2102 left), 0.833333 -13.2102 dendrogram 0.75 -13.2103 m ≥ 0.636364 -13.2105 0.333333 -13.2108 press. 0.833333 -13.2108 60k 0.833333 -13.2108 clda, 0.833333 -13.2108 restaurant, 0.833333 -13.2108 0.240 0.833333 -13.2108 patient) 0.833333 -13.2108 (absolute) 0.833333 -13.2108 mirkin 0.318182 -13.2111 improvements on 0.473684 -13.2113 philadelphia, 0.75 -13.2114 commons 0.75 -13.2114 v30] 0.833333 -13.2114 foods 0.833333 -13.2114 earthquakes, 0.833333 -13.2114 parent/child 0.833333 -13.2114 850 0.833333 -13.2114 3our 0.833333 -13.2114 tf·idf 0.833333 -13.2114 character) 0.325581 -13.2117 α. 0.5 -13.2117 ?c 0.636364 -13.2118 both tasks, 0.533333 -13.212 (bunescu 0.833333 -13.212 circumflexed 0.833333 -13.212 in-vehicle 0.833333 -13.212 fellow 0.833333 -13.212 figure 4(b) 0.833333 -13.212 undergone 0.833333 -13.212 1141 0.833333 -13.212 59.1% 0.833333 -13.212 0.801 0.833333 -13.212 true; 0.833333 -13.212 obligations 0.833333 -13.212 modifier 0.833333 -13.2126 partial, 0.833333 -13.2126 binarizing 0.833333 -13.2126 rewriting, 0.833333 -13.2126 0.667 0.833333 -13.2126 kleinberg, 0.833333 -13.2126 rlm1 0.833333 -13.2126 categories) 0.636364 -13.2127 λ3 0.253521 -13.213 are relatively 0.423077 -13.2132 8). 0.833333 -13.2132 targetdependent 0.833333 -13.2132 λ4 0.833333 -13.2132 sim(i,j) 0.833333 -13.2132 mirella 0.833333 -13.2132 (karakos 0.833333 -13.2132 perspectives, 0.833333 -13.2132 isue 0.833333 -13.2132 sieves, 0.2375 -13.2133 entity, 0.75 -13.2136 inhibits 0.666667 -13.2137 variants: 0.666667 -13.2137 yes. 0.473684 -13.2137 same way. 0.833333 -13.2138 wmt’11 0.833333 -13.2138 translations), 0.833333 -13.2138 hindered 0.833333 -13.2138 ‘+’ 0.833333 -13.2138 alabama 0.833333 -13.2138 aho-corasick 0.833333 -13.2138 (bender, 0.833333 -13.2138 60.6 0.833333 -13.2138 rte5 0.833333 -13.2144 (words, 0.833333 -13.2144 survival, 0.833333 -13.2144 0.441 0.833333 -13.2144 intelligence tasks 0.833333 -13.2144 inflecting 0.833333 -13.2144 variance, 0.833333 -13.2144 w2] 0.363636 -13.2145 squaresolid 0.666667 -13.2146 yorktown heights, 0.24359 -13.2148 more important 0.833333 -13.2149 41.3 0.833333 -13.2149 (consisting 0.833333 -13.2149 sorry 0.833333 -13.2149 italian, spanish, 0.833333 -13.2149 pronominal, 0.833333 -13.2149 lm-seeded 0.833333 -13.2149 1.20 0.833333 -13.2149 verhagen 0.833333 -13.2149 good-turing 0.5 -13.2149 validate our 0.423077 -13.215 7.3 0.833333 -13.2155 overlooked 0.833333 -13.2155 entrainment, 0.833333 -13.2155 annoying 0.833333 -13.2155 0.882 0.833333 -13.2155 relations” 0.636364 -13.2159 h) 0.583333 -13.2161 latent-variable 0.833333 -13.2161 contend 0.833333 -13.2161 more amenable 0.833333 -13.2161 show that: 0.833333 -13.2161 hh 0.833333 -13.2161 species, 0.833333 -13.2161 (tillmann, 0.833333 -13.2161 dras, 0.833333 -13.2161 1a) 0.833333 -13.2161 subtype, 0.833333 -13.2161 µ2 0.833333 -13.2161 5.2) 0.833333 -13.2161 super-entities 0.833333 -13.2161 epoch, 0.833333 -13.2161 9.9 0.533333 -13.2162 (ast) 0.342105 -13.2162 sentences). 0.636364 -13.2164 ncbc 0.0292226 -13.2164 figure 0.666667 -13.2166 (rieser 0.833333 -13.2167 1606 0.833333 -13.2167 amplified 0.833333 -13.2167 (entities 0.75 -13.2168 product: 0.583333 -13.2169 mul 0.571429 -13.2173 d’ 0.833333 -13.2173 revises 0.833333 -13.2173 585 0.833333 -13.2173 61.9% 0.833333 -13.2173 (kbp) 0.833333 -13.2173 vandalism, 0.833333 -13.2173 realising 0.833333 -13.2173 dwell 0.583333 -13.2174 joakim 0.666667 -13.2175 boyer, 0.583333 -13.2178 rosé, 0.75 -13.2179 variances 0.833333 -13.2179 modeled, 0.833333 -13.2179 deciphering 0.833333 -13.2179 dir(β) 0.833333 -13.2179 (akin 0.833333 -13.2179 an annotation tool 0.833333 -13.2179 atributes 0.833333 -13.2179 62.5% 0.342105 -13.2181 and vogel, 0.666667 -13.2185 arson 0.666667 -13.2185 lm type 0.833333 -13.2185 zaenen, 0.833333 -13.2185 cuay´ahuitl 0.833333 -13.2185 (7 0.833333 -13.2185 precision), 0.666667 -13.2189 system detects 0.75 -13.2189 agreement rates 0.583333 -13.219 (1a) 0.583333 -13.219 retrieval model 0.454545 -13.2191 crafted 0.636364 -13.2191 3,720m 0.833333 -13.2191 bloom 0.833333 -13.2191 size) 0.833333 -13.2191 (vadas 0.833333 -13.2191 92.9% 0.833333 -13.2191 (z) 0.833333 -13.2191 -0.32 0.533333 -13.2193 discourse context 0.351351 -13.2194 miwa 0.058432 -13.2197 presented 0.533333 -13.2197 pain 0.833333 -13.2197 modal, 0.833333 -13.2197 demonstration, 0.833333 -13.2197 depression 0.833333 -13.2197 773 0.833333 -13.2197 sub-optimal 0.833333 -13.2197 rlm2 0.833333 -13.2197 0.742 0.833333 -13.2197 (2005, 0.833333 -13.2197 7000 0.833333 -13.2197 solved, 0.407407 -13.2197 are displayed 0.666667 -13.2199 0.686 0.583333 -13.2199 candidate answers 0.75 -13.22 lawyers 0.473684 -13.2201 core arguments 0.173333 -13.2202 seen as 0.833333 -13.2203 1985) 0.833333 -13.2203 21.2 0.833333 -13.2203 side-effect 0.833333 -13.2203 hsu 0.833333 -13.2203 (yeh 0.833333 -13.2203 targets: 0.833333 -13.2203 leaning 0.833333 -13.2203 (gabrilovich 0.833333 -13.2203 0.740 0.833333 -13.2203 chua 0.5 -13.2203 avery 0.454545 -13.2204 subspace 0.5 -13.2207 densities 0.533333 -13.2208 distinctions, 0.833333 -13.2209 489 0.833333 -13.2209 dongdong 0.833333 -13.2209 remedial 0.833333 -13.2209 invisible 0.833333 -13.2209 minimal, 0.833333 -13.2209 subordinate-clause 0.833333 -13.2209 echihabi, 0.571429 -13.221 are positively 0.583333 -13.2211 2-grams 0.571429 -13.2214 semantic mapping 0.833333 -13.2215 monothematic 0.833333 -13.2215 word sequences, 0.833333 -13.2215 (thomas 0.833333 -13.2215 clouds 0.833333 -13.2215 pre-processing, 0.833333 -13.2215 katz, 0.833333 -13.2215 imaging 0.833333 -13.2215 more conservative 0.666667 -13.2218 implicit relations 0.666667 -13.2218 (strube 0.583333 -13.222 55.2 0.833333 -13.2221 ur-en 0.833333 -13.2221 mpc 0.833333 -13.2221 unequal 0.833333 -13.2221 ppool 0.833333 -13.2221 cohort 0.833333 -13.2221 acomplish 0.150754 -13.2222 class. 0.666667 -13.2223 targetlanguage 0.666667 -13.2223 spontaneously 0.666667 -13.2223 pobj 0.533333 -13.2224 (erkan 0.434783 -13.2225 tuning set 0.75 -13.2227 ely 0.833333 -13.2227 syntactic roles 0.833333 -13.2227 audiences 0.833333 -13.2227 0.071 0.833333 -13.2227 raw-to-clean 0.833333 -13.2227 content; 0.833333 -13.2227 popescu, 0.833333 -13.2227 28.12 0.833333 -13.2227 used; 0.833333 -13.2227 l-bfgs-b 0.666667 -13.2228 (lu 0.666667 -13.2228 terppara 0.533333 -13.2231 radicalbig 0.454545 -13.2233 both tasks 0.75 -13.2233 ˜ao 0.833333 -13.2233 zheng, 0.833333 -13.2233 postposition, 0.833333 -13.2233 parameters), 0.833333 -13.2233 noun’s 0.666667 -13.2237 email corpus 0.666667 -13.2237 .59 0.75 -13.2238 nnps 0.75 -13.2238 gprime 0.392857 -13.2238 6we 0.392857 -13.2238 discriminative model 0.833333 -13.2239 usergenerated 0.833333 -13.2239 20) 0.833333 -13.2239 around, 0.833333 -13.2239 errorful 0.351351 -13.2245 spearman 0.833333 -13.2245 rows, 0.833333 -13.2245 publication, 0.833333 -13.2245 systemt, 0.833333 -13.2245 anecdotal 0.833333 -13.2245 24.5 0.833333 -13.2245 avoided, 0.325581 -13.2246 original english 0.5 -13.2246 when working 0.666667 -13.2247 synsetcj 0.159091 -13.2247 contribute 0.0706522 -13.2249 instance 0.833333 -13.2251 cores, 0.833333 -13.2251 underlined, 0.833333 -13.2251 transtac 0.833333 -13.2251 full-long 0.833333 -13.2251 coercion, 0.833333 -13.2251 expletive it 0.833333 -13.2251 island, 0.833333 -13.2251 813.50 0.140351 -13.2253 idea of 0.533333 -13.2254 pam 0.833333 -13.2257 a-n 0.833333 -13.2257 task’s 0.833333 -13.2257 heuristic: 0.833333 -13.2257 “source” 0.833333 -13.2257 coconstruction 0.833333 -13.2257 hockenmaier, 0.833333 -13.2257 germany, 0.833333 -13.2257 (2006) report 0.833333 -13.2257 mxnlu 0.583333 -13.2258 76.9 0.151515 -13.226 are very 0.666667 -13.2262 ter−4bleu 0.666667 -13.2262 deltas 0.583333 -13.2262 michel 0.240506 -13.2263 suffer 0.833333 -13.2263 booking 0.833333 -13.2263 accusative, 0.833333 -13.2263 fdbk 0.833333 -13.2263 1.0% 0.833333 -13.2263 positing 0.833333 -13.2263 upto 0.636364 -13.2265 wsem 0.583333 -13.2266 expected value 0.280702 -13.2268 4 evaluation 0.833333 -13.2269 0.228 0.833333 -13.2269 46.6 0.833333 -13.2269 500m 0.833333 -13.2269 text-hypothesis 0.833333 -13.2269 [not 0.833333 -13.2269 ...> 0.833333 -13.2269 let's 0.833333 -13.2269 bve+mdl 0.833333 -13.2269 18.8 0.833333 -13.2269 fu 0.833333 -13.2269 grammaticalinference 0.833333 -13.2269 rarity 0.75 -13.227 unsupervised manner 0.363636 -13.227 yang and 0.666667 -13.2271 (y) 0.169811 -13.2271 a complete 0.325581 -13.2272 word frequency 0.311111 -13.2273 to leverage 0.833333 -13.2275 conceive 0.833333 -13.2275 tweets; 0.833333 -13.2275 pnnl 0.833333 -13.2275 routinely 0.833333 -13.2275 cp, 0.833333 -13.2275 insideoutside 0.833333 -13.2275 52.2% 0.833333 -13.2275 characters; 0.833333 -13.2275 vectorr 0.833333 -13.2275 84.9 0.833333 -13.2275 parallels 0.285714 -13.2278 offset 0.636364 -13.2278 ppl 0.533333 -13.2278 (2009) showed 0.666667 -13.2281 rounding 0.833333 -13.2282 steep 0.833333 -13.2282 phraseology 0.833333 -13.2282 protect 0.833333 -13.2282 arc-factored 0.833333 -13.2282 2005)) 0.833333 -13.2282 left-to-right, 0.833333 -13.2282 kn:bn-small 0.833333 -13.2282 editor’s 0.833333 -13.2282 contact, 0.833333 -13.2282 (highly 0.454545 -13.2286 2010a; 0.265625 -13.2286 english wikipedia 0.833333 -13.2288 should reflect 0.833333 -13.2288 53.51 0.833333 -13.2288 0.026 0.833333 -13.2288 5these 0.833333 -13.2288 em-like 0.833333 -13.2288 comparability, 0.833333 -13.2294 stances 0.833333 -13.2294 milton 0.833333 -13.2294 generators, 0.833333 -13.2294 norel 0.833333 -13.2294 (trained 0.833333 -13.2294 lund 0.833333 -13.2294 much, 0.833333 -13.2294 aclia 0.583333 -13.2296 rel-lda1 0.583333 -13.2296 general vocabulary 0.13913 -13.2298 center 0.666667 -13.23 37.9 0.833333 -13.23 weld 0.533333 -13.2301 headlines. 0.636364 -13.2302 varies from 0.833333 -13.2306 mid-range 0.833333 -13.2306 (baldridge 0.833333 -13.2306 lectures, 0.833333 -13.2306 itself; 0.833333 -13.2306 mansion 0.833333 -13.2306 carefully selected 0.833333 -13.2306 376 0.75 -13.2309 spoken utterances 0.833333 -13.2312 “good 0.833333 -13.2312 0.687 0.833333 -13.2312 736 0.833333 -13.2312 rule’s source 0.583333 -13.2313 94.2 0.583333 -13.2313 .49 0.583333 -13.2313 each chunk 0.666667 -13.2315 penalized for 0.583333 -13.2317 (szpektor 0.583333 -13.2317 our shift-reduce parser 0.333333 -13.2317 lda model 0.833333 -13.2318 kicked 0.833333 -13.2318 def 0.833333 -13.2318 by chance, 0.833333 -13.2318 relabel 0.833333 -13.2318 approve 0.833333 -13.2318 (2005)’s 0.833333 -13.2318 coherence: 0.5 -13.2319 metaphors 0.636364 -13.232 nagata 0.454545 -13.2322 million sentences 0.028869 -13.2322 based 0.666667 -13.2324 dropped from 0.666667 -13.2324 (fr) 0.533333 -13.2324 steps: 1. 0.833333 -13.2324 preliminary, 0.833333 -13.2324 ezekiel 0.833333 -13.2324 passage, 0.833333 -13.2324 armed 0.833333 -13.2324 multinuclear 0.833333 -13.2324 concept; 0.833333 -13.2324 becoming available 0.583333 -13.2325 ranges, 0.583333 -13.2325 tourist 0.636364 -13.2329 345 0.833333 -13.233 komachi 0.833333 -13.233 provides strong 0.833333 -13.233 (mpqa) 0.833333 -13.233 “+” 0.833333 -13.233 500,000 0.833333 -13.233 -0.05 0.833333 -13.233 breaks, 0.833333 -13.233 414 0.833333 -13.233 left; 0.833333 -13.233 rf 0.833333 -13.233 insight, 0.833333 -13.233 english–german 0.833333 -13.233 habitual 0.75 -13.2336 [arg2 0.833333 -13.2336 (bp) 0.833333 -13.2336 linking, 0.833333 -13.2336 threaded 0.833333 -13.2336 tc, 0.833333 -13.2336 pdscore 0.833333 -13.2336 0.962 0.833333 -13.2336 habitats 0.833333 -13.2336 corectnes 0.833333 -13.2336 out-forest 0.833333 -13.2336 synthesizer 0.833333 -13.2336 listwise 0.833333 -13.2336 bi-tags 0.833333 -13.2336 melville 0.833333 -13.2336 writers, 0.583333 -13.2338 bayesian approach 0.3 -13.2339 our research 0.583333 -13.2342 router’s 0.833333 -13.2342 carol 0.833333 -13.2342 papineni 0.833333 -13.2342 paraphrase-augmented 0.833333 -13.2342 shooting 0.833333 -13.2342 customary 0.833333 -13.2342 argn 0.833333 -13.2342 minute, 0.833333 -13.2342 86.3 0.666667 -13.2344 wordends 0.833333 -13.2348 -based 0.833333 -13.2348 lex, 0.833333 -13.2348 morris 0.833333 -13.2348 circumvented 0.833333 -13.2348 11] 0.833333 -13.2348 linguistic-based 0.833333 -13.2348 massachusetts, amherst 0.833333 -13.2348 casual 0.833333 -13.2348 3.1.4 0.833333 -13.2348 medstractplus 0.833333 -13.2348 step; 0.833333 -13.2348 nullk 0.833333 -13.2348 bilexical 0.833333 -13.2348 volunteers, 0.833333 -13.2348 lexicon; 0.666667 -13.2349 reward function, 0.666667 -13.2349 sentence-ranking 0.342105 -13.235 (banerjee 0.533333 -13.2351 sarah 0.392857 -13.2352 were included 0.666667 -13.2354 0.734 0.833333 -13.2354 0.562 0.833333 -13.2354 certificate 0.833333 -13.2354 target–landmark 0.833333 -13.2354 repulsion 0.833333 -13.2354 0.989 0.833333 -13.2354 matching/approximate 0.833333 -13.2354 597 0.636364 -13.2357 raymond 0.0726846 -13.2357 ∈ 0.75 -13.2358 second baseline, 0.75 -13.2358 complex morphology 0.666667 -13.2359 is maintained 0.533333 -13.2359 label candidates 0.833333 -13.236 oepen 0.833333 -13.236 mising 0.833333 -13.236 mcs 0.833333 -13.236 “high” 0.833333 -13.236 patrons 0.833333 -13.236 judges’ 0.833333 -13.236 brands, 0.833333 -13.236 (chance) 0.833333 -13.236 rouge-2, 0.666667 -13.2363 97.9 0.407407 -13.2365 projection. 0.833333 -13.2366 groupings, 0.833333 -13.2366 unhelpful 0.833333 -13.2366 dialogue: 0.833333 -13.2366 web-test 0.833333 -13.2366 oren 0.833333 -13.2366 frey, 0.833333 -13.2366 bisazza 0.666667 -13.2368 subpos 0.257576 -13.2368 have similar 0.25 -13.2369 studies have 0.636364 -13.2371 maximises 0.833333 -13.2373 lynn 0.833333 -13.2373 le, 0.833333 -13.2373 mixed-initiative 0.833333 -13.2373 musicians 0.833333 -13.2373 practical purposes. 0.833333 -13.2373 too big 0.636364 -13.2376 thomson 0.375 -13.2376 neutral. 0.833333 -13.2379 aperiodicity 0.833333 -13.2379 1935 0.833333 -13.2379 store, 0.833333 -13.2379 (tr) 0.833333 -13.2379 cplex 0.833333 -13.2379 3.5.1 0.833333 -13.2379 incurring 0.833333 -13.2379 scrambled 0.833333 -13.2379 14.3% 0.833333 -13.2379 24, 0.833333 -13.2379 53.0 0.833333 -13.2379 39% 0.75 -13.238 θ∗ 0.75 -13.238 feature extraction. 0.636364 -13.238 rte3 0.75 -13.2385 265 0.833333 -13.2385 labeling images 0.833333 -13.2385 mneg 0.833333 -13.2385 plethora 0.833333 -13.2385 takers 0.833333 -13.2385 weischedel 0.833333 -13.2385 31.44 0.833333 -13.2385 resolvers 0.0841121 -13.2386 conclusion 0.666667 -13.2388 tweets: 0.75 -13.2391 get, 0.75 -13.2391 we call such 0.833333 -13.2391 different, but 0.833333 -13.2391 ambiguous: 0.833333 -13.2391 years) 0.833333 -13.2391 echoes 0.833333 -13.2391 (kambhatla, 0.833333 -13.2391 personphoneall 0.833333 -13.2391 tag-triple 0.407407 -13.2392 impasse 0.636364 -13.2394 step: 0.75 -13.2396 tdt5 0.0519713 -13.2397 sets 0.833333 -13.2397 borderline 0.833333 -13.2397 φ(x) 0.0493298 -13.2398 distribution 0.318182 -13.2401 elements are 0.666667 -13.2402 frame: 0.666667 -13.2402 0.346 0.75 -13.2402 wm,n 0.533333 -13.2402 by decreasing 0.833333 -13.2403 s-neut 0.833333 -13.2403 78.32 0.833333 -13.2403 wordalignment 0.833333 -13.2403 chiao 0.833333 -13.2403 rutgers 0.666667 -13.2407 training/testing 0.833333 -13.2409 24.29 0.833333 -13.2409 newspaper, 0.833333 -13.2409 relational database 0.833333 -13.2409 grenager 0.833333 -13.2409 damage, 0.666667 -13.2412 -um 0.666667 -13.2412 0.714 0.75 -13.2413 w: 0.833333 -13.2415 n-word 0.833333 -13.2415 41.19 0.833333 -13.2415 “but” 0.833333 -13.2415 gothenburg 0.833333 -13.2415 irregularities 0.833333 -13.2415 presses 0.833333 -13.2415 i;h;j 0.833333 -13.2415 6for 0.666667 -13.2417 recently introduced 0.833333 -13.2421 conflation 0.833333 -13.2421 job, 0.833333 -13.2421 (yoshida 0.833333 -13.2421 pmc 0.833333 -13.2421 underrepresented 0.833333 -13.2421 pushpak 0.833333 -13.2421 38.5 0.833333 -13.2421 δθ(x) 0.666667 -13.2427 43.5 0.636364 -13.2427 sentence-level sentiment 0.833333 -13.2428 3-best 0.833333 -13.2428 aclia1 0.833333 -13.2428 checkitout+ 0.833333 -13.2428 0.756 0.833333 -13.2428 perception, 0.205607 -13.243 we treat 0.666667 -13.2432 blei, 0.833333 -13.2434 switched 0.833333 -13.2434 clean, 0.833333 -13.2434 14.9 0.833333 -13.2434 cross-modal 0.833333 -13.244 translation”’ 0.833333 -13.244 gold: 0.833333 -13.244 30%, 0.833333 -13.244 before) 0.833333 -13.244 committees 0.666667 -13.2442 p ≤ 0.666667 -13.2442 unparsed 0.666667 -13.2442 2 methodology 0.666667 -13.2442 ordered, 0.342105 -13.2442 cardie 0.636364 -13.2446 +pos 0.833333 -13.2446 dilemma 0.833333 -13.2446 0.790 0.833333 -13.2446 ⊗ 0.833333 -13.2446 “-” 0.833333 -13.2446 5-word 0.37931 -13.2448 (palmer 0.583333 -13.2449 alignment-oriented 0.636364 -13.2451 large improvements 0.833333 -13.2452 misuse 0.833333 -13.2452 whichever 0.833333 -13.2452 nell, 0.833333 -13.2452 α: 0.833333 -13.2452 (fraser 0.833333 -13.2452 framework: 0.833333 -13.2452 sheds 0.833333 -13.2452 (italy) 0.666667 -13.2456 660 0.75 -13.2457 annotation, which 0.533333 -13.2457 vector addition 0.833333 -13.2458 supertagging, 0.833333 -13.2458 disjunctions 0.833333 -13.2458 netscape 0.833333 -13.2458 confines 0.833333 -13.2458 15) 0.833333 -13.2458 non-relational 0.583333 -13.2462 entailment detection 0.833333 -13.2464 ned, 0.833333 -13.2464 interlocutor 0.833333 -13.2464 445 0.833333 -13.2464 floorleftsfloorright[i,j) 0.833333 -13.2464 surely 0.833333 -13.2464 perturbations 0.833333 -13.2464 langlais 0.833333 -13.2464 surface-based 0.833333 -13.2464 apply, 0.666667 -13.2466 +poslm 0.666667 -13.2471 48.0 0.583333 -13.2471 machine translations 0.833333 -13.2471 bootstrapping procedure 0.833333 -13.2471 reusability 0.833333 -13.2471 bagof-word 0.833333 -13.2471 deepest 0.833333 -13.2471 statuses 0.833333 -13.2471 professionals 0.636364 -13.2474 (b 0.833333 -13.2477 nielsen 0.833333 -13.2477 minimalist 0.833333 -13.2477 indiana university 0.833333 -13.2477 ntcir-6 0.833333 -13.2477 arg0, 0.311111 -13.248 θ. 0.833333 -13.2483 departs 0.833333 -13.2483 fnr 0.833333 -13.2483 ttt 0.833333 -13.2483 m&l, 0.833333 -13.2483 intensifiers 0.833333 -13.2483 iii: 0.833333 -13.2483 1to 0.833333 -13.2483 investing 0.833333 -13.2483 speak, 0.833333 -13.2483 (yuan 0.583333 -13.2484 slot4], 0.666667 -13.2486 “you” 0.833333 -13.2489 1hmm 0.833333 -13.2489 1978), 0.833333 -13.2489 assembly 0.833333 -13.2489 p&d 0.833333 -13.2489 qt1 0.833333 -13.2489 katherine 0.833333 -13.2489 (several 0.833333 -13.2489 relata, 0.833333 -13.2495 bv 0.833333 -13.2495 (aligned 0.833333 -13.2495 (primarily 0.833333 -13.2495 extents 0.833333 -13.2495 finalscore 0.833333 -13.2495 226 0.75 -13.2496 ddi} 0.75 -13.2496 interoperability 0.75 -13.2496 eprime∈eh 0.75 -13.2496 computers. 0.75 -13.2496 face. 0.75 -13.2496 dialect. 0.75 -13.2496 herman 0.75 -13.2496 〈det〉the 0.75 -13.2496 landmark’s 0.75 -13.2496 shouldn’t 0.75 -13.2496 genpex. 0.75 -13.2496 statistics-based 0.75 -13.2496 ave. 0.75 -13.2496 edition. 0.75 -13.2496 furry 0.75 -13.2496 covers. 0.45 -13.2499 another example 0.181818 -13.2499 8: 0.833333 -13.2501 basque-spanish 0.833333 -13.2501 o) 0.833333 -13.2501 pentateuch 0.666667 -13.2506 sms data. 0.833333 -13.2508 eneko 0.833333 -13.2508 0.409 0.833333 -13.2508 portions, 0.833333 -13.2508 1.22 0.833333 -13.2508 volatility 0.833333 -13.2508 quest, 0.352941 -13.2512 breath 0.0428336 -13.2513 probability 0.833333 -13.2514 92.85 0.833333 -13.2514 50.5% 0.833333 -13.2514 (directed) 0.833333 -13.2514 examples; 0.833333 -13.2514 between-system 0.833333 -13.2514 lexicographically 0.833333 -13.2514 systematicity 0.0984683 -13.2516 each word 0.636364 -13.2517 unvoiced 0.147059 -13.2518 framework for 0.833333 -13.252 discontinuous, 0.833333 -13.252 winnow 0.833333 -13.252 left/right 0.833333 -13.252 its usefulness 0.833333 -13.252 frontend 0.833333 -13.252 02–21 0.833333 -13.252 touched 0.833333 -13.252 zesch 0.833333 -13.252 201), 0.833333 -13.252 n11 0.833333 -13.252 interregnum 0.666667 -13.2521 xing, 0.636364 -13.2521 language phenomena 0.45 -13.2522 corpus includes 0.266667 -13.2523 a large set 0.666667 -13.2526 40th 0.833333 -13.2526 comunicators 0.833333 -13.2526 seginer’s 0.833333 -13.2526 uncovers 0.833333 -13.2526 collocate 0.833333 -13.2526 turkers’ 0.833333 -13.2526 863 0.201835 -13.2528 has two 0.318182 -13.2528 we analyzed 0.473684 -13.2531 semi-markov 0.45 -13.2532 expects 0.833333 -13.2532 0.461 0.833333 -13.2532 polarity: 0.833333 -13.2532 one-tailed 0.833333 -13.2532 payao 0.833333 -13.2532 monitors 0.407407 -13.2533 engaged 0.636364 -13.2535 routed 0.583333 -13.2536 nn) 0.833333 -13.2539 ragazzo 0.833333 -13.2539 council, 0.666667 -13.254 (bicici 0.352941 -13.2543 (pang et al., 0.294118 -13.2544 seek to 0.583333 -13.2544 milk 0.833333 -13.2545 mecab 0.833333 -13.2545 160k 0.833333 -13.2545 duh 0.833333 -13.2545 m`arquez 0.833333 -13.2545 parse forests 0.533333 -13.2548 dec 0.833333 -13.2551 low-cost 0.833333 -13.2551 544 0.833333 -13.2551 two teams, 0.833333 -13.2551 cleanser 0.833333 -13.2551 though: 0.833333 -13.2551 lok 0.833333 -13.2551 individual adjective 0.833333 -13.2551 acts: 0.833333 -13.2551 37% 0.833333 -13.2557 0.4) 0.833333 -13.2557 space), 0.833333 -13.2557 0.406 0.666667 -13.256 amino 0.416667 -13.2561 each annotator 0.473684 -13.2562 preservation, 0.833333 -13.2563 great, 0.833333 -13.2563 partly funded 0.833333 -13.2563 tptp 0.833333 -13.2563 1.42 0.833333 -13.2563 breadth-first 0.833333 -13.2563 thei-th 0.833333 -13.2563 enthymemes 0.833333 -13.2563 superscripts 0.833333 -13.2563 c1...cn 0.311111 -13.2565 the next section, we 0.45 -13.2569 (lee 0.666667 -13.257 alphabetical 0.833333 -13.257 volume, 0.833333 -13.257 h2, 0.833333 -13.257 (hybrid) 0.45 -13.2572 godis 0.636364 -13.2573 dolan 0.636364 -13.2573 jordan” 0.636364 -13.2573 “are 0.473684 -13.2573 pattern grammar 0.5 -13.2573 lexical forms 0.666667 -13.2575 model (vogel 0.45 -13.2576 report, 0.833333 -13.2576 killer 0.833333 -13.2576 msol 0.833333 -13.2576 xia and 0.833333 -13.2576 lexicalizing 0.833333 -13.2576 bohnet 0.833333 -13.2576 9%, 0.533333 -13.258 2007a). 0.833333 -13.2582 b-matching graph 0.833333 -13.2582 unexpressed 0.833333 -13.2582 qsb 0.833333 -13.2582 2.56 0.833333 -13.2582 sense distinctions, 0.833333 -13.2582 (dependency 0.833333 -13.2582 low-density 0.583333 -13.2584 grefenstette 0.666667 -13.2585 unobtrusive 0.266667 -13.2588 markov models 0.186992 -13.2588 nodes are 0.833333 -13.2588 28.2 0.833333 -13.2588 daily, 0.833333 -13.2588 41.28 0.833333 -13.2588 (lr) 0.833333 -13.2588 surnames 0.833333 -13.2588 downloading 0.833333 -13.2588 137k 0.833333 -13.2588 (short 0.666667 -13.259 36.6 0.666667 -13.2595 non-projectable 0.833333 -13.2595 679 0.833333 -13.2595 w’s 0.833333 -13.2595 307 0.833333 -13.2595 24.1 0.833333 -13.2595 2250 0.833333 -13.2595 value: 0.833333 -13.2595 ‘actionable’ 0.833333 -13.2595 participle, 0.833333 -13.2595 0.749 0.583333 -13.2597 (dt) 0.583333 -13.2597 leacock, 0.666667 -13.26 suzuki, 0.666667 -13.26 finger 0.833333 -13.2601 stanza, 0.833333 -13.2601 “wrong” 0.833333 -13.2601 rarer 0.833333 -13.2601 linear map 0.833333 -13.2601 sk, 0.833333 -13.2601 soliciting 0.45 -13.2606 adjudication 0.833333 -13.2607 length-based 0.833333 -13.2607 1.6% 0.833333 -13.2607 o(t) 0.833333 -13.2607 356 0.833333 -13.2607 10−4 0.833333 -13.2607 (true) 0.833333 -13.2607 kong, 0.666667 -13.261 65.8 0.21 -13.261 another. 0.533333 -13.2612 evidence about 0.434783 -13.2612 can exploit 0.833333 -13.2613 23.1 0.833333 -13.2613 scan with 0.833333 -13.2613 lately 0.833333 -13.2613 -n 0.833333 -13.2613 thresholded 0.833333 -13.2613 high-scoring 0.666667 -13.2615 1 -13.3673 online at: 1 -13.3673 merge. 1 -13.3673 statistical, 1 -13.3673 scc. 1 -13.3673 (§3.1). 1 -13.3673 wallis, 1 -13.3673 adam meyers. 1 -13.3673 (fundel 1 -13.3673 kobus 1 -13.3673 ts.t. 1 -13.3673 person) 1 -13.3673 composites 1 -13.3673 f1). 1 -13.3673 early-update 1 -13.3673 [3]. 1 -13.3673 89.0% 1 -13.3673 ads. 1 -13.3673 composes 1 -13.3673 where { geo:new 1 -13.3673 allen. 1 -13.3673 ongoing. 1 -13.3673 capacity. 1 -13.3673 inequality. 1 -13.3673 iht. 1 -13.3673 spanish). 1 -13.3673 0.71. 1 -13.3673 boek 1 -13.3673 jagadeesh jagarlamudi 1 -13.3673 prep. 1 -13.3673 traversals 1 -13.3673 partialromance 1 -13.3673 expert swsd 1 -13.3673 ryan mcdonald, 1 -13.3673 risk. 1 -13.3673 continue. 1 -13.3673 1.17 1 -13.3673 cars grew 1 -13.3673 fs. 1 -13.3673 zodiac-408 cipher. 1 -13.3673 quotation marks 1 -13.3673 wud 1 -13.3673 model4. 1 -13.3673 jst. 1 -13.3673 segment). 1 -13.3673 a1ev, a2ev, 1 -13.3673 (bilingual) 1 -13.3673 6note 1 -13.3673 resnik (1999) 1 -13.3673 languages”. 1 -13.3673 uncertain. 1 -13.3673 sentiment: 1 -13.3673 sibling. 1 -13.3673 minkov 1 -13.3673 thai, 1 -13.3673 mcc courses 1 -13.3673 192). 1 -13.3673 descendants), 1 -13.3673 cathedral. 1 -13.3673 (auc). 1 -13.3673 (cho, 1 -13.3673 claim. 1 -13.3673 usd. 1 -13.3673 .",' 1 -13.3673 peperkamp 1 -13.3673 1030 1 -13.3673 predict, 1 -13.3673 picture? 1 -13.3673 b.p. 1 -13.3673 interfaces. 1 -13.3673 (ratnaparkhi 1 -13.3673 searches. 1 -13.3673 quotient algebras 1 -13.3673 (k). 1 -13.3673 relaxed. 1 -13.3673 textrunner. 1 -13.3673 lawyer 1 -13.3673 similaritybased 1 -13.3673 /h. 1 -13.3673 non-directional easy-first 1 -13.3673 c[i:e] 1 -13.3673 concordancing 1 -13.3673 e.g, 1 -13.3673 iwsd. 1 -13.3673 promise. 1 -13.3673 wonderful ideas. 1 -13.3673 out-perform 1 -13.3673 fairy tales, 1 -13.3673 <.01 1pre-test 1 -13.3673 φd 1 -13.3673 (haffari 1 -13.3673 systematic, 1 -13.3673 politeness, 1 -13.3673 country’s 1 -13.3673 utilises 1 -13.3673 dist sen 1 -13.3673 time progresses, 1 -13.3673 sourced 1 -13.3673 validated. 1 -13.3673 mih 1 -13.3673 traversal. 1 -13.3673 dissimilarity. 1 -13.3673 1-24. 1 -13.3673 outnumbering 1 -13.3673 gradability 1 -13.3673 coders, 1 -13.3673 narayanan. 1 -13.3673 majumdar 1 -13.3673 non-understanding, 1 -13.3673 unaccented roman 1 -13.3673 problemin 1 -13.3673 langmead, 1 -13.3673 functionality, 1 -13.3673 editor. 1 -13.3673 fallback 1 -13.3673 maturity. 1 -13.3673 filt. 1 -13.3673 rentoumi 1 -13.3673 out-of-domain, 1 -13.3673 grac¸a 1 -13.3673 bodrumlu 1 -13.3673 product-moment 1 -13.3673 .]〉 1 -13.3673 afterwards. 1 -13.3673 93.00 1 -13.3673 boundary) 1 -13.3673 francisco casacuberta 1 -13.3673 dligach 1 -13.3673 music player. 1 -13.3673 outlinks 1 -13.3673 eprime countt(f,eprime) 1 -13.3673 scholars, 1 -13.3673 survey). 1 -13.3673 68 32 1 -13.3673 right-attach 1 -13.3673 pa, 15260 1 -13.3673 ˆy. 1 -13.3673 cluster; 1 -13.3673 “better” 1 -13.3673 barto, 1 -13.3673 2003b; 1 -13.3673 std. 1 -13.3673 frame’s 1 -13.3673 lvc. 1 -13.3673 〈num〉sg 1 -13.3673 recalculated. 1 -13.3673 addedtothe 1 -13.3673 szarvas, 1 -13.3673 benchmark, 1 -13.3673 well-performing 1 -13.3673 (2006a) 1 -13.3673 verbobject 1 -13.3673 purver 1 -13.3673 [gillenwater 1 -13.3673 prioritizing 1 -13.3673 tokenizer, 1 -13.3673 meteor’s 1 -13.3673 (reynar 1 -13.3673 tfst 1 -13.3673 1966), 1 -13.3673 tagged, 1 -13.3673 surround 1 -13.3673 tsukahara, 1 -13.3673 way? 1 -13.3673 architectures. 1 -13.3673 in proceedings of acl2010. 1 -13.3673 o(n2). 1 -13.3673 (x,y). 1 -13.3673 fire. 1 -13.3673 97.3% 1 -13.3673 thelwall 1 -13.3673 substring, 1 -13.3673 web). 1 -13.3673 1–11, 1 -13.3673 (jurish, 1 -13.3673 bitext, 1 -13.3673 converged. 1 -13.3673 39.86 1 -13.3673 cancellation. 1 -13.3673 letter-by-letter 1 -13.3673 mulogo 1 -13.3673 nseed 1 -13.3673 progresses. 1 -13.3673 collaboration-based 1 -13.3673 released. 1 -13.3673 phase). 1 -13.3673 approach). 1 -13.3673 compactness. 1 -13.3673 terry koo, 1 -13.3673 bloodgood 1 -13.3673 by. 1 -13.3673 ’m 1 -13.3673 execution. 1 -13.3673 (sds) 1 -13.3673 per-category 1 -13.3673 house’. 1 -13.3673 58.2 30.5 1 -13.3673 2-5 1 -13.3673 (remote) 1 -13.3673 - loria 1 -13.3673 knight@isi.edu 1 -13.3673 continuation. 1 -13.3673 se. 1 -13.3673 schuller 1 -13.3673 ments, 1 -13.3673 vowel harmony, 1 -13.3673   [request‐ 1 -13.3673 (4.08%) 1 -13.3673 pre-calculated 1 -13.3673 descendants. 1 -13.3673 (metaweb 1 -13.3673 (nullnullnullnullnullnullnullnull (nullnullnullnull 1 -13.3673 dk(q). 1 -13.3673 diagonal. 1 -13.3673 standardised 1 -13.3673 factoid. 1 -13.3673 (hmm). 1 -13.3673 compressor, 1 -13.3673 6.2, 1 -13.3673 webpages. 1 -13.3673 seed. 1 -13.3673 harvesting, 1 -13.3673 fan-out. 1 -13.3673 debes pagar 1 -13.3673 decontamination 1 -13.3673 2008]. 1 -13.3673 (middle) 1 -13.3673 54.91 1 -13.3673 repeats. 1 -13.3673 (optional). 1 -13.3673 generations, 1 -13.3673 ptk. 1 -13.3673 indef. 1 -13.3673 npov, 1 -13.3673 coincide. 1 -13.3673 cultures, 1 -13.3673 i−. 1 -13.3673 lattices) 1 -13.3673 plaintext. 1 -13.3673 mcdonald, crammer, 1 -13.3673 hiberno-english. 1 -13.3673 meth, 1 -13.3673 ci,j, pi,j, 1 -13.3673 intelligible. 1 -13.3673 initializations. 1 -13.3673 8,200 1 -13.3673 integer programming 1 -13.3673 bigram) 1 -13.3673 hapaxes, 1 -13.3673 knowledge), 1 -13.3673 ball. 1 -13.3673 (bohus, 1 -13.3673 c++. 1 -13.3673 investigate. 1 -13.3673 skywlkr 1 -13.3673 gsc, 1 -13.3673 higher-quality 1 -13.3673 scale). 1 -13.3673 ˆpmle. 1 -13.3673 partials returned 1 -13.3673 leaf. 1 -13.3673 parsim, 1 -13.3673 vickrey 1 -13.3673 rte-5, 1 -13.3673 (system prediction: 1 -13.3673 chin-yew lin. 1 -13.3673 sensitivity. 1 -13.3673 cutoff. 1 -13.3673 24. 1 -13.3673 swap-rate 1 -13.3673 health. 1 -13.3673 stripe. 1 -13.3673 lemmatized. 1 -13.3673 wikipedia). 1 -13.3673 tm. 1 -13.3673 ns si 1 -13.3673 serious, 1 -13.3673 ox1 1 -13.3673 00005 1 -13.3673 smoothing; 1 -13.3673 synonyms) 1 -13.3673 ideal. 1 -13.3673 non-transliterations. 1 -13.3673 translationese: 1 -13.3673 30–38, 1 -13.3673 offsets. 1 -13.3673 sets). 1 -13.3673 real-time. 1 -13.3673 hashimoto 1 -13.3673 3.5. 1 -13.3673 /) () 1 -13.3673 1a). 1 -13.3673 strategy). 1 -13.3673 relationally-clustered svs 1 -13.3673 tones 1 -13.3673 nps) 1 -13.3673 nomenclature. 1 -13.3673 6×104 ue 1 -13.3673 different-genre 1 -13.3673 204). 1 -13.3673 undermines 1 -13.3673 werden. 1 -13.3673 proceedings. 1 -13.3673 confident. 1 -13.3673 cprime(ng) 1 -13.3673 duration: 1 -13.3673 60-iteration 1 -13.3673 .0001). 1 -13.3673 parks 1 -13.3673 methodologies. 1 -13.3673 serena. 1 -13.3673 100). 1 -13.3673 (rae) 1 -13.3673 null@null 1 -13.3673 emnlp. 1 -13.3673 41.62 1 -13.3673 degrades. 1 -13.3673 stopped. 1 -13.3673 poss. 1 -13.3673 modularized 1 -13.3673 & roth’s 1 -13.3673 wmt overview 1 -13.3673 possibility. 1 -13.3673 tourist resort 1 -13.3673 encyclopedia. 1 -13.3673 fiths 1 -13.3673 ...”. 1 -13.3673 rubin, 1 -13.3673 volunteers. 1 -13.3673 plateaus 1 -13.3673 ever had. 1 -13.3673 survive 1 -13.3673 jack mostow 1 -13.3673 modules; 1 -13.3673 indexed. 1 -13.3673 signals. 1 -13.3673 phonotactics 1 -13.3673 postprocess 1 -13.3673 treebank-trained 1 -13.3673 adj. 1 -13.3673 answers). 1 -13.3673 vandenberghe, 1 -13.3673 links). 1 -13.3673 example; 1 -13.3673 nomoto, 1 -13.3673 pers. 1 -13.3673 vmm, 1 -13.3673 histories. 1 -13.3673 sdss. 1 -13.3673 communication). 1 -13.3673 cimec, university 1 -13.3673 guarantees. 1 -13.3673 absence) 1 -13.3673 conservative. 1 -13.3673 collect, 1 -13.3673 material). 1 -13.3673 (rogers, 1 -13.3673 trainer. 1 -13.3673 headwords. 1 -13.3673 adequately. 1 -13.3673 (sapena 1 -13.3673 web-mined 1 -13.3673 unsurprisingly 1 -13.3673 bridge. 1 -13.3673 weir, 1 -13.3673 eyebrows. 1 -13.3673 1968). 1 -13.3673 [b,c] a∼[β,γ] 1 -13.3673 error-annotated 1 -13.3673 0005 1 -13.3673 headdependent 1 -13.3673 88.72 1 -13.3673 inserted? 1 -13.3673 termination. 1 -13.3673 calculations, 1 -13.3673 2b. 1 -13.3673 cross-learning 1 -13.3673 favourability. 1 -13.3673 ubiu’s 1 -13.3673 python. 1 -13.3673 annotate. 1 -13.3673 1974). 1 -13.3673 equivalents, 1 -13.3673 debugging 1 -13.3673 clip. 1 -13.3673 (advp (rb 1 -13.3673 (spoustov´a 1 -13.3673 rectangle 1 -13.3673 avoid. 1 -13.3673 3qd, 1 -13.3673 language.” 1 -13.3673 asymmetric, 1 -13.3673 “bad”. 1 -13.3673 riley, 1 -13.3673 τs. 1 -13.3673 perceptibly 1 -13.3673 metonymy. 1 -13.3673 mc. 1 -13.3673 program, darpa/cmo 1 -13.3673 endings. 1 -13.3673 librarian. 1 -13.3673 other? 1 -13.3673 annotators). 1 -13.3673 curriculum 1 -13.3673 thesis. 1 -13.3673 am-fm, 1 -13.3673 patterns” 1 -13.3673 emnlp’08. 1 -13.3673 anthology. 1 -13.3673 rblms 1 -13.3673 tightly coupled 1 -13.3673 peaked 1 -13.3673 deny 1 -13.3673 instructors. 1 -13.3673 l1(s) +λrq(s) 1 -13.3673 95%, 1 -13.3673 ’07, 1 -13.3673 ferent 1 -13.3673 political violence 1 -13.3673 scaling. 1 -13.3673 0.81. 1 -13.3673 word-frequency 1 -13.3673 f-measure; 1 -13.3673 sempos. 1 -13.3673 65%. 1 -13.3673 demographics. 1 -13.3673 authorship attribution. 1 -13.3673 literature; 1 -13.3673 figuring 1 -13.3673 soon. 1 -13.3673 (e4). 1 -13.3673 graph gd(s) 1 -13.3673 29.74 1 -13.3673 s.append( 1 -13.3673 non-experts. 1 -13.3673 ease. 1 -13.3673 louis. 1 -13.3673 circles. 1 -13.3673 (cahill 1 -13.3673 relatively narrow 1 -13.3673 expansion) 1 -13.3673 function). 1 -13.3673 essen ‘that 1 -13.3673 competency 1 -13.3673 russia. 1 -13.3673 human beings. 1 -13.3673 cn-lvcs. 1 -13.3673 gigaword, 1 -13.3673 prediction-based 1 -13.3673 gd(s). 1 -13.3673 sentiment). 1 -13.3673 query-click 1 -13.3673 nonunderstanding 1 -13.3673 message-level 1 -13.3673 blur 1 -13.3673 mouse genome 1 -13.3673 explain. 1 -13.3673 documents’ 1 -13.3673 chart[b, e] 1 -13.3673 v). 1 -13.3673 mihalcea. 1 -13.3673 long-span 1 -13.3673 condition). 1 -13.3673 regression (svr) 1 -13.3673 informality. 1 -13.3673 utt. 1 -13.3673 interested, 1 -13.3673 proces. 1 -13.3673 .cm/d 1 -13.3673 delicious. 1 -13.3673 (3.6) 1 -13.3673 afrl 1 -13.3673 collect. 1 -13.3673 adam lopez 1 -13.3673 (monolingual 1 -13.3673 designs, 1 -13.3673 unanchored 1 -13.3673 j3 1 -13.3673 kybots. 1 -13.3673 ??x 1 -13.3673 boost. 1 -13.3673 34508 1 -13.3673 article; 1 -13.3673 (gap+) [first-pron] 1 -13.3673 (mart´ınez 1 -13.3673 84.21% 1 -13.3673 window). 1 -13.3673 tain 1 -13.3673 realistic. 1 -13.3673 recognize. 1 -13.3673 combi. 1 -13.3673 realisations. 1 -13.3673 complementation 1 -13.3673 (dhillon 1 -13.3673 (3k) 1 -13.3673 system.2 1 -13.3673 shorter. 1 -13.3673 29.78 1 -13.3673 hull, 1 -13.3673 squiggleright [b1,c1]...[bk,ck] 1 -13.3673 1957). 1 -13.3673 sep. 1 -13.3673 systematically. 1 -13.3673 fine-tune 1 -13.3673 anchor. 1 -13.3673 2b). 1 -13.3673 polaritybearing 1 -13.3673 lm). 1 -13.3673 ia. 1 -13.3673 anm fod 1 -13.3673 0.05), 1 -13.3673 halves, 1 -13.3673 experiments.) 1 -13.3673 soricut 1 -13.3673 system? 1 -13.3673 hirschberg’s 1 -13.3673 yu. 1 -13.3673 meta-evaluate 1 -13.3673 maintain. 1 -13.3673 (graehl 1 -13.3673 win: 1 -13.3673 ece. 1 -13.3673 ir. 1 -13.3673 subclause 1 -13.3673 env. 1 -13.3673 envisage 1 -13.3673 differentiable. 1 -13.3673 item, ι, 1 -13.3673 lsi-context … 1 -13.3673 (gamon, 1 -13.3673 nonstop 1 -13.3673 (mani, 1 -13.3673 (sp¨arck 1 -13.3673 mc rb 1 -13.3673 component). 1 -13.3673 ji; ;a/ 1 -13.3673 one-to-one. 1 -13.3673 nulli. 1 -13.3673 verb forms: 1 -13.3673 undertook 1 -13.3673 luxburg, 1 -13.3673 modularization. 1 -13.3673 corporation. 1 -13.3673 axis. 1 -13.3673 reply. 1 -13.3673 kingdom, 1 -13.3673 slider 1 -13.3673 itself). 1 -13.3673 (all). 1 -13.3673 pseudo-word. 1 -13.3673 h∈h. 1 -13.3673 w*. 1 -13.3673 materials. 1 -13.3673 leter 1 -13.3673 (none). 1 -13.3673 staged 1 -13.3673 vectorw. 1 -13.3673 57.84 1 -13.3673 exposition. 1 -13.3673 experiences. 1 -13.3673 timegraph. 1 -13.3673 constants. 1 -13.3673 smeaton, 1 -13.3673 z(s) 1 -13.3673 tendency. 1 -13.3673 (finlayson, 1 -13.3673 86.08 1 -13.3673 treetagger. 1 -13.3673 combined system‘s 1 -13.3673 w. chapman, 1 -13.3673 sub-genre. 1 -13.3673 journals, 1 -13.3673 srs 1 -13.3673 reconcile, 1 -13.3673 tikk 1 -13.3673 human-labeled 1 -13.3673 weight). 1 -13.3673 sample) 1 -13.3673 pitchaccent 1 -13.3673 selecting, 1 -13.3673 janota. 1 -13.3673 cited. 0.666667 -13.3674 making. 0.666667 -13.3674 krahmer, 0.666667 -13.3674 clustering: 0.666667 -13.3674 yvon 0.666667 -13.3674 joanis 0.444444 -13.3677 universidad 0.714286 -13.368 amn’t 0.8 -13.368 combinations). 0.8 -13.368 logging 0.8 -13.368 non-name 0.8 -13.368 provisions 0.8 -13.368 “. 0.8 -13.368 (0.6) 0.8 -13.368 emotinet. 0.8 -13.368 included). 0.8 -13.368 (brown, 0.8 -13.368 (torii 0.8 -13.368 evaluation) 0.8 -13.368 3-4 0.8 -13.368 recent. 0.8 -13.368 events), 0.8 -13.368 fewer. 0.8 -13.368 fusion, 0.8 -13.368 (80.3%) 0.8 -13.368 committee. 0.8 -13.368 terms). 0.8 -13.368 inference-based 0.8 -13.368 ir group 0.8 -13.368 linear. 0.8 -13.368 deterministic. 0.8 -13.368 85%. 0.8 -13.368 .c/d 0.8 -13.368 overlooked. 0.8 -13.368 singular. 0.8 -13.368 expected-reconstruction-error 0.8 -13.368 appr. 0.8 -13.368 intensity. 0.8 -13.368 bible. 0.8 -13.368 lacoste-julien 0.8 -13.368 connotation. 0.8 -13.368 samuel, 0.8 -13.368 (joachims, 1999), 0.8 -13.368 interpersonal communication. 0.8 -13.368 int. 0.8 -13.368 formalise 0.8 -13.368 +mst 0.8 -13.368 factor). 0.8 -13.368 dk, 0.8 -13.368 weighs 0.8 -13.368 ˆc. 0.8 -13.368 surveys. 0.8 -13.368 dump. 0.8 -13.368 recover. 0.8 -13.368 (§4) 0.8 -13.368 owl. 0.8 -13.368 optimizer. 0.8 -13.368 majority. 0.8 -13.368 continuous scales 0.8 -13.368 204) 0.8 -13.368 co-referenced. 0.8 -13.368 syllable. 0.8 -13.368 (herbrich 0.8 -13.368 wordnetbased 0.8 -13.368 magnitudes 0.8 -13.368 introduction. 0.8 -13.368 1...m 0.8 -13.368 m). 0.8 -13.368 67.00 0.8 -13.368 w). 0.8 -13.368 compressors 0.8 -13.368 inquiry points. 0.8 -13.368 person), 0.8 -13.368 81.0% 0.8 -13.368 germane 0.8 -13.368 substrings, 0.8 -13.368 nlu, 0.8 -13.368 election. 0.8 -13.368 (2006)). 0.8 -13.368 0.90. 0.8 -13.368 abilities. 0.8 -13.368 participated. 0.8 -13.368 humorous 0.8 -13.368 discovered brands 0.8 -13.368 shriberg, 0.8 -13.368 40.61 0.8 -13.368 concept). 0.8 -13.368 surveyed 0.8 -13.368 10–19, 0.8 -13.368 yamangil 0.8 -13.368 1spd. 0.8 -13.368 1,000. 0.8 -13.368 tier. 0.8 -13.368 event-sentiment 0.8 -13.368 7%. 0.8 -13.368 asked workers 0.8 -13.368 middle. 0.8 -13.368 wülfing, 0.8 -13.368 .967 0.8 -13.368 information), 0.8 -13.368 nonpronominal 0.8 -13.368 matchings. 0.8 -13.368 .67 0.8 -13.368 shop. 0.8 -13.368 low). 0.8 -13.368 gene nomenclature 0.8 -13.368 (§2). 0.8 -13.368 biased. 0.8 -13.368 era. 0.8 -13.368 (thelen 0.8 -13.368 fhmms 0.8 -13.368 lse. 0.8 -13.368 tighter integration 0.5 -13.3681 5a 0.5 -13.3685 producttext 0.5 -13.3685 and martin 0.714286 -13.3686 practically, 0.714286 -13.3686 briefly described 0.714286 -13.3686 smart(x) 0.714286 -13.3686 ˆd 0.714286 -13.3686 what degree 0.8 -13.3686 jaky jaok 0.8 -13.3686 salakoski. 0.8 -13.3686 guided bootstrapping 0.8 -13.3686 gender attribution, 0.8 -13.3686 sems 0.8 -13.3686 baldwin, 1998), 0.8 -13.3686 l-bfgs. 0.8 -13.3686 2005a). 0.8 -13.3686 kn06, 0.8 -13.3686 (martineau 0.8 -13.3686 predictor β 0.8 -13.3686 set.1 0.8 -13.3686 unigrams oov 0.8 -13.3686 quit 0.8 -13.3686 intended. 0.8 -13.3686 latin. 0.8 -13.3686 acquisition bottleneck 0.8 -13.3686 semantic orientation. 0.8 -13.3686 formulae, 0.8 -13.3686 contrastive submission 0.8 -13.3686 parse tree; 0.8 -13.3686 log. 0.8 -13.3686 intrathoracic 0.5 -13.369 might lead 0.714286 -13.3691 (χ2 0.226667 -13.3693 k-nn 0.8 -13.3693 time being, 0.8 -13.3693 problems; 0.8 -13.3693 forty paradigm 0.8 -13.3693 90.4% 0.8 -13.3693 sequential minimal 0.8 -13.3693 rooth 0.8 -13.3693 outgoing arcs 0.8 -13.3693 pathologists 0.8 -13.3693 george a. 0.226667 -13.3696 zhang et 0.8 -13.37 qt2 0.8 -13.37 anaphoric references 0.8 -13.37 hypothesis “α 0.8 -13.37 0.797 0.8 -13.37 90.2% 0.8 -13.37 by human) 0.8 -13.37 svm-mc 0.8 -13.37 proven successful 0.8 -13.37 joined together 0.8 -13.37 “attack” 0.8 -13.37 osborne (2009) 0.8 -13.37 online communities. 0.8 -13.37 ndcg scores 0.8 -13.3706 morphbleu 0.8 -13.3706 cumulative gain 0.8 -13.3706 medical record 0.8 -13.3706 abbasi 0.8 -13.3706 64, 0.315789 -13.3713 precision-recall 0.8 -13.3713 fable 0.8 -13.3713 rounds, 0.8 -13.3713 incorporating syntax 0.8 -13.3713 p. & 0.8 -13.3713 head-finding rules 0.8 -13.3713 ancestor chain 0.8 -13.3713 fully inflected 0.254237 -13.3716 our framework 0.235294 -13.3716 this results in 0.6 -13.3718 average rank 0.8 -13.372 neighbour analysis 0.8 -13.372 splitting compounds 0.8 -13.372 (section 6), 0.8 -13.372 centroid-based 0.8 -13.372 factors affecting 0.714286 -13.3721 (fp7/2007-2013) 0.6 -13.3723 realized with 0.6 -13.3723 several interesting 0.8 -13.3726 three bipolar 0.8 -13.3726 emnlp, pages 0.8 -13.3726 80.5% 0.8 -13.3726 progress, 0.8 -13.3726 inserted, 0.8 -13.3726 bubs 0.8 -13.3726 regularized joint 0.8 -13.3726 positive integer 0.8 -13.3726 (magerman, 0.8 -13.3726 list1 0.166667 -13.3727 opposed to 0.714286 -13.3727 shell 0.714286 -13.3727 akan 0.714286 -13.3733 (0, 0.8 -13.3733 reported, 0.8 -13.3733 feminine, 0.8 -13.3733 0.442 0.8 -13.3733 (language 0.8 -13.3733 preprocessor 0.8 -13.3733 grishman (2008) 0.714286 -13.3739 µj 0.714286 -13.3739 sugests 0.714286 -13.3739 bracehtipuprightbracehtipupleft 0.302326 -13.374 n best 0.8 -13.374 graph-based optimization 0.8 -13.374 let’s consider 0.8 -13.374 almost perfect 0.6 -13.3743 afective 0.6 -13.3743 centering 0.8 -13.3746 t1 t2 0.8 -13.3746 peripheral element 0.8 -13.3746 (µ = 0.8 -13.3746 fluctuates 0.354839 -13.3747 marsi, 0.295455 -13.3748 occur at 0.8 -13.3753 after filtering, 0.8 -13.3753 0.530 0.8 -13.3753 entrain 0.8 -13.3753 motion event 0.216867 -13.3754 tokens are 0.254237 -13.3756 based on these 0.714286 -13.3757 largest improvement 0.444444 -13.3759 s-patterns 0.324324 -13.376 tool. 0.8 -13.376 786 0.8 -13.376 (glickman 0.8 -13.376 diet 0.8 -13.376 mood, 0.8 -13.376 similarity measurements 0.8 -13.376 85.52 0.216867 -13.3761 this experiment, 0.714286 -13.3763 recurrent neural 0.714286 -13.3763 (tsochantaridis et 0.444444 -13.3766 utility. 0.204301 -13.3767 they may 0.8 -13.3767 non-european 0.8 -13.3767 (bies 0.8 -13.3773 0.011 0.8 -13.3773 mukherjee 0.8 -13.3773 frame; 0.8 -13.3773 (hirst 0.8 -13.3773 comp.arg2 0.6 -13.3774 epidemics 0.139303 -13.3778 depend 0.6 -13.3779 several languages 0.8 -13.378 (genzel, 0.8 -13.378 newstest2011 0.8 -13.378 chalenges 0.714286 -13.3781 5m 0.714286 -13.3781 correspondingto 0.5 -13.3786 q(s,a) 0.8 -13.3787 halting 0.8 -13.3787 available: 0.8 -13.3787 (halliday 0.8 -13.3787 79.2% 0.8 -13.3787 tree-to-string rules. 0.8 -13.3787 gfqg 0.8 -13.3787 an lscript2 0.8 -13.3787 single-task 0.8 -13.3787 dictionary lookup. 0.6 -13.3789 =− 0.714286 -13.3793 decipherment method 0.714286 -13.3793 0.059 0.714286 -13.3793 table: 0.8 -13.3794 occasions 0.8 -13.3794 33.9 0.8 -13.3794 abdul 0.538462 -13.3797 ., 0.6 -13.3799 korhonen 0.8 -13.38 twitter sentiment classification 0.8 -13.38 meurers, 0.8 -13.38 market-oriented 0.8 -13.38 (belz 0.8 -13.38 bi-text, 0.8 -13.38 automatic summarizers, 0.8 -13.38 longest matching 0.6 -13.3804 interprets 0.6 -13.3804 mst+malt 0.714286 -13.3805 fixed effect 0.538462 -13.3807 some errors 0.8 -13.3807 hiero-style systems 0.8 -13.3807 source dictionaries. 0.714286 -13.3811 labour 0.714286 -13.3811 (initial 0.714286 -13.3811 adnominal 0.714286 -13.3811 oe 0.8 -13.3814 27.42 0.8 -13.3814 mwc 0.8 -13.3814 formed, 0.8 -13.3814 (km) 0.8 -13.3814 predominant sense 0.8 -13.3814 descriptions: 0.8 -13.3814 0.626 0.538462 -13.3816 ser 0.714286 -13.3817 21: 0.409091 -13.3819 evaluation measure 0.6 -13.3819 carroll 0.8 -13.3821 (scha, 0.8 -13.3821 stk, 0.8 -13.3821 (data not 0.8 -13.3821 (garera 0.8 -13.3821 (similar) 0.8 -13.3821 (kouylekov 0.8 -13.3821 (engelbrecht 0.8 -13.3821 arcs; 0.8 -13.3821 kow, 0.8 -13.3821 46.2 0.8 -13.3821 without retraining 0.409091 -13.3822 (2008) proposed 0.470588 -13.3824 lower precision 0.142105 -13.3827 function. 0.8 -13.3827 realizations, 0.8 -13.3827 0.736 0.8 -13.3827 lambda calculus. 0.8 -13.3827 becomes extremely 0.8 -13.3827 highest gain 0.8 -13.3827 61.55 0.8 -13.3827 (geman 0.8 -13.3827 explorative 0.8 -13.3827 456 0.8 -13.3827 0.383 0.8 -13.3827 corston-oliver 0.8 -13.3827 globally trained 0.37037 -13.3829 λ2 0.714286 -13.383 eulas 0.714286 -13.383 must belong 0.333333 -13.383 (in this 0.6 -13.3834 trofi 0.444444 -13.3834 xj 0.8 -13.3834 15.3 0.8 -13.3834 (nakov, 0.8 -13.3834 2.11 0.8 -13.3834 ment, 0.8 -13.3834 grand-sibling 0.8 -13.3834 (bamman 0.264151 -13.3835 enrich 0.714286 -13.3836 e∗ 0.714286 -13.3836 throughout, 0.470588 -13.384 wpp 0.8 -13.3841 (dtk) 0.8 -13.3841 similar pronunciations 0.8 -13.3841 genetag 0.8 -13.3841 tp, 0.8 -13.3841 35.8 0.8 -13.3841 bunt, 0.8 -13.3841 surpassing 0.8 -13.3841 horvitz, 0.6 -13.3845 zhao-eng 0.0883534 -13.3846 expressed 0.28 -13.3847 baroni 0.714286 -13.3848 “and 0.714286 -13.3848 periodically 0.384615 -13.3848 lbj 0.8 -13.3848 shi, 0.8 -13.3848 0.067 0.8 -13.3848 human-readable 0.8 -13.3848 (allan 0.8 -13.3848 genitive, 0.8 -13.3848 r/o 0.8 -13.3848 selection; 0.8 -13.3848 35.1 0.8 -13.3848 self-labeled 0.428571 -13.385 adding new 0.264151 -13.3853 available from 0.714286 -13.3854 bracehtipdownright 0.714286 -13.3854 o[the] 0.8 -13.3855 s6 0.8 -13.3855 2007)) 0.8 -13.3855 training exercises, 0.428571 -13.3857 it outperforms 0.714286 -13.386 0.912 0.206522 -13.386 entries in 0.8 -13.3861 89.5% 0.8 -13.3861 (earley, 0.8 -13.3861 year” 0.8 -13.3861 hj 0.8 -13.3861 treebank (xue 0.8 -13.3861 18.0 0.6 -13.3865 svr-rec 0.714286 -13.3866 several strategies 0.714286 -13.3866 hair 0.8 -13.3868 19.5 0.8 -13.3868 coma 0.8 -13.3868 albias 0.8 -13.3868 (dale 0.8 -13.3868 8in 0.8 -13.3868 level; 0.8 -13.3868 481 0.8 -13.3868 publicly-available 0.8 -13.3868 emami 0.714286 -13.3872 user utterance. 0.214286 -13.3873 even with 0.8 -13.3875 mainichi 0.8 -13.3875 monods, 0.8 -13.3875 unavoidably 0.8 -13.3875 probe 0.8 -13.3875 nullnull1 0.5 -13.3879 gone 0.0439506 -13.3881 the best 0.8 -13.3882 segc 0.8 -13.3882 probability; 0.8 -13.3882 sumo 0.8 -13.3882 0.824 0.8 -13.3882 onr grant 0.8 -13.3882 simulator 0.8 -13.3882 (chieu 0.714286 -13.3884 =∅ 0.470588 -13.3888 hello 0.5 -13.3888 duluth-1 0.8 -13.3889 afforded 0.8 -13.3889 41.8 0.8 -13.3889 considerably, 0.8 -13.3889 0.904 0.8 -13.3889 clusteringbased 0.8 -13.3889 0.539 0.8 -13.3889 (g), 0.8 -13.3889 combine, 0.8 -13.3889 〈a,n〉 0.8 -13.3889 0.453 0.714286 -13.389 “at 0.714286 -13.389 “was 0.258621 -13.3891 correctly. 0.5 -13.3892 may introduce 0.538462 -13.3893 indiana 0.8 -13.3895 artist, 0.8 -13.3895 meantime, 0.8 -13.3895 benefit, 0.8 -13.3895 lated 0.714286 -13.3896 λ-production 0.714286 -13.3896 394 0.714286 -13.3896 multext 0.357143 -13.3896 comparative sentences 0.6 -13.3896 homophones, 0.6 -13.3896 several aspects 0.150602 -13.3899 comparison, 0.714286 -13.3902 references: 0.8 -13.3902 senders 0.8 -13.3902 0.731 0.8 -13.3902 sense-inventory 0.8 -13.3902 1.65 0.8 -13.3902 liwc, 0.8 -13.3902 duality 0.8 -13.3902 apparatus 0.444444 -13.3905 f-scores of 0.714286 -13.3908 presumably, 0.8 -13.3909 laid 0.8 -13.3909 re-estimated 0.8 -13.3909 pos-wise 0.8 -13.3909 np’s 0.8 -13.3909 γ.vi 0.8 -13.3909 “target” 0.8 -13.3909 content-word 0.8 -13.3909 treebased 0.8 -13.3909 coalescence 0.8 -13.3909 promising direction 0.8 -13.3909 colective 0.6 -13.3911 nl+nlp 0.0954654 -13.3913 tasks. 0.195876 -13.3915 6th 0.714286 -13.3915 beg 0.714286 -13.3915 you please 0.8 -13.3916 precompute 0.8 -13.3916 2008; thater 0.8 -13.3916 interdisciplinary 0.8 -13.3916 specificity, 0.8 -13.3916 our crowdsourced data 0.8 -13.3916 6-gram lm 0.8 -13.3916 dissimilarity, 0.714286 -13.3921 jproductdisplay 0.6 -13.3921 kbs 0.8 -13.3923 iu 0.8 -13.3923 friends, 0.8 -13.3923 folows: 0.8 -13.3923 jindigo 0.8 -13.3923 laundry 0.8 -13.3923 sibling, 0.8 -13.3923 49.8 0.8 -13.3923 0.810 0.714286 -13.3927 aus 0.8 -13.3929 rtl 0.8 -13.3929 ameliorate 0.8 -13.3929 lifted 0.8 -13.3929 sprimei 0.8 -13.3929 yy 0.8 -13.3929 +4 0.8 -13.3929 (wmt 0.8 -13.3929 posibly 0.8 -13.3929 matching: 0.8 -13.3929 sk+1 0.714286 -13.3933 java, 0.538462 -13.3935 verbs, adjectives 0.8 -13.3936 confounders 0.8 -13.3936 lay 0.8 -13.3936 2011b; 0.8 -13.3936 appreciated 0.8 -13.3936 (mst) 0.8 -13.3936 30.2% 0.8 -13.3936 em: 0.8 -13.3936 m2m-aligner 0.8 -13.3936 professions 0.8 -13.3936 nrc emotion 0.8 -13.3936 92.44 0.8 -13.3936 73.6% 0.714286 -13.3939 palestinians 0.295455 -13.3939 between pairs 0.6 -13.3942 commitments 0.8 -13.3943 0.386 0.8 -13.3943 69.2% 0.8 -13.3943 action pops 0.8 -13.3943 monologic 0.8 -13.3943 nc-test2008 0.8 -13.3943 synonymous, 0.8 -13.3943 engine’s 0.8 -13.3943 lucene, 0.8 -13.3943 inversions 0.538462 -13.3944 duluth-3 0.714286 -13.3945 453 0.15528 -13.3946 they do 0.8 -13.395 personal, 0.8 -13.395 dvds, 0.8 -13.395 (dictionary) 0.8 -13.395 3813 0.8 -13.395 2.3) 0.8 -13.395 fahrr¨ader” 0.8 -13.395 sen. 0.8 -13.395 varma, 0.714286 -13.3951 d(p,q) 0.5 -13.3952 5000 forbes 0.538462 -13.3953 argument classification 0.202128 -13.3957 and lee, 0.714286 -13.3957 bacteria, 0.6 -13.3957 ling 0.6 -13.3957 95.0 0.8 -13.3957 backchannel, 0.8 -13.3957 17.8 0.8 -13.3957 post), 0.8 -13.3957 284 0.8 -13.3957 differentiable, 0.8 -13.3957 vocabulary-based 0.8 -13.3957 then retrained 0.8 -13.3957 embody 0.8 -13.3957 time-consuming, 0.8 -13.3957 0.579 0.333333 -13.3963 is essential 0.8 -13.3964 0.465 0.8 -13.3964 labor-intensive 0.8 -13.3964 θna 0.8 -13.3964 st-onge, 0.8 -13.3964 43.01 0.8 -13.3964 flurry 0.8 -13.3964 nodules 0.8 -13.3964 zeroanaphora 0.8 -13.3964 expanded, 0.8 -13.3964 riding 0.8 -13.3964 historic 0.8 -13.3964 zettlemoyer, 0.8 -13.3964 ι 0.8 -13.3964 that), 0.176 -13.3966 concerned 0.714286 -13.397 significant: 0.139175 -13.3971 decided 0.8 -13.3971 (o’connor 0.8 -13.3971 fluency-oriented 0.8 -13.3971 more importantly 0.8 -13.3971 object) 0.8 -13.3971 leveled 0.8 -13.3971 tf-idf weighting 0.8 -13.3971 34.57 0.8 -13.3971 mcnemar 0.8 -13.3971 1p, 0.384615 -13.3973 riedel et 0.6 -13.3973 circleplusdisplay 0.8 -13.3977 95.57 0.8 -13.3977 present study. 0.8 -13.3977 inspiring 0.8 -13.3977 .05, 0.8 -13.3977 ghemawat, 0.8 -13.3977 landeghem 0.8 -13.3977 *scope* 0.8 -13.3977 0.691 0.8 -13.3977 procedure; 0.8 -13.3977 size; 0.8 -13.3977 cunit([a1,a2],[b1,b2]) 0.8 -13.3977 variable) 0.8 -13.3977 .8 0.8 -13.3977 counterproductive 0.8 -13.3977 co-occurrences, 0.8 -13.3977 archived 0.139896 -13.3978 2001; 0.204545 -13.3978 to answer 0.176 -13.3981 measure. 0.307692 -13.3984 intuition behind 0.470588 -13.3984 geneva 0.133028 -13.3984 differs 0.176471 -13.3984 will also 0.8 -13.3984 trivial, 0.8 -13.3984 1975) 0.8 -13.3984 matroids 0.8 -13.3984 0.566 0.8 -13.3984 absorbing 0.8 -13.3984 0.436 0.8 -13.3984 ×100 0.8 -13.3984 hara 0.8 -13.3984 interpretability 0.8 -13.3984 function-word 0.8 -13.3984 emerges as 0.8 -13.3984 ℓ 0.8 -13.3984 ei items 0.538462 -13.3985 2007b; 0.714286 -13.3988 comercial 0.8 -13.3991 62.60 0.8 -13.3991 re-used 0.8 -13.3991 enright 0.8 -13.3991 character; 0.8 -13.3991 multi-stream 0.8 -13.3991 eze 0.8 -13.3991 widdows, 0.8 -13.3991 (macleod 0.8 -13.3991 recognising 0.8 -13.3991 awkward 0.8 -13.3991 seddah 0.714286 -13.3994 frame semantics. 0.714286 -13.3994 sur 0.8 -13.3998 cumulative, 0.8 -13.3998 inside, 0.8 -13.3998 44.17 0.8 -13.3998 fate 0.8 -13.3998 -0.35 0.8 -13.3998 morphs 0.8 -13.3998 uk+1 0.8 -13.3998 0.631 0.8 -13.3998 subtopic 0.470588 -13.4 boys 0.714286 -13.4 x∗ 0.714286 -13.4 (¯e,¯a) 0.8 -13.4005 0.854 0.8 -13.4005 clauses) 0.8 -13.4005 2t 0.8 -13.4005 0.969 0.8 -13.4005 (koppel 0.8 -13.4005 marry 0.8 -13.4005 boldface, 0.8 -13.4005 14.2 0.8 -13.4005 (asher 0.8 -13.4005 directive sentences 0.8 -13.4005 robust cross-domain 0.8 -13.4005 cairo, 0.714286 -13.4006 articles published 0.8 -13.4012 no-par 0.8 -13.4012 pakhomov 0.8 -13.4012 (nodes 0.8 -13.4012 sr, 0.8 -13.4012 kenlm 0.8 -13.4012 kaji 0.8 -13.4012 (bisani 0.8 -13.4012 galanis 0.8 -13.4012 0.571 0.8 -13.4012 heavier 0.8 -13.4012 2006a), 0.8 -13.4012 subtree, 0.8 -13.4012 gmcγ 0.8 -13.4012 abstractive compressions 0.8 -13.4012 0.328 0.714286 -13.4013 y”, 0.8 -13.4019 ciprimej = 0.8 -13.4019 reproduction 0.8 -13.4019 parameters; 0.8 -13.4019 masculine, 0.8 -13.4019 (smith, 0.8 -13.4019 terminal isr, 0.8 -13.4019 a1) 0.8 -13.4019 standardize 0.8 -13.4019 66.07 0.8 -13.4019 worsened 0.8 -13.4019 an interlingual representation 0.8 -13.4019 refined, 0.8 -13.4019 bellare 0.8 -13.4019 inaugural 0.121569 -13.4026 is able 0.8 -13.4026 (cca) 0.8 -13.4026 entertainment, 0.8 -13.4026 abdul-rauf 0.8 -13.4026 in-sentence 0.8 -13.4026 fra, 0.8 -13.4026 0.578 0.8 -13.4026 prf 0.8 -13.4026 on un-observed 0.8 -13.4026 buchholz 0.8 -13.4026 diversified 0.8 -13.4026 206) 0.714286 -13.4031 (base 0.384615 -13.4032 indicators for 0.8 -13.4033 tan. 0.8 -13.4033 lima 0.8 -13.4033 acuracy, 0.8 -13.4033 wolfson 0.8 -13.4033 functional tags 0.8 -13.4033 84.15 0.8 -13.4033 master’s 0.8 -13.4033 0.574 0.8 -13.4033 pronounce 0.8 -13.4033 .01, 0.8 -13.4033 lex – 0.714286 -13.4037 interrater 0.428571 -13.4037 michigan 0.211765 -13.4038 chose to 0.444444 -13.4039 goyal 0.8 -13.404 (arora 0.8 -13.404 np-hardness 0.8 -13.404 burn 0.8 -13.404 689 0.8 -13.404 compounds) 0.8 -13.404 (adding 0.8 -13.404 0.874 0.8 -13.404 40.46 0.8 -13.404 aplies 0.8 -13.404 ‘-’ 0.8 -13.404 moses-nogc 0.8 -13.404 vocabulary instruction 0.8 -13.404 remediations 0.8 -13.404 0.501 0.8 -13.404 (7b) 0.8 -13.404 δ, 0.8 -13.404 carrier 0.714286 -13.4043 plugging 0.8 -13.4046 aplication 0.8 -13.4046 (elsner 0.8 -13.4046 (tong 0.8 -13.4046 edit cost 0.8 -13.4046 z¨urich 0.8 -13.4046 latvian, 0.8 -13.4046 birch, 0.8 -13.4046 defeasibly 0.37037 -13.4048 there are multiple 0.714286 -13.405 ranks, 0.5 -13.4051 discriminative features 0.8 -13.4053 2004)) 0.8 -13.4053 diakopoulos 0.8 -13.4053 term-weighting 0.8 -13.4053 fiscal 0.8 -13.4053 entangled 0.8 -13.4053 3’s 0.8 -13.4053 (2006)) 0.8 -13.4053 smaler 0.8 -13.4053 distill 0.8 -13.4053 naming, 0.8 -13.4053 khemphuthe 0.714286 -13.4056 shrinking 0.264151 -13.4056 which contain 0.8 -13.406 38th 0.8 -13.406 givoni 0.8 -13.406 (optionally) 0.8 -13.406 (litkowski 0.8 -13.406 visweswariah 0.8 -13.406 actions) 0.8 -13.406 87.46 0.8 -13.406 navigational 0.8 -13.406 24k 0.8 -13.406 pronouns) 0.8 -13.406 t), 0.8 -13.406 (cover 0.8 -13.406 context-dependent, 0.8 -13.406 jer 0.8 -13.406 sourcing 0.8 -13.406 flatten 0.8 -13.406 contacted 0.8 -13.406 (schuller 0.714286 -13.4062 height-front 0.0759878 -13.4062 estimate 0.5 -13.4064 (6 0.8 -13.4067 null4, 0.8 -13.4067 xtract 0.8 -13.4067 pagerank, 0.8 -13.4067 different perspectives. 0.8 -13.4067 kamm, 0.8 -13.4067 α parameter 0.8 -13.4067 plan, 0.8 -13.4067 61.38 0.8 -13.4067 (popovi´c 0.8 -13.4067 (viterbi) 0.8 -13.4067 founder 0.8 -13.4067 supercomputer 0.8 -13.4067 timetable 0.8 -13.4067 external, 0.714286 -13.4068 contain, 0.714286 -13.4074 (assuming 0.8 -13.4074 spammers 0.8 -13.4074 sprimej 0.8 -13.4074 brockett 0.8 -13.4074 observ 0.8 -13.4074 politécnica 0.8 -13.4074 tree-sequence 0.714286 -13.4081 null(null 0.714286 -13.4081 richer representation 0.8 -13.4081 sacrifices 0.8 -13.4081 highest-weighted 0.8 -13.4081 cherry, 0.8 -13.4081 semantic informativeness 0.8 -13.4081 karttunen, 0.8 -13.4081 upon, 0.8 -13.4081 edits, 0.8 -13.4081 expressions; 0.8 -13.4081 automatically acquire 0.8 -13.4081 zechner, 0.8 -13.4081 vised 0.8 -13.4081 flaw 0.8 -13.4081 96.3 0.8 -13.4081 (2009, 0.8 -13.4081 “tweet” 0.8 -13.4081 sfa 0.8 -13.4081 l2-norm 0.8 -13.4081 both systems were 0.8 -13.4081 0.679 0.714286 -13.4087 [to 0.6 -13.4087 stem+morph 0.8 -13.4088 sociolinguistics, 0.8 -13.4088 hargraves, 0.8 -13.4088 backoff interpolation 0.8 -13.4088 cvandal 0.288889 -13.4091 this problem is 0.391304 -13.4092 republic 0.5 -13.4094 node u 0.8 -13.4095 baeza-yates 0.8 -13.4095 treex 0.8 -13.4095 ical 0.8 -13.4095 (schwarzschild, 0.8 -13.4095 1.69 0.8 -13.4095 nlp tools, 0.8 -13.4095 sub-module 0.8 -13.4095 rejected, 0.8 -13.4095 biologically 0.8 -13.4095 0.247 0.8 -13.4095 refuse 0.8 -13.4095 sym 0.8 -13.4095 np-internal 0.276596 -13.4101 we obtain a 0.8 -13.4102 possibility, 0.8 -13.4102 top: 0.8 -13.4102 eukaryotic 0.8 -13.4102 0.245 0.8 -13.4102 rbmt system 0.8 -13.4102 mfv 0.8 -13.4102 parallelize 0.8 -13.4102 0.295 0.8 -13.4102 (landis 0.8 -13.4102 meyer, 0.8 -13.4102 1.03 0.8 -13.4102 well-trained 0.323529 -13.4109 z = 0.315789 -13.4109 translation model, 0.8 -13.4109 726 0.8 -13.4109 0.364 0.8 -13.4109 derivational, 0.8 -13.4109 that) 0.8 -13.4109 high-coverage 0.8 -13.4109 has-part 0.8 -13.4109 deem 0.8 -13.4109 tokenbased 0.8 -13.4109 0.769 0.8 -13.4109 hmm’s 0.8 -13.4109 low (2004), 0.8 -13.4109 english-japanese 0.15894 -13.4114 are generated 0.391304 -13.4116 lms. 0.8 -13.4116 neutron 0.8 -13.4116 set-up, 0.8 -13.4116 panels 0.8 -13.4116 alternations, 0.8 -13.4116 duverle 0.8 -13.4116 tempeval-2 0.6 -13.4123 topic: 0.8 -13.4123 21.3% 0.8 -13.4123 kruijff, 0.8 -13.4123 jesus 0.8 -13.4123 dictates 0.8 -13.4123 wurden 0.8 -13.4123 83.4% 0.8 -13.4123 concessions 0.8 -13.4123 prop-dep 0.8 -13.4123 mbdp-1, 0.8 -13.4123 english-spanish, 0.8 -13.4123 felicitous 0.8 -13.4123 0.560 0.8 -13.4123 drawbacks: 0.8 -13.4123 semcor/ptb 0.8 -13.4123 50.6% 0.202247 -13.4128 creating a 0.714286 -13.413 shawe-taylor, 0.8 -13.413 64.7% 0.8 -13.413 ends, 0.8 -13.413 portion, 0.8 -13.413 φ2 0.8 -13.413 reyle, 0.357143 -13.4133 however, unlike 0.714286 -13.4137 loves 0.206897 -13.4137 strategies for 0.8 -13.4137 accross 0.8 -13.4137 capacities 0.8 -13.4137 “web” 0.8 -13.4137 4b 0.8 -13.4137 1995, 0.8 -13.4137 unitary 0.8 -13.4137 (fp) 0.8 -13.4137 english-german, 0.8 -13.4137 (against 0.8 -13.4137 subforum 0.8 -13.4137 kbp2010 0.8 -13.4137 (feldman 0.8 -13.4137 (textual 0.6 -13.4144 position within 0.8 -13.4144 faithful 0.8 -13.4144 quotes, 0.8 -13.4144 (turian 0.8 -13.4144 secondly 0.8 -13.4144 grf 0.8 -13.4144 ambiguity level 0.8 -13.4144 scale) 0.8 -13.4144 visualizes 0.8 -13.4144 bandyopadhyay, 0.8 -13.4144 581 0.6 -13.4149 sensors 0.8 -13.4151 apprenticeship 0.8 -13.4151 ontonotes, 0.8 -13.4151 meza-ruiz 0.8 -13.4151 piwek, 0.8 -13.4151 qs2 0.8 -13.4151 crego 0.8 -13.4151 ccl’s 0.8 -13.4151 *, 0.8 -13.4151 (general) 0.8 -13.4151 t1: 0.8 -13.4151 12.5 0.6 -13.4154 victoria 0.8 -13.4158 seg-1 0.8 -13.4158 (dumais 0.8 -13.4158 /?/ 0.8 -13.4158 92.52 0.8 -13.4158 duc-03 0.8 -13.4158 synonym, 0.8 -13.4158 65m 0.8 -13.4158 so: 0.8 -13.4158 lexica, 0.8 -13.4158 elaborating 0.8 -13.4158 (kamp 0.8 -13.4158 92.02 0.25 -13.416 pieces 0.6 -13.416 gnu 0.288889 -13.4161 observations. 0.714286 -13.4162 “word 0.8 -13.4165 0.009 0.8 -13.4165 58.22 0.8 -13.4165 french-to-english 0.8 -13.4165 0.111 0.8 -13.4165 christensen 0.8 -13.4165 1,000, 0.8 -13.4165 summary-based 0.8 -13.4165 unsimplified 0.8 -13.4165 jill 0.8 -13.4165 53.85 0.714286 -13.4168 +loclm 0.6 -13.417 (dialogue 0.545455 -13.4172 c∈c 0.8 -13.4172 fulfilled 0.8 -13.4172 intensities 0.8 -13.4172 (chappelier 0.8 -13.4172 crane, 0.8 -13.4172 15000 0.8 -13.4172 (nullnullnullnullnull 0.8 -13.4172 lightlysupervised 0.8 -13.4172 retweets 0.8 -13.4172 (polanyi 0.8 -13.4172 100%, 0.8 -13.4172 sheet 0.8 -13.4172 possible; 0.8 -13.4172 = 0.5). 0.8 -13.4172 pythy 0.8 -13.4172 designed, 0.8 -13.4172 onishi 0.357143 -13.4174 supported by nsf 0.545455 -13.4176 do: 0.8 -13.4179 modifier(s) 0.8 -13.4179 9.2% 0.8 -13.4179 0.206 0.8 -13.4179 figure 3: sample 0.8 -13.4179 σj 0.8 -13.4179 chunk-based 0.8 -13.4179 handing 0.8 -13.4179 1.01 0.5 -13.4181 kapa 0.538462 -13.4182 this point. 0.8 -13.4186 (totally 0.8 -13.4186 1,815 0.8 -13.4186 serbia 0.8 -13.4186 klementiev, 0.8 -13.4186 task-related 0.8 -13.4186 (klapaftis 0.8 -13.4186 asumed 0.8 -13.4186 wcn, 0.8 -13.4186 hom 0.8 -13.4186 parser-based 0.8 -13.4186 0.413 0.8 -13.4186 “3” 0.8 -13.4193 zuidema, 0.8 -13.4193 hyderabad 0.8 -13.4193 yard 0.8 -13.4193 parsers) 0.8 -13.4193 (dietterich, 0.8 -13.4193 symbols: 0.8 -13.4193 sporleder 0.8 -13.4193 base-phrases 0.8 -13.4193 daume 0.8 -13.4193 a3s 0.8 -13.4193 presence/absence 0.8 -13.4193 handaligned 0.8 -13.4193 disregards 0.8 -13.4193 expectation, 0.714286 -13.4199 certain degree 0.8 -13.42 object’s 0.8 -13.42 ancestors, 0.8 -13.42 (sense) 0.8 -13.42 nomoto 0.8 -13.42 (erjavec 0.8 -13.42 markable, 0.8 -13.42 cohesion, 0.8 -13.42 ofk 0.8 -13.42 aiding 0.8 -13.42 experiment evaluates 0.8 -13.42 comes out 0.8 -13.42 exposition 0.8 -13.42 woepos 0.8 -13.42 nsf, 0.8 -13.42 round-trip 0.8 -13.42 category c, 0.8 -13.42 8.0% 0.545455 -13.4205 imageability and 0.538462 -13.4205 0.45 0.5 0.538462 -13.4205 ‘ . 0.409091 -13.4206 or, usa 0.0545886 -13.4206 data, 0.6 -13.4207 inequality 0.8 -13.4207 ku-ka pap-ul 0.8 -13.4207 corpus2 0.8 -13.4207 epr 0.8 -13.4207 scheme’s 0.8 -13.4207 wn++, 0.8 -13.4207 null0’s 0.8 -13.4207 (lewis, 0.8 -13.4207 propbank-style 0.8 -13.4207 clarified 0.8 -13.4207 retail 0.8 -13.4207 tolerant 0.545455 -13.421 our method achieves 0.714286 -13.4212 (7), 0.714286 -13.4212 indicates that, 0.6 -13.4212 0.003 0.8 -13.4214 paste 0.8 -13.4214 fewer, 0.8 -13.4214 75.3% 0.8 -13.4214 skipping 0.8 -13.4214 goutte, 0.8 -13.4214 443 0.8 -13.4214 mknlm 0.8 -13.4221 coarse, 0.8 -13.4221 0null075 0.8 -13.4221 systems), 0.8 -13.4221 enthymeme 0.8 -13.4221 phrasepair 0.8 -13.4221 characters), 0.8 -13.4221 estimators 0.8 -13.4221 judge’s 0.8 -13.4221 wim 0.8 -13.4221 6these 0.8 -13.4221 92.7% 0.8 -13.4221 s3vm 0.8 -13.4221 cross-validated 0.8 -13.4221 strikingly 0.8 -13.4221 contexts) 0.8 -13.4221 (ace, 0.175 -13.4224 here, we 0.714286 -13.4224 exponential models 0.217949 -13.4226 emnlp 0.6 -13.4228 f8 0.8 -13.4228 hpam 0.8 -13.4228 repeated, 0.8 -13.4228 alen 0.8 -13.4228 quicker 0.8 -13.4228 extractive, 0.8 -13.4228 (extended 0.8 -13.4228 input; 0.8 -13.4228 tap 0.8 -13.4228 (bar-haim 0.8 -13.4228 0.582 0.538462 -13.4229 curves are 0.344828 -13.4235 card 0.8 -13.4235 lavergne 0.8 -13.4235 0.644 0.8 -13.4235 unintelligible 0.8 -13.4235 max-score 0.8 -13.4235 (vertical 0.8 -13.4235 holdings 0.8 -13.4235 italics, 0.8 -13.4235 50.0% 0.8 -13.4235 sheng 0.714286 -13.4237 cyril 0.276596 -13.4239 different values 0.0469484 -13.4241 value 0.8 -13.4242 bi-lingual 0.8 -13.4242 change; 0.8 -13.4242 security. 0.8 -13.4242 (naseem 0.8 -13.4242 sβ>γ 0.8 -13.4242 0.902 0.8 -13.4242 thadani 0.8 -13.4242 treebanking 0.8 -13.4242 mullen 0.8 -13.4242 96.6 0.714286 -13.4243 table 1 describes 0.25 -13.4243 environment. 0.538462 -13.4243 (deerwester 0.109677 -13.4248 this work, 0.8 -13.4249 rocky 0.8 -13.4249 complies 0.8 -13.4249 major drawback 0.8 -13.4249 distinguish, 0.8 -13.4249 utters 0.8 -13.4249 0.689 0.8 -13.4249 foundin 0.8 -13.4249 2-2 0.8 -13.4249 officer 0.8 -13.4249 wife 0.8 -13.4249 dcs, 0.8 -13.4249 lean 0.8 -13.4249 (genest 0.8 -13.4249 42.11 0.8 -13.4249 schemata 0.8 -13.4249 semantic) 0.5 -13.4252 0.8 -13.4913 restrained 0.625 -13.492 µ1 0.545455 -13.492 asia, 0.8 -13.4921 16.6% 0.8 -13.4921 tara 0.8 -13.4921 da-level summarization 0.8 -13.4921 role-semantic 0.8 -13.4921  with 0.8 -13.4921 0.149 0.8 -13.4921 phrase-extraction 0.8 -13.4921 (cfg) 0.8 -13.4921 manli-exact 0.8 -13.4921 asp 0.8 -13.4921 857 0.8 -13.4921 toulouse, 0.8 -13.4921 log-likelihood, 0.103343 -13.4922 asked 0.625 -13.4926 (lp), 0.466667 -13.4929 rule application 0.8 -13.4929 descendants, 0.8 -13.4929 informationstatus 0.8 -13.4929 (uzuner, 0.8 -13.4929 24.03 0.8 -13.4929 88.34 0.8 -13.4929 john, 0.8 -13.4929 “user 0.8 -13.4929 both) 0.8 -13.4929 realvalued 0.8 -13.4929 off” 0.391304 -13.4932 the game. 0.8 -13.4936 tune, 0.8 -13.4936 gs-80 0.8 -13.4936 reverses 0.8 -13.4936 53.37 0.8 -13.4936 weights) 0.8 -13.4936 question-answering, 0.8 -13.4936 windows, 0.8 -13.4936 (viz., 0.8 -13.4936 with和 0.8 -13.4936 president’ 0.8 -13.4936 binds 0.8 -13.4936 index) 0.8 -13.4936 thek-best 0.8 -13.4936 shadowing 0.172414 -13.494 0 0 0 0.625 -13.4943 kulkarni 0.297297 -13.4943 (schmid, 0.292683 -13.4944 objective. 0.8 -13.4944 (ferraresi 0.8 -13.4944 pitfalls 0.8 -13.4944 only: 0.8 -13.4944 983 0.8 -13.4944 ‘suffix’ 0.8 -13.4944 ski 0.8 -13.4944 mt3 0.8 -13.4944 0.863 0.8 -13.4944 mcdonald’s 0.8 -13.4944 rightmost column 0.8 -13.4944 ceaf-e 0.8 -13.4944 grasp, 0.0202576 -13.4947 use 0.625 -13.4949 sadrzadeh, 0.8 -13.4951 node: 0.8 -13.4951 degrading 0.8 -13.4951 disappointingly, 0.8 -13.4951 vygen, 0.8 -13.4951 card, 0.8 -13.4951 hfw 0.8 -13.4951 statesplitting 0.8 -13.4951 “low” 0.8 -13.4951 john’s 0.8 -13.4951 elaboration, 0.625 -13.4955 673 0.625 -13.4955 promised 0.625 -13.4955 regularizing 0.625 -13.4955 45th 0.8 -13.4959 baseline(en) 0.8 -13.4959 touchable 0.8 -13.4959 1978) 0.8 -13.4959 +0.04 0.8 -13.4959 87.5% 0.8 -13.4959 grammar; 0.297297 -13.496 zhou et 0.545455 -13.4962 parameter c 0.545455 -13.4962 about 30 0.36 -13.4964 observed word 0.8 -13.4966 nullnullnullnullnullnullnullnullnullnull 0.8 -13.4966 cancer’ 0.8 -13.4966 (sher, 0.8 -13.4966 1.43 0.8 -13.4966 gay 0.8 -13.4966 (qazvinian 0.8 -13.4966 single-link 0.8 -13.4966 avatar’s 0.8 -13.4966 discounting, 0.8 -13.4966 (paraphrase 0.625 -13.4973 hybhsum 0.8 -13.4974 clustering) 0.8 -13.4974 kallen, 0.8 -13.4974 “every 0.8 -13.4974 elected 0.8 -13.4974 hopping 0.8 -13.4974 2009: 0.8 -13.4974 97.7 0.5 -13.498 100.0% 0.8 -13.4982 2006b), 0.8 -13.4982 provably 0.8 -13.4982 1.5, 0.8 -13.4982 ‘i’ 0.8 -13.4982 40.2 0.8 -13.4982 wildcards, 0.8 -13.4982 translate) 0.8 -13.4989 digraph 0.8 -13.4989 conjugate 0.8 -13.4989 surer 0.8 -13.4989 gabor 0.8 -13.4989 2all 0.8 -13.4989 0.272 0.8 -13.4989 1350 0.8 -13.4989 non-terminal, 0.8 -13.4989 friedman, 0.8 -13.4989 47.02 0.4 -13.499 class based 0.625 -13.4991 1e+ 0.36 -13.4994 be extracted. 0.8 -13.4997 fullimpl 0.8 -13.4997 responsiveness, 0.8 -13.4997 y”) 0.8 -13.4997 string; 0.8 -13.4997 mgi’s 0.8 -13.4997 datask 0.8 -13.4997 add-one 0.8 -13.4997 development: 0.8 -13.4997 gesture, 0.8 -13.4997 diagram, 0.8 -13.4997 nodes: 0.8 -13.4997 124k 0.202381 -13.4999 point, 0.421053 -13.5 rna 0.322581 -13.5002 material. 0.625 -13.5003 ∈ σ∗ 0.8 -13.5004 interpretation: 0.8 -13.5004 chinese/english 0.8 -13.5004 siddharth 0.8 -13.5004 tf-itf 0.8 -13.5004 history: 0.8 -13.5004 (korte 0.8 -13.5004 sacks, 0.8 -13.5004 emphasis, 0.210526 -13.5007 mentions are 0.625 -13.5009 an arity 0.8 -13.5012 stuff 0.8 -13.5012 0.779 0.8 -13.5012 910 0.8 -13.5012 gelbukh, 0.8 -13.5012 counting, 0.8 -13.5012 topbs 0.8 -13.5012 bleu% 0.8 -13.5012 procesed 0.8 -13.5012 buildings, 0.8 -13.5012 tracks: 0.8 -13.5012 1990s, 0.8 -13.5012 [x,y] 0.8 -13.5012 revisiting 0.8 -13.5012 third-party 0.8 -13.5012 lm: 0.237288 -13.5013 for measuring 0.5 -13.5018 country, 0.5 -13.5018 instituto 0.545455 -13.5019 opi score 0.8 -13.502 phenomena: 0.8 -13.502 egyptian 0.8 -13.502 c3-f 0.8 -13.502 lms: 0.8 -13.502 ow, 0.8 -13.502 c3-t 0.8 -13.502 precision(%) 0.8 -13.502 sentiwordnet synsets 0.5 -13.5022 director of 0.545455 -13.5024 cn1 0.8 -13.5027 hurdles 0.8 -13.5027 cancellation 0.8 -13.5027 suddenly 0.8 -13.5027 7.9% 0.8 -13.5027 nuances 0.8 -13.5027 non-traditional 0.8 -13.5027 (langkilde 0.8 -13.5027 “dogs 0.8 -13.5027 brute-force 0.333333 -13.5028 nmod 0.0102341 -13.503 as 0.4375 -13.503 extractor. 0.153846 -13.5031 event. 0.8 -13.5035 reflected 0.8 -13.5035 dependency-parsed 0.8 -13.5035 arbitrary-image 0.8 -13.5035 communicators 0.8 -13.5035 predefined threshold 0.8 -13.5043 book), 0.8 -13.5043 (google 0.8 -13.5043 outputs: 0.8 -13.5043 τ.sp 0.8 -13.5043 per-tweet 0.8 -13.5043 similarity; 0.8 -13.5043 ¢¢¢ 0.8 -13.5043 how, 0.625 -13.5044 wi ∈ 0.466667 -13.5044 works better 0.164062 -13.5048 we note that 0.8 -13.505 panasonic 0.8 -13.505 messenger 0.8 -13.505 getoor, 0.8 -13.505 language/cultural 0.8 -13.505 39.19 0.8 -13.505 87.73 0.8 -13.505 non-consecutive 0.8 -13.505 independently-trained 0.8 -13.505 parentheses, 0.8 -13.505 detract 0.8 -13.505 pervasiveness 0.8 -13.505 editdist(u,v;x) 0.8 -13.5058 consumption, 0.8 -13.5058 distillation 0.8 -13.5058 condition’s 0.8 -13.5058 should, 0.8 -13.5058 ting 0.8 -13.5058 < .001). 0.8 -13.5058 english-to-dutch 0.25 -13.5059 proceed 0.333333 -13.5063 sentence contains 0.8 -13.5066 method2 0.8 -13.5066 synthesizes 0.8 -13.5066 physiological 0.8 -13.5066 poles 0.8 -13.5066 64.50 0.8 -13.5066 advise 0.8 -13.5066 11.7% 0.8 -13.5066 representation; 0.625 -13.5068 1965; 0.625 -13.5068 macro-avg 0.8 -13.5073 unspoken 0.8 -13.5073 honest 0.8 -13.5073 game’s 0.8 -13.5073 58.4% 0.8 -13.5073 “approximate 0.8 -13.5073 405 0.8 -13.5073 0.318 0.8 -13.5073 body, 0.8 -13.5073 frobenius 0.344828 -13.5078 finkel 0.8 -13.5081 duty 0.8 -13.5081 𝑌 0.8 -13.5081 remedies 0.8 -13.5081 dr-da 0.8 -13.5081 unit; 0.8 -13.5081 fuzziness 0.8 -13.5081 tailed 0.8 -13.5081 (2011b) 0.8 -13.5081 98.6% 0.8 -13.5081 complementation, 0.8 -13.5081 modes, 0.8 -13.5081 ppolkn 0.625 -13.5086 clustering coefficient 0.8 -13.5089 c5 0.8 -13.5089 smartcorect 0.8 -13.5089 08540 0.8 -13.5089 28.06 0.8 -13.5089 negations, 0.8 -13.5089 :b 0.8 -13.5089 pre-procesing 0.625 -13.5092 income 0.625 -13.5092 avoir 0.545455 -13.5092 dpi 0.8 -13.5096 t op 0.8 -13.5096 388 0.8 -13.5096 presence, 0.8 -13.5096 (fernando, 0.8 -13.5096 49.6% 0.8 -13.5096 holonymy 0.8 -13.5096 accomplish, 0.4 -13.5097 that arise 0.625 -13.5098 assigned, 0.314286 -13.5102 biotope 0.8 -13.5104 “with” 0.8 -13.5104 post-social 0.8 -13.5104 “going 0.8 -13.5104 alto 0.8 -13.5104 54.30 0.8 -13.5104 planing 0.8 -13.5104 t’s 0.126168 -13.5105 we train 0.545455 -13.5107 and riloff 0.4375 -13.5109 (2006) use 0.237288 -13.5111 edinburgh 0.8 -13.5112 742 0.8 -13.5112 (proper 0.8 -13.5112 build, 0.8 -13.5112 levine, 0.8 -13.5112 (constant 0.8 -13.5112 managers 0.8 -13.5119 event; 0.8 -13.5119 hist 0.8 -13.5119 87.57 0.8 -13.5119 (as, 0.8 -13.5119 bilingual dictionaries. 0.8 -13.5119 ihr 0.8 -13.5119 feel, 0.8 -13.5119 hispanic 0.8 -13.5119 biomed 0.8 -13.5127 1967; 0.8 -13.5127 hockey 0.8 -13.5127 skyline 0.8 -13.5127 verbmobil 0.8 -13.5127 verb), 0.8 -13.5127 yamada, 0.625 -13.5128 mwa model 0.625 -13.5128 (repeated) 0.545455 -13.5128 word wj 0.237288 -13.5129 the next section, 0.13 -13.5132 nodes in 0.8 -13.5135 26.52 0.8 -13.5135 interfering 0.8 -13.5135 nelson 0.8 -13.5135 m&h-f 0.8 -13.5135 rouge-w 0.8 -13.5135 992 0.8 -13.5135 340 0.110714 -13.514 versions 0.8 -13.5142 vivid 0.8 -13.5142 impasse state 0.8 -13.5142 parsers; 0.8 -13.5142 adv: 0.8 -13.5142 inefficient, 0.8 -13.5142 measure; 0.8 -13.5142 kobayashi, 0.8 -13.5142 watson, 0.8 -13.5142 basque, 0.8 -13.5142 sato 0.8 -13.5142 kpbl1 0.8 -13.5142 manabu 0.289474 -13.5146 observed. 0.322581 -13.5148 novel method 0.181818 -13.5149 all. 0.8 -13.515 25.90 0.8 -13.515 2.) 0.8 -13.515 esslli 0.8 -13.515 0.516 0.8 -13.515 customizing 0.12069 -13.5153 treat 0.191489 -13.5158 translations are 0.8 -13.5158 inand 0.8 -13.5158 caveat 0.8 -13.5158 semnet 0.8 -13.5158 2s 0.8 -13.5158 0.684 0.8 -13.5158 non-positive 0.8 -13.5158 fruitfully 0.8 -13.5158 prefabricated 0.8 -13.5158 (written 0.8 -13.5158 states: 0.8 -13.5158 lingpipe 0.625 -13.5165 (tc) 0.4 -13.5166 ,x 0.8 -13.5166 little, 0.8 -13.5166 25.40 0.8 -13.5166 precise definition 0.8 -13.5166 title: 0.8 -13.5166 word+3-gram 0.8 -13.5166 73.50 0.8 -13.5166 hashes 0.466667 -13.517 morfessor 0.625 -13.5171 497 0.545455 -13.5171 maximality 0.8 -13.5173 details, see 0.8 -13.5173 vocabulary) 0.8 -13.5173 modulation 0.8 -13.5173 traversal, 0.8 -13.5173 drdas. 0.8 -13.5173 captain 0.8 -13.5173 “tree 0.8 -13.5173 sociology, 0.4375 -13.5177 eprime 0.8 -13.5181 576 0.8 -13.5181 thus: 0.8 -13.5181 caa 0.8 -13.5181 genbank 0.8 -13.5181 0.331 0.8 -13.5181 mahesh 0.8 -13.5181 isocat 0.8 -13.5181 1.40 0.8 -13.5181 --------- 0.0442529 -13.5184 there is 0.375 -13.5185 several methods 0.5 -13.5186 propagating 0.8 -13.5189 programming-based 0.8 -13.5189 reliant 0.4375 -13.519 dif 0.36 -13.5195 precision (p), 0.545455 -13.5197 langlais, 0.8 -13.5197 86.32 0.8 -13.5197 4.52 0.8 -13.5197 (nps) 0.8 -13.5197 latencies 0.8 -13.5197 phases, 0.8 -13.5197 fj, 0.8 -13.5197 explorer 0.8 -13.5197 children; 0.8 -13.5197 thought, 0.8 -13.5197 type+arg+sal 0.8 -13.5197 0.572 0.8 -13.5197 resultative 0.213333 -13.5199 (nivre 0.5 -13.52 federico, 0.333333 -13.5201 mann 0.8 -13.5204 injure 0.8 -13.5204 extra-linguistic 0.8 -13.5204 y) ◦ 0.8 -13.5204 “run” 0.8 -13.5212 schneider, 0.8 -13.5212 2001): 0.8 -13.5212 position; 0.8 -13.5212 outweighed 0.8 -13.5212 slots: 0.8 -13.5212 0.033 0.8 -13.5212 41.29 0.8 -13.5212 zc 0.8 -13.5212 change-trend 0.8 -13.5212 entity) 0.8 -13.5212 (ct) 0.8 -13.5212 baseline’s 0.8 -13.5212 submodules 0.8 -13.5212 affixation, 0.8 -13.5212 select, 0.8 -13.5212 domain-general 0.625 -13.5213 0.848 0.4375 -13.5215 english/hindi 0.545455 -13.5218 a snapshot 0.8 -13.522 doublets 0.8 -13.522 correlation-based 0.625 -13.5225 vd(fn) 0.4 -13.5228 week. 0.466667 -13.5228 earth 0.8 -13.5228 nulla 0.8 -13.5228 98.7% 0.8 -13.5228 1.49 0.8 -13.5228 90.90 0.8 -13.5228 bng 0.8 -13.5228 neurological 0.8 -13.5228 value), 0.545455 -13.5229 0.2. 0.102167 -13.5229 the output of 0.305556 -13.5236 answers to 0.8 -13.5236 acoustic correlates 0.8 -13.5236 important; 0.8 -13.5236 experiential 0.8 -13.5236 katrin 0.8 -13.5236 (difference 0.8 -13.5236 connotation, 0.8 -13.5236 ordinate 0.8 -13.5236 subframe-parallel 0.8 -13.5236 crimes 0.8 -13.5236 gesturing 0.8 -13.5236 0 0.014562 0.8 -13.5236 losielle 0.545455 -13.524 grew 0.3125 -13.5241 unsupervised data 0.625 -13.5243 unsupervised morphology 0.8 -13.5243 marius 0.8 -13.5243 furthest 0.8 -13.5243 persuade 0.8 -13.5243 1.60 0.8 -13.5243 (joshi, 0.8 -13.5243 0.219 0.8 -13.5243 pos: 0.289474 -13.5244 or equal 0.8 -13.5251 954 0.8 -13.5251 vegetables 0.8 -13.5251 pcfgs, 0.8 -13.5251 hyphen, 0.8 -13.5251 volvo 0.8 -13.5251 w1,...,wn 0.625 -13.5256 arrives 0.333333 -13.5259 first proposed 0.8 -13.5259 prettenhofer 0.8 -13.5259 2∨tp 0.8 -13.5259 0.052 0.8 -13.5259 sblr 0.8 -13.5259 automaton, 0.466667 -13.5265 massachusetts, 0.8 -13.5267 security, 0.8 -13.5267 saturated 0.8 -13.5267 nonadaptive 0.8 -13.5267 alignments) 0.8 -13.5267 1998); 0.8 -13.5267 678 0.8 -13.5267 flavors 0.625 -13.5268 eggs, 0.5 -13.5269 d∈d 0.8 -13.5275 648 0.8 -13.5275 pg 0.8 -13.5275 homology 0.8 -13.5275 summarized, 0.8 -13.5275 (ps) 0.8 -13.5275 hoang 0.8 -13.5275 markovian 0.8 -13.5275 einem 0.8 -13.5275 postpositional 0.8 -13.5275 fishlov 0.8 -13.5275 biologicalevent 0.8 -13.5275 patron’s 0.8 -13.5275 taking, 0.8 -13.5275 88.8% 0.4375 -13.5279 hierarchical models 0.625 -13.528 prompted 0.8 -13.5282 (std) 0.8 -13.5282 jolla, 0.8 -13.5282 korean) 0.8 -13.5282 [p] 0.8 -13.5282 passive edges 0.346154 -13.5284 any additional 0.545455 -13.5287 tf*itf 0.5 -13.5288  0.8 -13.529 cally 0.8 -13.529 flipped 0.8 -13.529 regularizer, 0.8 -13.529 con, 0.8 -13.529 literals, 0.625 -13.5292 appreciates 0.625 -13.5298 87.70 0.8 -13.5298 uninflected 0.8 -13.5298 (strongly 0.8 -13.5298 g: 0.8 -13.5298 (abstract 0.8 -13.5298 (kennedy 0.8 -13.5298 outermost 0.380952 -13.5301 two points 0.375 -13.5302 evaluation methods 0.625 -13.5305 .34 0.4 -13.5306 relatedness. 0.8 -13.5306 laptop, 0.8 -13.5306 piij 0.8 -13.5306 generation) 0.8 -13.5306 kunnen 0.100304 -13.5307 well. 0.5 -13.5308 encn 0.625 -13.5311 and henderson 0.8 -13.5314 cable 0.8 -13.5314 61.1% 0.8 -13.5314 913 0.8 -13.5314 checker, 0.625 -13.5317 rig 0.545455 -13.532 sentiment classifiers 0.421053 -13.5321 broke 0.8 -13.5322 designations 0.8 -13.5322 β3 0.545455 -13.5325 [first-pron] 0.421053 -13.533 dist 0.8 -13.533 konstanz 0.8 -13.533 amplitude 0.8 -13.533 voca, 0.8 -13.533 nlp: 0.8 -13.533 qa, 0.8 -13.533 solution: 0.8 -13.533 mari 0.8 -13.533 tag’s 0.8 -13.533 k1a 0.346154 -13.5332 feedback from 0.380952 -13.5334 overcome these 0.625 -13.5335 x), 0.8 -13.5337 xie, 0.8 -13.5337 standard format 0.8 -13.5337 etc. (see 0.8 -13.5337 concatenations 0.625 -13.5341 mcore 0.4375 -13.5344 λ2, 0.8 -13.5345 renormalized 0.8 -13.5345 ε, 0.8 -13.5345 anoop 0.8 -13.5345 aol 0.8 -13.5345 mtmp 0.8 -13.5345 assimilation 0.8 -13.5345 (verbs) 0.8 -13.5345 58.13 0.8 -13.5345 subunit-complex 0.8 -13.5345 (pos, 0.8 -13.5345 oficial 0.4 -13.5346 preferences for 0.8 -13.5353 gbytes 0.8 -13.5353 (asr 0.8 -13.5353 biochemistry 0.8 -13.5353 (sucre) 0.8 -13.5353 indistinguishable, 0.132275 -13.5354 al., 2005; 0.421053 -13.5354 p(c/¬a) 0.625 -13.5354 els 0.625 -13.5354 1available 0.5 -13.5357 predicates like 0.4375 -13.5361 tiered 0.8 -13.5361 +0.27 0.8 -13.5361 opposites 0.8 -13.5361 synsetid 0.8 -13.5361 91.56 0.8 -13.5361 keypoints 0.8 -13.5361 1015 0.8 -13.5361 3.80 0.8 -13.5361 1970) 0.8 -13.5361 3.21 0.545455 -13.5363 ceafm 0.333333 -13.5367 column. 0.8 -13.5369 (allowing 0.8 -13.5369 (+: 0.8 -13.5369 state; 0.8 -13.5369 homogeneous, 0.8 -13.5369 estrogen 0.8 -13.5369 ‘n’ 0.8 -13.5369 copula, 0.8 -13.5377 91.8% 0.8 -13.5377 ying 0.8 -13.5377 representable 0.8 -13.5377 nov 0.8 -13.5377 crosses 0.8 -13.5377 26.25 0.8 -13.5377 memetube 0.8 -13.5377 (pado 0.8 -13.5377 ppol0 0.8 -13.5377 catalog, 0.8 -13.5377 persistence 0.625 -13.5378 so-wtp 0.188889 -13.5379 bleu (papineni 0.545455 -13.5379 system finds 0.289474 -13.538 attracted 0.27907 -13.5381 we adopt a 0.5 -13.5382 u,v 0.5 -13.5382 a 5-fold 0.8 -13.5385 “interest” 0.8 -13.5385 21.47 0.8 -13.5385 clipping 0.8 -13.5385 analysts, 0.8 -13.5385 flatter 0.25 -13.539 numbers are 0.8 -13.5393 28.6% 0.8 -13.5393 truncate 0.8 -13.5393 inconsistency, 0.8 -13.5393 subgradient-based 0.8 -13.5393 ¯θ) 0.8 -13.5393 878 0.8 -13.5393 bag-of-words, 0.8 -13.5393 (combined 0.8 -13.5393 aug 0.8 -13.5393 content-bearing 0.625 -13.5397 +0.6 0.625 -13.5397 x2, 0.8 -13.5401 verbalisation 0.8 -13.5401 {t 0.8 -13.5401 muscle 0.8 -13.5401 bandyopadhyay 0.8 -13.5401 bugs 0.8 -13.5401 (9 0.380952 -13.5406 gao, 0.8 -13.5408 widen 0.8 -13.5408 scopefinder 0.8 -13.5408 mails 0.8 -13.5408 stored, 0.8 -13.5408 biography, 0.625 -13.5409 tape 0.205128 -13.5413 4th 0.8 -13.5416 bic 0.8 -13.5416 1193 0.8 -13.5416 interviews, 0.8 -13.5416 720 0.625 -13.5422 confirm this 0.625 -13.5422 definition question 0.8 -13.5424 speaker) 0.8 -13.5424 p(α 0.8 -13.5424 postag 0.625 -13.5428 greece 0.8 -13.5432 information, including 0.8 -13.5432 religion, 0.8 -13.5432 cd-comb 0.8 -13.5432 41.16 0.8 -13.5432 subspaces 0.8 -13.5432 point: 0.8 -13.5432 0.735 0.8 -13.5432 compactness, 0.4375 -13.5439 despite these 0.8 -13.544 doubled 0.8 -13.544 “find 0.8 -13.544 32.65 0.8 -13.544 00.1 0.8 -13.544 avp 0.8 -13.544 goods 0.8 -13.544 dictionaries: 0.3125 -13.5447 bullets 0.8 -13.5448 extra-sentential 0.8 -13.5448 graph) 0.8 -13.5448 conversants 0.8 -13.5448 function), 0.8 -13.5448 ungar, 0.8 -13.5448 cell-closing 0.625 -13.5453 blackberry 0.625 -13.5453 abbreviates 0.545455 -13.5454 buckwalter 0.466667 -13.5454 ur 0.8 -13.5456 42.77 0.8 -13.5456 (s[wq]/(s[dcl]\np))/n, 0.8 -13.5456 805 0.8 -13.5456 meteor-next 0.8 -13.5456 acurate 0.8 -13.5456 nwire 0.5 -13.5456 pem 0.625 -13.5459 fei 0.625 -13.5459 p(a 0.5 -13.5461 777 0.8 -13.5464 ross 0.8 -13.5464 restored 0.8 -13.5464 (10), 0.8 -13.5464 “speech 0.8 -13.5464 amazing 0.8 -13.5464 three-level 0.8 -13.5464 (swier 0.625 -13.5465 stime 0.113725 -13.5466 2002; 0.333333 -13.547 wordnet: 0.625 -13.5472 ramage, 0.8 -13.5472 “does 0.8 -13.5472 28.17 0.8 -13.5472 tukey’s 0.8 -13.5472 walter 0.8 -13.5472 verlag, 0.8 -13.5472 xi(d) 0.8 -13.5472 suffering 0.8 -13.5472 dtus 0.26087 -13.5476 large. 0.8 -13.548 maynard 0.8 -13.548 florida, 0.8 -13.548 juxtaposition 0.8 -13.548 any stereotype 0.8 -13.548 facet 0.8 -13.548 numbers: 0.4375 -13.5482 which applies 0.8 -13.5488 curious 0.8 -13.5488 actor, 0.8 -13.5488 terminology) 0.8 -13.5488 63.71 0.346154 -13.549 parallel sentences. 0.282051 -13.5493 be understood 0.545455 -13.5493 ‘x 0.8 -13.5496 toponyms 0.8 -13.5496 disaster, 0.8 -13.5496 well-formedness 0.8 -13.5496 bsoa,property 0.197531 -13.5498 instance. 0.545455 -13.5498 qui 0.555556 -13.5502 idafa 0.625 -13.5503 smile 0.625 -13.5503 slovenian 0.666667 -13.5504 f1 auc 0.666667 -13.5504 h0 0.666667 -13.5504 indicates, 0.666667 -13.5504 (10%) 0.666667 -13.5504 nullnullnullnullnullnullnullnullnullnullnullnullnull 0.666667 -13.5504 p(s 0.666667 -13.5504 maise is 0.8 -13.5504 january, 0.8 -13.5504 cook, 0.8 -13.5504 parsing; 0.8 -13.5504 elsner, 0.8 -13.5504 universe 0.333333 -13.5506 were tuned 0.625 -13.5509 list; 0.666667 -13.551 tension 0.666667 -13.551 carve 0.666667 -13.551 highand 0.202532 -13.5512 learning methods 0.8 -13.5512 tweeting 0.8 -13.5512 verb; 0.8 -13.5512 woz-08 + 0.8 -13.5512 shih, 0.8 -13.5512 (further 0.8 -13.5512 classifiers; 0.8 -13.5512 ask, 0.8 -13.5512 (hovy, 0.8 -13.5512 respectable 0.8 -13.5512 bos, 0.8 -13.5512 11this 0.8 -13.5512 23.44 0.555556 -13.5513 project: 0.346154 -13.5514 (dempster et 0.625 -13.5515 interchangeable 0.666667 -13.5517 (shriberg et 0.666667 -13.5517 congress 0.555556 -13.5519 s(t) 0.555556 -13.5519 (chapman 0.8 -13.552 i'm 0.8 -13.552 80,000 0.8 -13.552 737 0.8 -13.552 1.10 0.8 -13.552 {y, 0.8 -13.552 nullt,k 0.625 -13.5522 limitation: 0.666667 -13.5524 (estimated 0.666667 -13.5524 “n” 0.666667 -13.5524 argmaxy 0.4 -13.5525 context features, 0.4375 -13.5526 use cases 0.8 -13.5528 87.44 0.8 -13.5528 discourse” 0.8 -13.5528 λ0 0.8 -13.5528 name] 0.8 -13.5528 (calculated 0.8 -13.5528 words/phrases 0.8 -13.5528 func 0.8 -13.5528 co-decoding 0.8 -13.5528 marked-up 0.8 -13.5528 linkage, 0.10989 -13.5532 step, 0.625 -13.5534 precision: 0.333333 -13.5536 mcclosky et 0.8 -13.5536 truck 0.8 -13.5536 dependences 0.8 -13.5536 shortest-derivation 0.8 -13.5536 tenenbaum, 0.8 -13.5536 upload 0.8 -13.5536 surpassed 0.8 -13.5536 variants) 0.8 -13.5536 37.6 0.666667 -13.5537 serif 0.545455 -13.5542 linear svm 0.666667 -13.5544 39.7 0.666667 -13.5544 aus, 0.666667 -13.5544 records: 0.666667 -13.5544 –coref 0.666667 -13.5544 nouns: 0.666667 -13.5544 prosodic, 0.8 -13.5544 nulla,bnull 0.8 -13.5544 one-token 0.8 -13.5544 703 0.8 -13.5544 application-specific 0.8 -13.5544 ok, 0.8 -13.5544 random: 0.8 -13.5544 (t,n) 0.8 -13.5544 chalenge 0.555556 -13.5547 sorry, 0.555556 -13.5547  0.36 -13.555 cand 0.666667 -13.555 bat, 0.666667 -13.555 preps 0.666667 -13.555 (landman, 0.666667 -13.555 brighton, 0.8 -13.5552 (1)) 0.8 -13.5552 show: 0.8 -13.5552 predictors, 0.8 -13.5552 (201), 0.8 -13.5552 political, 0.8 -13.5552 santorini, 0.8 -13.5552 multi-stage 0.8 -13.5552 rwn 0.545455 -13.5553 za 0.346154 -13.5555 β, 0.666667 -13.5557 jeong 0.666667 -13.5557 q) 0.333333 -13.5557 a triple 0.165289 -13.5558 years, 0.346154 -13.5559 wpred 0.625 -13.5559 clusters: 0.8 -13.556 raco 0.8 -13.556 relaxed, 0.8 -13.556 q0 0.666667 -13.5564 […] 0.666667 -13.5564 0.711 0.666667 -13.5564 style transformation 0.666667 -13.5564 ijk 0.555556 -13.5564 kp 0.36 -13.5565 x ∈ 0.8 -13.5568 software: 0.8 -13.5568 (yan et 0.8 -13.5568 0.139 0.8 -13.5568 409 0.8 -13.5568 unsure 0.8 -13.5568 accord 0.666667 -13.557 (so-called 0.5 -13.5571 40.9 0.12381 -13.5573 necessary to 0.555556 -13.5575 count: 0.466667 -13.5575 priori. 0.8 -13.5576 serdyukov et 0.8 -13.5576 censored 0.8 -13.5576 68.56 0.8 -13.5576 509 0.666667 -13.5577 eye-tracking 0.666667 -13.5577 trainedon 0.666667 -13.5577 english-to-english 0.666667 -13.5577 (corpus 0.666667 -13.5577 conjugation 0.625 -13.5578 any special 0.555556 -13.558 à 0.466667 -13.558 nicht 0.5 -13.5581 47.4 0.666667 -13.5584 371 0.8 -13.5584 s(x) 0.8 -13.5584 agricultural 0.8 -13.5584 dws 0.8 -13.5584 0.001, 0.555556 -13.5586 stream specific 0.555556 -13.5586 chairman 0.545455 -13.5586 lookup. 0.545455 -13.5586 data preprocessing 0.545455 -13.5586 scfg-hmd 0.666667 -13.5591 (identified 0.666667 -13.5591 re-ranker 0.666667 -13.5591 enabled, 0.380952 -13.5592 ensembl 0.8 -13.5593 tory 0.8 -13.5593 55.01 0.8 -13.5593 case markers 0.8 -13.5593 languagepair 0.8 -13.5593 lowercased, 0.8 -13.5593 (500 0.8 -13.5593 2011a), 0.8 -13.5593 club, 0.8 -13.5593 u-shaped 0.555556 -13.5597 jjs 0.8 -13.5601 pathogenic 0.8 -13.5601 es gibt 0.8 -13.5601 1973; 0.8 -13.5601 hiero-re 0.8 -13.5601 praise, 0.8 -13.5601 88.25 0.8 -13.5601 determination, 0.8 -13.5601 predicating 0.8 -13.5601 (1a), 0.8 -13.5601 (2a), 0.8 -13.5601 (1977), 0.8 -13.5601 ψprime 0.202532 -13.5602 limitation 0.666667 -13.5604 257 0.666667 -13.5604 suggest, 0.666667 -13.5604 non-reference 0.666667 -13.5604 ubs 0.666667 -13.5604 hllda 0.155556 -13.5606 size. 0.8 -13.5609 86.8% 0.8 -13.5609 29.1% 0.8 -13.5609 geopolitical 0.8 -13.5609 473 0.8 -13.5609 log, 0.8 -13.5609 log2 0.625 -13.561 (2c) 0.666667 -13.5611 beneficial, 0.666667 -13.5611 shrink 0.666667 -13.5611 house’ 0.666667 -13.5611 aε 0.625 -13.5616 nam 0.208333 -13.5616 project. 0.8 -13.5617 716 0.8 -13.5617 correctness, 0.8 -13.5617 (iaa) 0.8 -13.5617 ±1 0.8 -13.5617 dow 0.8 -13.5617 bernhard 0.8 -13.5617 wicopaco 0.5 -13.5617 duluth-2 0.666667 -13.5618 size 7. 0.666667 -13.5624 (hmm), 0.666667 -13.5624 chance baseline 0.666667 -13.5624 s&b 0.666667 -13.5624 wake 0.666667 -13.5624 c+ 0.240741 -13.5624 translation probability 0.8 -13.5625 cum 0.282051 -13.5626 marneffe et 0.0793037 -13.563 be used to 0.126214 -13.563 values for 0.666667 -13.5631 “≥others” 0.666667 -13.5631 grow, 0.666667 -13.5631 accuracy), 0.666667 -13.5631 11we 0.8 -13.5633 factuality 0.8 -13.5633 seating 0.8 -13.5633 three runs 0.8 -13.5633 sage 0.8 -13.5633 (detailed 0.8 -13.5633 clear distinction 0.8 -13.5633 open, 0.8 -13.5633 [black 0.8 -13.5633 73.33 0.625 -13.5635 che 0.666667 -13.5638 orwell’s 0.666667 -13.5638 0.019 0.666667 -13.5638 erd 0.666667 -13.5638 (person 0.380952 -13.5639 using linear 0.8 -13.5641 27.88 0.8 -13.5641 1.4% 0.8 -13.5641 motivation: 0.8 -13.5641 (majority 0.8 -13.5641 “a”, 0.8 -13.5641 group-level 0.8 -13.5641 wikipedia2 0.666667 -13.5645 matsoukas 0.666667 -13.5645 kent 0.666667 -13.5645 doubt, 0.217391 -13.5646 been done 0.555556 -13.5648 θt 0.625 -13.5648 circuit 0.625 -13.5648 snow, 0.8 -13.5649 johnson’s 0.8 -13.5649 coding, 0.8 -13.5649 italian sign 0.666667 -13.5651 created: 0.555556 -13.5653 87.1 0.555556 -13.5653 france, 0.466667 -13.5655 in (2), 0.236364 -13.5656 it shows 0.8 -13.5657 (2007) showed 0.8 -13.5657 σij 0.666667 -13.5658 (above 0.282051 -13.5663 morphological features 0.4375 -13.5663 subsections, 0.8 -13.5666 999 0.8 -13.5666 u.k. 0.8 -13.5666 filed 0.8 -13.5666 into: 0.625 -13.5667 96.0 0.168142 -13.5671 regarded 0.666667 -13.5672 corollary 0.666667 -13.5672 resample 0.666667 -13.5672 wcn 0.666667 -13.5672 large-margin 0.625 -13.5673 nicolas 0.8 -13.5674 '(y) 0.8 -13.5674 spanish, portuguese, 0.8 -13.5674 minh 0.8 -13.5674 deneefe, 0.8 -13.5674 boris 0.4375 -13.5676 core12 0.555556 -13.5676 local classification 0.555556 -13.5676 (gi) 0.666667 -13.5679 organisations 0.666667 -13.5679 backchannel-preceding 0.666667 -13.5679 race 0.625 -13.5679 0.591 0.545455 -13.568 award. 0.8 -13.5682 kenji 0.8 -13.5682 adjectives) 0.8 -13.5682 candito, 0.8 -13.5682 (total) 0.8 -13.5682 non-lexicalized 0.8 -13.5682 correspondsto 0.8 -13.5682 8m 0.666667 -13.5685 multi-reference 0.666667 -13.5685 woodsend 0.666667 -13.5685 568 0.4375 -13.5685 wizard: 0.625 -13.5686 1 (c) 0.109434 -13.5688 (koehn 0.282051 -13.5689 dataset was 0.8 -13.569 educated 0.8 -13.569 viewpoint, 0.8 -13.569 t; 0.8 -13.569 sfst 0.8 -13.569 u’s 0.8 -13.569 iv: 0.8 -13.569 reportedin 0.666667 -13.5692 linnaeus 0.666667 -13.5692 vision, 0.555556 -13.5693 clinical text 0.363636 -13.5695 we also investigate 0.466667 -13.5698 querrien 0.466667 -13.5698 sim.user: 0.8 -13.5698 uj/z 0.8 -13.5698 present; 0.8 -13.5698 precomputing 0.8 -13.5698 orthographically 0.666667 -13.5699 lr-hb4 0.666667 -13.5699 maker 0.666667 -13.5699 tcss 0.666667 -13.5699 query-dependent 0.555556 -13.5699 (steps 0.0544783 -13.57 this paper, 0.411765 -13.5703 francisco 0.0988024 -13.5704 the total number of 0.3125 -13.5705 tokyo 0.666667 -13.5706 391 0.666667 -13.5706 (artstein et 0.666667 -13.5706 one-sense-per-discourse 0.666667 -13.5706 sumo/milo 0.8 -13.5706 91.5% 0.8 -13.5706 examinations 0.8 -13.5706 (1965) 0.266667 -13.5707 exposure 0.5 -13.5708 mexico, 0.666667 -13.5713 z∗(i)q 0.666667 -13.5713 283 0.666667 -13.5713 “30 0.107527 -13.5713 believe that 0.18 -13.5714 vector. 0.8 -13.5714 composer 0.8 -13.5714 (subject, 0.346154 -13.5717 (1991) 0.625 -13.5718 wnf 0.666667 -13.5719 (2008, 0.666667 -13.5719 d[i,j 0.666667 -13.5719 equally good 0.666667 -13.5719 complicating 0.666667 -13.5719 wendy 0.126263 -13.5721 a pair of 0.8 -13.5723 team’s 0.8 -13.5723 drove 0.8 -13.5723 (significant 0.8 -13.5723 parameterizes 0.8 -13.5723 linoptimizer 0.8 -13.5723 (much 0.8 -13.5723 ˆxr 0.8 -13.5723 naw 0.625 -13.5724 (luo et 0.666667 -13.5726 desires 0.666667 -13.5726 69.77 0.10231 -13.5728 but it 0.8 -13.5731 subchallenge 0.8 -13.5731 (kawahara 0.8 -13.5731 “do” 0.8 -13.5731 criticize 0.8 -13.5731 (gb) 0.8 -13.5731 be) 0.8 -13.5731 rests 0.666667 -13.5733 citations) 0.666667 -13.5733 none) 0.0944444 -13.5734 our results 0.380952 -13.5736 sides. 0.625 -13.5737 projected parser 0.555556 -13.5738 previous example 0.8 -13.5739 robustness, 0.8 -13.5739 cold-start 0.8 -13.5739 on/p 0.8 -13.5739 0.48, 0.8 -13.5739 comlex, 0.666667 -13.574 “emotion” 0.411765 -13.5745 there. 0.205479 -13.5745 are created 0.666667 -13.5747 mais 0.666667 -13.5747 1979), 0.666667 -13.5747 4(a), 0.666667 -13.5747 1b) 0.666667 -13.5747 (boston,[bos]) 0.666667 -13.5747 mloc 0.666667 -13.5747 bernardini, 0.666667 -13.5747 fragments: 0.8 -13.5747 20.46 0.8 -13.5747 parentheses) 0.8 -13.5747 narratives, 0.8 -13.5747 lifetime 0.36 -13.5748 mode. 0.555556 -13.575 glass, 0.625 -13.575 pays 0.625 -13.575 loved 0.3125 -13.5753 conll-x 0.666667 -13.5753 often correspond 0.666667 -13.5753 pb-smt 0.8 -13.5755 lapurdian 0.8 -13.5755 reconcileconll, 0.8 -13.5755 underneath 0.8 -13.5755 whadvp 0.8 -13.5755 (bag-of-words) 0.8 -13.5755 ashish 0.555556 -13.5756 dfg 0.555556 -13.5756 eventlda 0.555556 -13.5756 bendersky 0.4375 -13.5757 projects, 0.322581 -13.5758 (2002). 0.666667 -13.576 write, 0.555556 -13.5761 0.180 0.555556 -13.5761 {n 0.625 -13.5763 madrid, 0.8 -13.5764 incidences 0.8 -13.5764 whom, 0.8 -13.5764 eik 0.4375 -13.5766 pyp 0.666667 -13.5767 niederhoffer 0.666667 -13.5767 dh 0.666667 -13.5767 recomputation 0.666667 -13.5767 perc 0.666667 -13.5767 ngd 0.666667 -13.5767 otherwise; 0.240741 -13.5772 concentrate 0.8 -13.5772 lmde 0.8 -13.5772 misclassified, 0.8 -13.5772 m’ 0.8 -13.5772 p(source 0.8 -13.5772 directly: 0.8 -13.5772 507 0.8 -13.5772 85.43 0.8 -13.5772 (203) 0.363636 -13.5773 (vogel 0.625 -13.5775 rice 0.8 -13.578 500, 0.8 -13.578 ray 0.8 -13.578 49.06 0.8 -13.578 gill 0.8 -13.578 predictive check 0.174757 -13.5781 computed as 0.666667 -13.5781 algorithm1 0.666667 -13.5781 fans 0.666667 -13.5781 encountered, 0.240741 -13.5786 states. 0.666667 -13.5788 782 0.4375 -13.5788 kul 0.8 -13.5788 bergsma, 0.8 -13.5788 opaque 0.555556 -13.579 lemma-discourse 0.666667 -13.5795 sws 0.625 -13.5795 table2: 0.625 -13.5795 receptor 0.555556 -13.5796 esel 0.555556 -13.5796 table4: 0.555556 -13.5796 0.872 0.8 -13.5797 twostage 0.8 -13.5797 ˆr 0.8 -13.5797 operation: 0.8 -13.5797 accomplishing 0.8 -13.5797 fake 0.8 -13.5797 23.67 0.8 -13.5797 αnull 0.0987654 -13.5798 shown that 0.322581 -13.5801 know, 0.5 -13.5801 domela 0.8 -13.5805 1750-1850 0.8 -13.5805 da, 0.8 -13.5805 mccc 0.8 -13.5805 rte1 0.8 -13.5805 user) 0.666667 -13.5808 figure 1 displays 0.666667 -13.5808 (please 0.666667 -13.5808 uni-gram 0.555556 -13.5813 (sense 0.411765 -13.5813 alia). 0.8 -13.5813 diversity: 0.8 -13.5813 monde 0.8 -13.5813 handannotated 0.8 -13.5813 sub-problem 0.8 -13.5813 wordnet synset. 0.8 -13.5813 another; 0.8 -13.5813 outlying 0.8 -13.5813 traverses 0.666667 -13.5815 sentences describing 0.666667 -13.5815 tails(e) 0.322581 -13.5818 extraction method 0.625 -13.582 subzone 0.8 -13.5821 93.71 0.8 -13.5821 clef 0.8 -13.5821 363 0.666667 -13.5822 banko, 0.666667 -13.5822 926 0.5 -13.5826 coreference features 0.625 -13.5827 0.280 0.666667 -13.5829 7in 0.555556 -13.583 semeval-2 0.8 -13.583 quoting 0.8 -13.583 whcr(t,h) 0.8 -13.583 patricia 0.625 -13.5833 linguistic dimensions 0.666667 -13.5836 2(c) 0.666667 -13.5836 spe 0.232143 -13.5838 feasibility 0.411765 -13.5838 q: 0.8 -13.5838 12the 0.8 -13.5838 conflating 0.8 -13.5838 haddow, 0.666667 -13.5843 pre-cluster 0.625 -13.5846 cha 0.8 -13.5846 label-preserving 0.8 -13.5846 expressives 0.8 -13.5846 litg 0.8 -13.5846 theirs, 0.8 -13.5846 mention pairs. 0.4375 -13.5847 alexandra 0.555556 -13.5847 == 0.555556 -13.5847 eh8 9ab, 0.666667 -13.585 745 0.666667 -13.585 mohamed 0.666667 -13.585 contours 0.282051 -13.5851 made. 0.202703 -13.5854 = arg 0.8 -13.5854 asr’s 0.8 -13.5854 24: 0.8 -13.5854 typing, 0.8 -13.5854 oqa&s 0.0848624 -13.5856 approach. 0.666667 -13.5857 profiling, 0.26087 -13.5858 resnik, 0.555556 -13.5859 smoothness 0.0236593 -13.5861 where 0.8 -13.5863 winners 0.8 -13.5863 “killed”, 0.666667 -13.5864 (roughly 0.666667 -13.5864 and rada 0.666667 -13.5864 roots, 0.380952 -13.5865 readymades 0.666667 -13.587 entity-origin 0.666667 -13.587 p(h) 0.8 -13.5871 (nmod 0.8 -13.5871 (split 0.411765 -13.5872 perplexity results 0.625 -13.5872 extrapolation 0.625 -13.5872 echihabi 0.555556 -13.5876 absolute, 0.666667 -13.5877 traditions 0.666667 -13.5877 (f, 0.666667 -13.5877 θv 0.8 -13.5879 50.15 0.8 -13.5879 fbk 0.4 -13.588 we consider only 0.666667 -13.5884 mistranslated 0.666667 -13.5884 subjective expressions 0.666667 -13.5884 particularly effective 0.666667 -13.5884 (other) 0.4375 -13.5888 they propose 0.555556 -13.5888 0.855 0.8 -13.5888 garoufi 0.8 -13.5888 (duc) 0.8 -13.5888 [that/?∅ 0.8 -13.5888 luckily, 0.666667 -13.5891 example-based 0.666667 -13.5891 ao 0.25 -13.5892 suggestions. 0.555556 -13.5893 chord 0.8 -13.5896 12,000 0.8 -13.5896 (whereas 0.8 -13.5896 “punctuation” 0.8 -13.5896 implicature 0.8 -13.5896 trs 0.8 -13.5896 obsolete 0.8 -13.5896 gwb66 0.666667 -13.5898 infrequently 0.411765 -13.5898 time step 0.4375 -13.5902 inference problem 0.4375 -13.5902 cornell 0.176471 -13.5903 then, we 0.236364 -13.5904 more useful 0.8 -13.5904 citation sentences, 0.8 -13.5904 39.56 0.8 -13.5904 prp, 0.8 -13.5904 α1 α2 perp. 0.8 -13.5904 c-model 0.8 -13.5904 rac-1 . 0.8 -13.5904 buch 0.666667 -13.5905 (c1 0.666667 -13.5905 5.2.4 0.666667 -13.5905 forum, 0.625 -13.5905 meronymy, 0.461538 -13.591 je 0.4375 -13.5911 we generalize 0.555556 -13.5911 andreevskaia 0.666667 -13.5912 irrational 0.555556 -13.5922 non-constituent 0.625 -13.5924 nguyen, 0.666667 -13.5926 gpa 0.666667 -13.5926 318 0.666667 -13.5926 laura 0.666667 -13.5933 similarity ratings 0.625 -13.5937 (sec) 0.058952 -13.5937 sentences. 0.666667 -13.594 ibm1hs 0.555556 -13.594 unigram model, 0.155039 -13.5942 is necessary 0.4375 -13.5943 damping 0.555556 -13.5945 subj, 0.555556 -13.5945 carlos 0.5 -13.5946 null vl, 0.666667 -13.5947 determinization 0.666667 -13.5947 affiliated 0.666667 -13.5947 hc 0.666667 -13.5947 forbid 0.666667 -13.5947 contextual, 0.4375 -13.5947 can play 0.555556 -13.5951 complementizer 0.555556 -13.5951 phil 0.666667 -13.5954 (nyt) 0.289474 -13.5954 programming. 0.666667 -13.5961 84.04 0.666667 -13.5961 arms 0.625 -13.5963 0.013 0.380952 -13.5965 (right) 0.162393 -13.5965 developed for 0.666667 -13.5968 wsj15 0.225806 -13.597 our results show 0.411765 -13.5971 strategic 0.666667 -13.5975 ptsg 0.666667 -13.5975 tverb 0.555556 -13.5975 2-fold 0.0649526 -13.5975 text. 0.294118 -13.5976 to mine 0.461538 -13.5977 quotient 0.363636 -13.5977 he has 0.5 -13.5977 word meanings 0.411765 -13.598 α2 0.666667 -13.5982 manifests 0.666667 -13.5982 3.91 0.110672 -13.5983 is computed 0.5 -13.5988 exercises, 0.666667 -13.5989 ∈r 0.666667 -13.5989 (wi) 0.666667 -13.5989 “it” 0.666667 -13.5996 takes values 0.625 -13.5996 interlanguage 0.282051 -13.5998 window. 0.0603949 -13.5999 models. 0.411765 -13.6001 fraser 0.666667 -13.6003 industry, 0.666667 -13.6003 he] 0.666667 -13.6003 npaper 0.666667 -13.6003 sentence-pair 0.666667 -13.6003 40.56 0.625 -13.6003 improve mt 0.5 -13.6004 sys2 0.666667 -13.601 961 0.411765 -13.601 our models, 0.380952 -13.6013 while being 0.625 -13.6016 pmid 0.30303 -13.6016 evoked 0.555556 -13.6021 siding 0.363636 -13.6023 time t 0.666667 -13.6031 vegematic 0.666667 -13.6031 (known 0.666667 -13.6031 non-word 0.4375 -13.6035 scfg rules. 0.461538 -13.6036 word-tree 0.461538 -13.6036 weblogs 0.625 -13.6036 tuples with 0.666667 -13.6038 eforts, 0.666667 -13.6038 reform 0.363636 -13.6042 summarization task 0.625 -13.6056 fever, 0.363636 -13.6057 jhu 0.5 -13.6062 shen, 0.0650407 -13.6067 side 0.625 -13.6069 possible sequences 0.625 -13.6069 fj1 0.346154 -13.607 variant. 0.5 -13.6072 (male) 0.666667 -13.6073 (sc) 0.666667 -13.608 pulled 0.666667 -13.608 defines 0.555556 -13.608 correlation analysis 0.666667 -13.6087 statistic, 0.666667 -13.6087 cluto 0.666667 -13.6094 a 0.4375 -13.6095 our approach differs 0.268293 -13.61 insightful 0.666667 -13.6101 (ex: 0.666667 -13.6101 chunker, 0.666667 -13.6101 (japanese 0.666667 -13.6101 gas 0.461538 -13.6104 dreyer 0.5 -13.6104 (11 0.0702875 -13.6109 sentence, 0.666667 -13.6109 d; 0.4375 -13.6109 onr 0.625 -13.6109 ct. 0.666667 -13.6116 trigger words, 0.666667 -13.6116 18.2% 0.666667 -13.6116 (5a) 0.666667 -13.6116 [adj] 0.275 -13.6116 since there are 0.5 -13.612 ndcg 0.555556 -13.6121 cancer. 0.285714 -13.6122 an attribute 0.4375 -13.6127 tagger (toutanova 0.461538 -13.6128 bits. 0.625 -13.6129 40%. 0.625 -13.6135 np↓ 0.625 -13.6135 substitutions. 0.5 -13.6136 first 100 0.0261826 -13.6137 approach 0.625 -13.6142 dj. 0.625 -13.6142 restricted. 0.625 -13.6142 inra 0.625 -13.6142 soviet 0.625 -13.6142 .01) 0.625 -13.6142 hungarian, 0.212121 -13.6143 concatenation 0.666667 -13.6144 “these 0.411765 -13.615 kernel function 0.197368 -13.6157 is implemented 0.666667 -13.6158 ilpscale 0.666667 -13.6158 happened, 0.666667 -13.6158 1990 0.666667 -13.6158 lui 0.666667 -13.6158 ωs 0.666667 -13.6165 25.2% 0.666667 -13.6165 f9 0.666667 -13.6165 non-unique 0.282051 -13.6166 figure 1 shows a 0.555556 -13.6169 r6 0.347826 -13.6173 match between 0.666667 -13.6173 versa) 0.141026 -13.6173 next, we 0.666667 -13.618 senator 0.5 -13.6184 better recall 0.363636 -13.6185 j-th 0.192771 -13.6186 mapping from 0.666667 -13.6187 yoshimasa 0.666667 -13.6187 reasoner 0.666667 -13.6187 (a,b) 0.666667 -13.6187 points: 0.555556 -13.6187 fishing 0.30303 -13.6189 when we use 0.153846 -13.6191 information on 0.666667 -13.6194 negative), 0.165138 -13.6197 area of 0.666667 -13.6201 imaginary 0.666667 -13.6201 map-em 0.461538 -13.6207 september, 0.388889 -13.6208 stopword 0.666667 -13.6208 what's 0.666667 -13.6215 continental 0.666667 -13.6215 affected, 0.666667 -13.6223 lisa 0.666667 -13.6223 hesitation 0.666667 -13.623 bagging 0.380952 -13.6232 then computed 0.555556 -13.6234 “apple” 0.666667 -13.6237 infobox, 0.666667 -13.6244 thou 0.666667 -13.6244 w3, 0.666667 -13.6244 important cue 0.461538 -13.6247 m4 0.310345 -13.6248 kullback-leibler 0.666667 -13.6251 side-by-side 0.666667 -13.6251 66.7% 0.183908 -13.6255 evidence for 0.0952381 -13.6256 baseline. 0.666667 -13.6259 announcements 0.666667 -13.6259 629 0.666667 -13.6259 a brief discussion 0.666667 -13.6259 difficult 0.215385 -13.6259 we begin 0.4375 -13.6264 dpmm, 0.666667 -13.6266 refused 0.666667 -13.6266 nssummationdisplay 0.666667 -13.6266 “they 0.666667 -13.6266 withthe 0.461538 -13.6267 dialogue structure 0.4375 -13.6268 were filtered 0.21875 -13.6268 treats 0.380952 -13.6269 employed. 0.666667 -13.628 rohit 0.555556 -13.6282 kruengkrai et 0.555556 -13.6288 vocabulary: 0.22807 -13.629 see table 0.388889 -13.6292 validation, 0.5 -13.6292 http: 0.5 -13.6292 pcls 0.555556 -13.6294 s3: 0.666667 -13.6302 i∈v 0.555556 -13.6306 concepts) 0.666667 -13.6309 npr 0.411765 -13.6309 dependency links 0.461538 -13.6312 non-translated 0.411765 -13.6314 dallas 0.5 -13.6314 qm 0.666667 -13.6316 •vp 0.555556 -13.6318 fp7 0.25 -13.6322 unsupervised word 0.5 -13.6324 henderson 0.555556 -13.633 bryan 0.275 -13.6335 (2010) use 0.388889 -13.6338 this approximation 0.666667 -13.6338 628 0.666667 -13.6338 norm, 0.268293 -13.634 arrive 0.142857 -13.6341 our goal is 0.411765 -13.6345 carreras, 0.148148 -13.6348 we introduce a 0.666667 -13.6352 (1998) proposed 0.666667 -13.6352 ace-owl 0.666667 -13.6352 lation 0.666667 -13.6352 428 0.5 -13.6352 acl-08: 0.555556 -13.6355 pos/chunking 0.666667 -13.636 lv 0.666667 -13.636 ask about 0.666667 -13.636 (nb) 0.555556 -13.6361 (acl 0.175258 -13.6366 expect that 0.666667 -13.6367 undirected, 0.555556 -13.6373 completely. 0.555556 -13.6373 lucent 0.333333 -13.6377 development test 0.666667 -13.6382 scatterplot 0.666667 -13.6382 0.130 0.666667 -13.6382 nal 0.666667 -13.6389 undirected-items 0.461538 -13.6393 (blaheta and 0.666667 -13.6396 epsilon1, 0.0491935 -13.6398 node 0.148148 -13.6398 our evaluation 0.555556 -13.6403 plos 0.5 -13.6406 are ready 0.666667 -13.6411 f(z) 0.666667 -13.6411 “event” 0.275 -13.6413 have: 0.555556 -13.6415 (university 0.124352 -13.6416 recently, 0.666667 -13.6418 6in 0.175258 -13.6419 errors are 0.363636 -13.6422 della 0.5 -13.6423 coindexed 0.666667 -13.6425 baron and 0.666667 -13.6425 53.15 0.666667 -13.6425 (perhaps 0.235294 -13.6425 think that 0.17 -13.6433 noun. 0.666667 -13.6433 articulated 0.666667 -13.6433 (id 0.666667 -13.6433 13.4% 0.666667 -13.6433 bicycle 0.294118 -13.6433 learning framework 0.347826 -13.6437 0.01). 0.666667 -13.644 respectively: 0.363636 -13.6442 mrs 0.461538 -13.6443 0.5 bleu 0.411765 -13.6444 different contexts. 0.666667 -13.6447 transform, 0.666667 -13.6447 ass 0.310345 -13.6448 interannotator 0.411765 -13.6449 search queries. 0.555556 -13.6452 (vi) 0.666667 -13.6455 80.75 0.666667 -13.6455 cover, 0.666667 -13.6455 subtreelet 0.666667 -13.6455 “so 0.411765 -13.6458 4 experiment 0.666667 -13.6462 (i,j,r) 0.363636 -13.6462 31, 0.666667 -13.6477 incidental 0.666667 -13.6484 (a): 0.461538 -13.6484 φ(x,y) 0.461538 -13.6489 christodoulopoulos 0.285714 -13.649 are normalized 0.666667 -13.6491 78.54 0.666667 -13.6491 normalising 0.666667 -13.6491 510 0.666667 -13.6491 zhejiang university, 0.555556 -13.6501 snover, 0.104089 -13.6502 general, 0.5 -13.6505 “gold 0.115044 -13.6508 document, 0.25 -13.6509 automatic speech 0.123711 -13.6514 0.277778 -13.6516  0.666667 -13.6528 invoking 0.666667 -13.6528 “5” 0.666667 -13.6528 causal event 0.666667 -13.6528 might, 0.555556 -13.6532 love, 0.186047 -13.6533 measures are 0.666667 -13.6536 γi 0.666667 -13.6536 gent, 0.555556 -13.6538 cambridge computer 0.555556 -13.6538 吼 0.388889 -13.6539 improvements were 0.666667 -13.6543 expected: 0.555556 -13.6544 prediction tasks 0.5 -13.6544 translators. 0.5 -13.6544 submission received: 0.162162 -13.6545 # of 0.666667 -13.655 defined above. 0.555556 -13.655 r∈c 1 -13.655 waited 1 -13.655 fom. 1 -13.655 kdd-t. 1 -13.655 jags@umiacs.umd.edu 1 -13.655 compressibility 1 -13.655 establishment. 1 -13.655 kar ‘do’ 1 -13.655 li, charles 1 -13.655 hsmq-learning 1 -13.655 convince 1 -13.655 accurate). 1 -13.655 (cr). 1 -13.655 pipelining 1 -13.655 .61 1 -13.655 izumi 1 -13.655 yeh, 1 -13.655 opca. 1 -13.655 klapaftis 1 -13.655 re. 1 -13.655 0.355±0.002 1 -13.655 (λ=0.2) 1 -13.655 non-locality 1 -13.655 1968. 1 -13.655 composition). 1 -13.655 neuter. 1 -13.655 marketing. 1 -13.655 precis. 1 -13.655 outlined. 1 -13.655 subfields. 1 -13.655 web-count 1 -13.655 ui. 1 -13.655 gim ´enez 1 -13.655 (davidov 1 -13.655 audio processing, 1 -13.655 olsson, 1 -13.655 newsgroups, 1 -13.655 93.13 1 -13.655 go off:s 1 -13.655 circles, 1 -13.655 ‘cp’ 1 -13.655 bmpr-ia. 1 -13.655 me? 1 -13.655 please contact 1 -13.655 part-of-speech) 1 -13.655 ungar 1 -13.655 licensed. 1 -13.655 schedules 1 -13.655 mneg. 1 -13.655 modeler 1 -13.655 one”. 1 -13.655 stuffy nose 1 -13.655 lucene. 1 -13.655 (kulkarni09). 1 -13.655 operating system, 1 -13.655 trouvain, 1 -13.655 alexandre dumas 1 -13.655 userdefined 1 -13.655 doddington 1 -13.655 guardians 1 -13.655 metagrammar, 1 -13.655 gramar. 1 -13.655 matrix). 1 -13.655 4-6 1 -13.655 bulgarian. 1 -13.655 kanzler chancellor 1 -13.655 carolyn penstein 1 -13.655 bombings 1 -13.655 trunk. 1 -13.655 2c +r2 1 -13.655 me too 1 -13.655 non-dominated 1 -13.655 properties (indexing 1 -13.655 snover, bonnie 1 -13.655 discoursereferents. 1 -13.655 highest-probability 1 -13.655 (dethlefs 1 -13.655 (carberry 1 -13.655 remains. 1 -13.655 hirschberg. 1 -13.655 cm,n 1 -13.655 tong. 1 -13.655 consistently. 1 -13.655 superclass 1 -13.655 component’s 1 -13.655 oxford: 1 -13.655 shape. 1 -13.655 nj. 1 -13.655 word-clustering 1 -13.655 comparativeness 1 -13.655 ribeiro-neto 1 -13.655 21.48 1 -13.655 genzel, 1 -13.655 mars@google.com 1 -13.655 2010a. 1 -13.655 λk. 1 -13.655 german/english 1 -13.655 selected). 1 -13.655 subfields 1 -13.655 females, 1 -13.655 (16). 1 -13.655 =∅. 1 -13.655 fairly, 1 -13.655 juho heimonen, 1 -13.655 iv. 1 -13.655 one? 1 -13.655 khalilov 1 -13.655 relabeling 1 -13.655 valentin@cs.stanford.edu 1 -13.655 superior. 1 -13.655 misses. 1 -13.655 (viz. 1 -13.655 npl {,} 1 -13.655 aac-specific 1 -13.655 parisien 1 -13.655 2011], 1 -13.655 chinese-to-english. 1 -13.655 time-consuming. 1 -13.655 accurate, compact, 1 -13.655 candlestick park”. 1 -13.655 emphasize, 1 -13.655 kriya 1 -13.655 americans. 1 -13.655 997 1 -13.655 [v+] w+] 1 -13.655 aims: 1 -13.655 definition). 1 -13.655 artificial inteligence, 1 -13.655 constructions). 1 -13.655 deadline 1 -13.655 simpler. 1 -13.655 obeying 1 -13.655 inadequate. 1 -13.655 1101 kitchawan 1 -13.655 0.0001), 1 -13.655 depriving 1 -13.655 representative, 1 -13.655 autogenerated 1 -13.655 guess. 1 -13.655 event)? 1 -13.655 scfgs. 1 -13.655 aggregated, 1 -13.655 φ(x,y), 1 -13.655 acoustics, 1 -13.655 systemt. 1 -13.655 error). 1 -13.655 ted pedersen 1 -13.655 sichuan 1 -13.655 c∈c(h) [1−p(t 1 -13.655 visualization. 1 -13.655 versed 1 -13.655 pentium 1 -13.655 hs, 1 -13.655 〈(x1,t1),...,(xn,tn)〉, 1 -13.655 monotonic. 1 -13.655 generations. 1 -13.655 brasil. 1 -13.655 〈0,=dv〉(20,21,23),〈1,v=>+ept〉(22)) 1 -13.655 0.733 0.29 1 -13.655 starts. 1 -13.655 reversely 1 -13.655 succession. 1 -13.655 updating. 1 -13.655 caches. 1 -13.655 sudheer kolachina 1 -13.655 catalogued 1 -13.655 itself), 1 -13.655 unaltered 1 -13.655 rights. 1 -13.655 dimple. 1 -13.655 base), 1 -13.655 [3..6] 1 -13.655 selectsize 1 -13.655 gajsek 1 -13.655 originating wiki 1 -13.655 lavie. 1 -13.655 initiative. 1 -13.655 streptomyces 1 -13.655 “shallow” 1 -13.655 trouble. 1 -13.655 open. 1 -13.655 regression). 1 -13.655 f.λx.f(x)∧ 1 -13.655 74.77% 1 -13.655 highly-ranked 1 -13.655 schemata. 1 -13.655 8-bit 1 -13.655 (lexical) 1 -13.655 sind billiger 1 -13.655 alignmenta. 1 -13.655 getötet. 1 -13.655 voices, 1 -13.655 (xie). 1 -13.655 civilization. 1 -13.655 chosen; 1 -13.655 4.2.2. 1 -13.655 requirements: 1 -13.655 dog”. 1 -13.655 attempted. 1 -13.655 reproducibility 1 -13.655 0.01% 1 -13.655 max-sum 1 -13.655 held-out. 1 -13.655 represents. 1 -13.655 70.83 1 -13.655 institutionen f¨or 1 -13.655 diner 1 -13.655 yannakakis 1 -13.655 scope-taking 1 -13.655 (oepen 1 -13.655 filler. 1 -13.655 wehrens, 1 -13.655 exceeded. 1 -13.655 monods. 1 -13.655 fabricated 1 -13.655 house.” 1 -13.655 temfull+f 1 -13.655 static. 1 -13.655 coarse. 1 -13.655 asked. 1 -13.655 deterministically, 1 -13.655 disambiguation). 1 -13.655 63.83 1 -13.655 cheat. 1 -13.655 (duh, 1 -13.655 tts. 1 -13.655 polysemy: 1 -13.655 adjoining, 1 -13.655 t-test). 1 -13.655 pseudoword. 1 -13.655 tidings 1 -13.655 prevails 1 -13.655 multi-headedness 1 -13.655 bars). 1 -13.655 sentences.) 1 -13.655 rgc6256/00e, 1 -13.655 marginals. 1 -13.655 (van durme 1 -13.655 ibm3 1 -13.655 parent-child-grandchild 1 -13.655 enrichment. 1 -13.655 hand-engineered 1 -13.655 passonneau. 1 -13.655 rarer. 1 -13.655 getnextbest 1 -13.655 (ned). 1 -13.655 julius 1 -13.655 unrealistic. 1 -13.655 beat. 1 -13.655 (9). 1 -13.655 (nps). 1 -13.655 causation. 1 -13.655 gemma boleda, 1 -13.655 lecture. 1 -13.655 95.67 1 -13.655 g.tec 1 -13.655 residence, 1 -13.655 decoder (vogel, 1 -13.655 system”. 1 -13.655 amsterdam. 1 -13.655 uks, 3t 1 -13.655 bies, 1 -13.655 ‘bad’. 1 -13.655 criteria) 1 -13.655 230,000 1 -13.655 gertjan 1 -13.655 ratnaparkhi 1 -13.655 (β) 1 -13.655 2.05 1 -13.655 manipulates 1 -13.655 bd, 1 -13.655 zhejiang, 1 -13.655 ascertaining 1 -13.655 anticipated. 1 -13.655 convention. 1 -13.655 faster, ... 1 -13.655 gesehen. 1 -13.655 prevention. 1 -13.655 “recommendations” 1 -13.655 arc-eager. 1 -13.655 10−6. 1 -13.655 our. 1 -13.655 hmt. 1 -13.655 macro/ 1 -13.655 pam, 1 -13.655 story.’ 1 -13.655 25.8% 1 -13.655 l←j}) 1 -13.655 marginally. 1 -13.655 gradually. 1 -13.655 bytes. 1 -13.655 martin.chodorow@hunter.cuny.edu 1 -13.655 conglomerate 1 -13.655 coreferents 1 -13.655 danish. 1 -13.655 hemoptysis 1 -13.655 � � 1 -13.655 growth. 1 -13.655 predictor’s 1 -13.655 implications, 1 -13.655 92–100, 1 -13.655 vided 1 -13.655 orlms 1 -13.655 diminished. 1 -13.655 alexandra balahur 1 -13.655 fscore. 1 -13.655 typicality 1 -13.655 λ-expressions. 1 -13.655 “justin bieber” 1 -13.655 fallen 1 -13.655 table). 1 -13.655 downloads, 1 -13.655 (rubin, 1 -13.655 bardblx−gsftbardbl2 +αtrace[(f−f0)tc1(f−f0)] 1 -13.655 statistic.pl 1 -13.655 lucian galescu, 1 -13.655 generics. 1 -13.655 switch. 1 -13.655 copa, 1 -13.655 lemmatizing 1 -13.655 svm-rank 1 -13.655 letter). 1 -13.655 33.91 1 -13.655 20-30 1 -13.655 infants. 1 -13.655 ??. 1 -13.655 no-rel. 1 -13.655 chodorow. 1 -13.655 3b, 1 -13.655 content-related 1 -13.655 orthogonality 1 -13.655 re-training. 1 -13.655 knight. 1 -13.655 struggles 1 -13.655 feat-iv, 1 -13.655 .\" 1 -13.655 ‘signatures’ 1 -13.655 45.89 1 -13.655 biographies. 1 -13.655 recall(%) prec.(%) 1 -13.655 l). 1 -13.655 (bhatt 1 -13.655 “cheapest” 1 -13.655 dynamics. 1 -13.655 ir-corpus 1 -13.655 fetcher 1 -13.655 80.83 1 -13.655 (scores 1 -13.655 centres. 1 -13.655 empty). 1 -13.655 disambig 1 -13.655 multi-faceted 1 -13.655 pji 1 -13.655 wex 1 -13.655 pairs? 1 -13.655 96.83 1 -13.655 (afrl), 1 -13.655 remains, 1 -13.655 converges, 1 -13.655 gap+ 1 -13.655 instantiated. 1 -13.655 supertagging. 1 -13.655 synthesiser. 1 -13.655 (mst). 1 -13.655 interpolated. 1 -13.655 wecan 1 -13.655 inconsistency. 1 -13.655 (2nd 1 -13.655 82.0% 1 -13.655 cut(.) 1 -13.655 dragos 1 -13.655 “law” 1 -13.655 clases. 1 -13.655 underspecification. 1 -13.655 venkatapathy 1 -13.655 penn-iii 1 -13.655 4.4.3 1 -13.655 shifters. 1 -13.655 92.93 1 -13.655 goave, 1 -13.655 [lite /–] 1 -13.655 oes, 1 -13.655 science-specific 1 -13.655 (bethard 1 -13.655 k-purity 1 -13.655 pave 1 -13.655 misinterpretation 1 -13.655 c(o) 1 -13.655 schmid’s 1 -13.655 confer 1 -13.655 mixed: 1 -13.655 kummerfeld 1 -13.655 (1.8%) 1 -13.655 test-takers 1 -13.655 technology”. 1 -13.655 join-semilattice 1 -13.655 yprime. 1 -13.655 —æj ¸@ 1 -13.655 subspaces. 1 -13.655 (3b). 1 -13.655 sub-types, 1 -13.655 uncertaintyadaptive 1 -13.655 adaptation? 1 -13.655 second-billed 1 -13.655 (crf). 1 -13.655 most-frequent 1 -13.655 ck. 1 -13.655 geneva geneva, 1 -13.655 possible) 1 -13.655 prisons. 1 -13.655 mans, 1 -13.655 extracted). 1 -13.655 epigenetic 1 -13.655 cov:parse 1 -13.655 kehler, 1 -13.655 applicative 1 -13.655 alchemy. 1 -13.655 formalization, 1 -13.655 self-score 1 -13.655 ?x, 1 -13.655 this). 1 -13.655 (gotoh 1 -13.655 hull. 1 -13.655 formosan, 1 -13.655 sub-space 1 -13.655 (jusczyk 1 -13.655 europarl 701,157 1 -13.655 5when 1 -13.655 (bellare 1 -13.655 sekine. 1 -13.655 [[yongkang zhou] 1 -13.655 discuss. 1 -13.655 notations. 1 -13.655 ified 1 -13.655 (left). 1 -13.655 dagan@cs.biu.ac.il 1 -13.655 wsj 777 1 -13.655 (freq) 1 -13.655 mae. 1 -13.655 name.first 1 -13.655 forrsooth’s 1 -13.655 dm. 1 -13.655 ££ordinal-description. 1 -13.655 expect. 1 -13.655 anonymized. 1 -13.655 barak09 1 -13.655 cs-svm. 1 -13.655 goodturing 1 -13.655 8%. 1 -13.655 sequential-alignment 1 -13.655 occurrences). 1 -13.655 to... 1 -13.655 centrality, 1 -13.655 raghavendra udupa 1 -13.655 (196). 1 -13.655 english→czech 1 -13.655 indirect, 1 -13.655 iis-05-34871. 1 -13.655 (3.5) 1 -13.655 co-citation, 1 -13.655 gland 1 -13.655 26. 1 -13.655 easing 1 -13.655 q(x). 1 -13.655 f.j. 1 -13.655 answer key, 1 -13.655 non-targets 1 -13.655 segw, 1 -13.655 le calvez 1 -13.655 count), 1 -13.655 visualise 1 -13.655 public! 1 -13.655 wonderland. 1 -13.655 (2007)), 1 -13.655 imported. 1 -13.655 decomposability. 1 -13.655 alternation. 1 -13.655 siddharthan, 1 -13.655 (15). 1 -13.655 bolded. 1 -13.655 drop. 1 -13.655 formation, 1 -13.655 [conceiver,intentional they] 1 -13.655 cohere 1 -13.655 aic, 1 -13.655 856 1 -13.655 vorfeld. 1 -13.655 (dligach 1 -13.655 a1). 1 -13.655 relative. 1 -13.655 imperatives. 1 -13.655 position) 1 -13.655 rooms. 1 -13.655 supported. 1 -13.655 s. bergler. 1 -13.655 tributions 1 -13.655 pairs” 1 -13.655 discocat 1 -13.655 parallelism, 1 -13.655 nugget. 1 -13.655 jrc 1 -13.655 conversions, 1 -13.655 set2. 1 -13.655 step), 1 -13.655 48.96 1 -13.655 israel. 1 -13.655 (san-segundo 1 -13.655 kmeans. 1 -13.655 “대선 후보들 1 -13.655 96.6% 1 -13.655 remediation. 1 -13.655 politician. 1 -13.655 kahane 1 -13.655 reverse) 1 -13.655 kendalls 1 -13.655 acronym’s 1 -13.655 certain. 1 -13.655 listbased 1 -13.655 nicknames. 1 -13.655 subordinating conjunctions 1 -13.655 back/ elab 1 -13.655 (mimacr). 1 -13.655 ne-pairs 1 -13.655 57,845 1 -13.655 mcnamee 1 -13.655 ´uplnˇe 1 -13.655 ostendorf. 1 -13.655 (0.5, 1 -13.655 carenini 1 -13.655 considers. 1 -13.655 canonicalized 1 -13.655 erg, 1 -13.655 (matuszek 1 -13.655 newsgroups. 1 -13.655 (.99) 1 -13.655 manual; 1 -13.655 afected 1 -13.655 posting. 1 -13.655 pct. 1 -13.655 # rules↑ 1 -13.655 non-regulation 1 -13.655 ecology 1 -13.655 a1, am-loc} 1 -13.655 annealing. 1 -13.655 boakye 1 -13.655 10-20% 1 -13.655 coercions, 1 -13.655 .). 1 -13.655 taskar, 1 -13.655 sfb 732. 1 -13.655 (model 2) 1 -13.655 args). 1 -13.655 hakkani-t¨ur, 1 -13.655 pairing. 1 -13.655 well-structured 1 -13.655 dominantly 1 -13.655 print [1/ 1 -13.655 ljubljana, slovenia 1 -13.655 6.3, 1 -13.655 combining. 1 -13.655 -. 1 -13.655 (lscript,s) 1 -13.655 “世界杯” 1 -13.655 32.91 1 -13.655 parsing). 1 -13.655 (women) kesariyaa6 1 -13.655 vm-ba 1 -13.655 moens. 1 -13.655 multi-criteria 1 -13.655 ear 1 -13.655 word-based. 1 -13.655 francisco. 1 -13.655 -0.74 1 -13.655 c(w31) 1 -13.655 obvious, 1 -13.655 both; 1 -13.655 poor) 1 -13.655 tendencies. 1 -13.655 schein’s 1 -13.655 experiments.5 1 -13.655 lengths). 1 -13.655 det. 1 -13.655 one-class-per-type 1 -13.655 promoting compound 1 -13.655 xue. 1 -13.655 discerning 1 -13.655 (d.ri 1 -13.655 extraction). 1 -13.655 data-driven, 1 -13.655 ended, 1 -13.655 (petrov, 1 -13.655 (pas¸ca 1 -13.655 pt(x). 1 -13.655 alpha. 1 -13.655 (fukumoto 1 -13.655 deverbal v/n 1 -13.655 corpus-dependent 1 -13.655 2011)). 1 -13.655 absence. 1 -13.655 al-youm al-sabe’ 1 -13.655 (follower-edges, maxent-seed) 1 -13.655 profile-based 1 -13.655 56.78 1 -13.655 (p<.05). 1 -13.655 (alfonseca 1 -13.655 hare 1 -13.655 englishchinese 1 -13.655 processors. 1 -13.655 timed 1 -13.655 iwslt’08 ct-ec 1 -13.655 derivatives. 1 -13.655 (freebase) 1 -13.655 1996a). 1 -13.655 emerged. 1 -13.655 unprocessed 1 -13.655 comprehension types, 1 -13.655 tricky, 1 -13.655 coster 1 -13.655 spannearquery 1 -13.655 >6. 1 -13.655 er. 1 -13.655 iiit hyderabad, 1 -13.655 eh8 9ab 1 -13.655 automaticallyderived 1 -13.655 o(n3), 1 -13.655 local. 1 -13.655 moat. 1 -13.655 microtexts, 1 -13.655 jn (2008) 1 -13.655 focus: 1 -13.655 cient 1 -13.655 productdisplay vi∈v 1 -13.655 to”. 1 -13.655 get. 1 -13.655 0null075) 1 -13.655 θuncovered 1 -13.655 apertium (forcada 1 -13.655 brants, 1 -13.655 66.91 1 -13.655 wn) 1 -13.655 hypothetical, historical, 1 -13.655 cssr. 1 -13.655 39-41. 1 -13.655 τφ. 1 -13.655 substep 1 -13.655 response) 1 -13.655 0.670 1 -13.655 summary). 1 -13.655 criticism. 1 -13.655 wildly 1 -13.655 aplications. 1 -13.655 abstractness. 1 -13.655 cancelled 1 -13.655 covariates. 1 -13.655 expert) 1 -13.655 reversals. 1 -13.655 uris 1 -13.655 nature; 1 -13.655 93.6% 1 -13.655 averagedifference 1 -13.655 nullinvite: ag(null)nullboy 1 -13.655 data.) 1 -13.655 attained. 1 -13.655 2008).1 1 -13.655 (a)(b), 1 -13.655 event-type 1 -13.655 misclassification. 1 -13.655 (steinberger 1 -13.655 arcstandard 1 -13.655 passages. 1 -13.655 in-car 1 -13.655 stand-in 1 -13.655 sacks 1 -13.655 progress). 1 -13.655 clause boundaries, 1 -13.655 socialturns 1 -13.655 callcut 1 -13.655 barr 1 -13.655 your talkative 1 -13.655 mpqa. 1 -13.655 1490 1 -13.655 unaddressed 1 -13.655 0.45, 1 -13.655 start. 1 -13.655 build. 1 -13.655 scribe 1 -13.655 glucksberg, 1 -13.655 happened. 1 -13.655 you”. 1 -13.655 5-gram. 1 -13.655 albnšklat2? 1 -13.655 [oov] 1 -13.655 teamscoringall(t, 1 -13.655 sentimatrix, 1 -13.655 lectures. 1 -13.655 imbalanced, 1 -13.655 bth 1 -13.655 819 1 -13.655 worlds: 1 -13.655 hits). 1 -13.655 90.1% 1 -13.655 manners. 1 -13.655 linking) 1 -13.655 thousands). 1 -13.655 yee seng 1 -13.655 beijing. 1 -13.655 ([v1,v2a]) 1 -13.655 “overlap”. 1 -13.655 “phrase 1 -13.655 gmds. 1 -13.655 “ad-hoc harmonic” 1 -13.655 ptt 1 -13.655 mle: mean 1 -13.655 adapted. 1 -13.655 cov. 1 -13.655 turktrain. 1 -13.655 ε. 1 -13.655 7.1. 1 -13.655 painstakingly 1 -13.655 contradictory. 1 -13.655 [v1,v2] 1 -13.655 clumps 1 -13.655 teaching. 1 -13.655 levi 1 -13.655 network). 1 -13.655 66.83 1 -13.655 drug router 1 -13.655 aroc 1 -13.655 http://rtw.ml.cmu.edu. 1 -13.655 vocabularies. 1 -13.655 cooccurrences. 1 -13.655 intruder. 1 -13.655 38.74 1 -13.655 distinctly. 1 -13.655 pseudo-word 1 -13.655 1986 1 -13.655 gradience 1 -13.655 voice spec. 1 -13.655 specificity: 1 -13.655 weasel, 1 -13.655 variablelength 1 -13.655 tine, 1 -13.655 concatenation, 1 -13.655 nouns/noun 1 -13.655 203). 1 -13.655 pbta). 1 -13.655 constraints), 1 -13.655 “naturally” 1 -13.655 rijsbergen, 1 -13.655 kelleher 1 -13.655 substructures. 1 -13.655 clwsd 1 -13.655 think? 1 -13.655 thresholding. 1 -13.655 problematic: 1 -13.655 word-by-word. 1 -13.655 aspell. 1 -13.655 kleinberg’s 1 -13.655 instance-instance graphs 1 -13.655 chinese– 1 -13.655 markovization. 1 -13.655 yad ‘memory’ 1 -13.655 naturallanguage 1 -13.655 variable’s 1 -13.655 contained. 1 -13.655 sense(i) 1 -13.655 61.84 1 -13.655 (moilanen 1 -13.655 hcr-dev 1 -13.655 same-genre 1 -13.655 ddtk. 1 -13.655 use? 1 -13.655 bottom-up. 1 -13.655 hand-crafting 1 -13.655 mediocre. 1 -13.655 hold’s. 1 -13.655 scopings, 1 -13.655 fuel 1 -13.655 protected 1 -13.655 te[pi]. 1 -13.655 90292, 1 -13.655 linguistically, 1 -13.655 kybot. 1 -13.655 basic. 1 -13.655 languagemodel. 1 -13.655 lesher 1 -13.655 〈e,t〉→ 1 -13.655 wong. 1 -13.655 url. 1 -13.655 gate (cunningham 1 -13.655 okay, 1 -13.655 humanities-based 1 -13.655 etim 1 -13.655 yashar 1 -13.655 universals. 1 -13.655 espresso: 1 -13.655 raw. 1 -13.655 turn-based 1 -13.655 many? 1 -13.655 72,628 listings 1 -13.655 7.2. 1 -13.655 sabine. 1 -13.655 context” 1 -13.655 1213 1 -13.655 excepting 1 -13.655 11,001 1 -13.655 influence 1 -13.655 translation-oriented 1 -13.655 40.89 1 -13.655 entropy-pruned 1 -13.655 ri). 1 -13.655 14). 1 -13.655 disk, 1 -13.655 bilal.m.sg=dat 1 -13.655 grouped. 1 -13.655 null(null))nullnullnull[null(null)nullwalk’(null)] 1 -13.655 (d2w) 1 -13.655 beyond. 1 -13.655 re-evaluating 1 -13.655 tracking, 1 -13.655 ychoi}@cs.stonybrook.edu 1 -13.655 > 1): 1 -13.655 relata. 1 -13.655 non-compositional, 1 -13.655 mt-08 1 -13.655 0.766 1 -13.655 parsert. 1 -13.655 anaphora-antecedent 1 -13.655 nl+nlp (pb) 1 -13.655 λx 1 -13.655 “not yet” 1 -13.655 andy. 1 -13.655 colin cherry 1 -13.655 (dp). 1 -13.655 bos. 1 -13.655 interchangeable. 1 -13.655 complementation. 1 -13.655 1.1~1.4 1 -13.655 6328 1 -13.655 4.4), 1 -13.655 25.9% 1 -13.655 ester ii 1 -13.655 75.14 1 -13.655 capped 1 -13.655 declaring 1 -13.655 rivi`ere 1 -13.655 (medelyan 1 -13.655 (kaminski) 1 -13.655 al) 1 -13.655 nullnullnull) 1 -13.655 yue.zhang@cl.cam.ac.uk 1 -13.655 1.3. 1 -13.655 delivery.n, 1 -13.655 van asch 1 -13.655 janice 1 -13.655 systemlevel 1 -13.655 wpred. 1 -13.655 grenager, 1 -13.655 s3h. 1 -13.655 trajectories. 1 -13.655 (paumier, 1 -13.655 (carletta 1 -13.655 2000a). 1 -13.655 savings, 1 -13.655 (slm) 1 -13.655 resources? 1 -13.655 parallelization 1 -13.655 arsenal. 1 -13.655 prepared. 1 -13.655 (allen, 1983) 1 -13.655 ordinals. 1 -13.655 korhonen computer 1 -13.655 4.2.3) 1 -13.655 testing). 1 -13.655 million). 1 -13.655 decide, 1 -13.655 #4: 1 -13.655 b2) 1 -13.655 underquantified 1 -13.655 dt) 1 -13.655 atempt 1 -13.655 church. 1 -13.655 bel(d, bel(c, 1 -13.655 (ekbal 1 -13.655 testsets. 1 -13.655 11,13be guaranteed 1 -13.655 bigger, 1 -13.655 (boutilier 1 -13.655 construction). 1 -13.655 saturday, 1 -13.655 engkoo. 1 -13.655 “was”. 1 -13.655 selector” 1 -13.655 recalls. 1 -13.655 searching. 1 -13.655 72.08 1 -13.655 criterion), 1 -13.655 apertium. 1 -13.655 keypoint 1 -13.655 metrics). 1 -13.655 maximum-entropy-based 1 -13.655 taging. 1 -13.655 88.89 1 -13.655 similarity/relatedness 1 -13.655 110–119, 1 -13.655 wc. 1 -13.655 02454 1 -13.655 §6. 1 -13.655 loglikelihood. 1 -13.655 labeled). 1 -13.655 something. 1 -13.655 racy 1 -13.655 love. 1 -13.655 sub-results. 1 -13.655 mobile. 1 -13.655 n-gramc 1 -13.655 worker-level 1 -13.655 preposition) 1 -13.655 “readability” 1 -13.655 out-of-english 1 -13.655 acquainted 1 -13.655 modelled. 1 -13.655 exponent. 1 -13.655 unfold 1 -13.655 718 1 -13.655 timebank, 1 -13.655 cohesiveness 1 -13.655 33%. 1 -13.655 acesible 1 -13.655 rouge: 1 -13.655 organelle, cell, 1 -13.655 haben. 1 -13.655 (svs) 1 -13.655 “fly” 1 -13.655 subzone. 1 -13.655 task p(%) 1 -13.655 (fhmms). 1 -13.655 darwish 1 -13.655 inparenthesis 1 -13.655 'j 1 -13.655 rudnicky. 1 -13.655 lowercasing 1 -13.655 terp (snover 1 -13.655 protein>. 1 -13.655 (2.7) 1 -13.655 (wn) 1 -13.655 franc¸ois yvon 1 -13.655 mutualinformation 1 -13.655 concordancer. 1 -13.655 buzek 1 -13.655 carpuat, 1 -13.655 instrument, 1 -13.655 1gb 1 -13.655 servers. 1 -13.655 quality.” 1 -13.655 unbalanced. 1 -13.655 left-hand-side (lhs) 1 -13.655 attached. 1 -13.655 clause” 1 -13.655 pre-lexical 1 -13.655 cu, 1 -13.655 (type) 1 -13.655 event), cause(core 1 -13.655 september. 1 -13.655 metonymies. 1 -13.655 sub-graphs 1 -13.655 ‘eval’ 1 -13.655 (clauset 1 -13.655 prominence. 1 -13.655 lower-cased. 1 -13.655 id’s, 1 -13.655 {among, between} 1 -13.655 river. 1 -13.655 friedman 1 -13.655 qg. 1 -13.655 determinernoun 1 -13.655 loadings 1 -13.655 (? 1 -13.655 (exponential 1 -13.655 [to] 1 -13.655 metrics.4 1 -13.655 63.08 1 -13.655 dysarthria. 1 -13.655 problems? 1 -13.655 vj. 1 -13.655 §3.1, 1 -13.655 barrett, 1 -13.655 alexandra balahur, 1 -13.655 slides. 1 -13.655 timelines, 1 -13.655 89%. 1 -13.655 latent. 1 -13.655 depression. 1 -13.655 sl. 1 -13.655 ltt. 1 -13.655 y∗(i) 1 -13.655 goldbej@eng.biu.ac.il 1 -13.655 (y). 1 -13.655 esp-game 1 -13.655 35. 1 -13.655 minqing hu 1 -13.655 tsao. 1 -13.655 turkers, 1 -13.655 (10%). 1 -13.655 orthogonal. 1 -13.655 data.4 1 -13.655 "n-grams" 1 -13.655 48.3 45.6 1 -13.655 nil, 1 -13.655 responsiveness. 1 -13.655 α-value 1 -13.655 “bin-2-lr” 1 -13.655 deontic 1 -13.655 dt. 1 -13.655 students; 1 -13.655 “eparl+nc+crawled2”. 1 -13.655 derivation’s 1 -13.655 ntcir8 1 -13.655 metrics kendall 1 -13.655 annotations). 1 -13.655 analyser. 1 -13.655 steinberger. 1 -13.655 q-query 1 -13.655 type-shifting 1 -13.655 transmission. 1 -13.655 makan tamir 1 -13.655 68.08 1 -13.655 cognition. 1 -13.655 frampton 1 -13.655 disambiguate. 1 -13.655 crfbased 1 -13.655 3.0). 1 -13.655 forgy 1 -13.655 (field 1 -13.655 (2003)), 1 -13.655 cyk-n 1 -13.655 applicable: 1 -13.655 (null1) 1 -13.655 disadvantages. 1 -13.655 semeval-2010. 1 -13.655 scisumm, 1 -13.655 zrcadla? 1 -13.655 kar ‘do’, 1 -13.655 arabic-english, 1 -13.655 interspeech, 1 -13.655 phrase-par. 1 -13.655 eng←eng1 1 -13.655 peak. 1 -13.655 downloaded. 1 -13.655 penalty). 1 -13.655 nearer 1 -13.655 alternative: 1 -13.655 boves, 1 -13.655 mary slapped 1 -13.655 schaefer 1 -13.655 igr. 1 -13.655 livejournal. 1 -13.655 (197). 1 -13.655 “prop. 1 -13.655 needless 1 -13.655 nonprofessional 1 -13.655 distraction 1 -13.655 adenopathy. 1 -13.655 sensible. 1 -13.655 collapsing, 1 -13.655 soroa 1 -13.655 /ada/, 1 -13.655 literally. 1 -13.655 tsang 1 -13.655 prolog. 1 -13.655 1p). 1 -13.655 non-salient 1 -13.655 negatively) 1 -13.655 hcr-train 1 -13.655 dependency-style 1 -13.655 neglecting 1 -13.655 y”. 1 -13.655 twsss. 1 -13.655 (dinarelli 1 -13.655 domainspecific. 1 -13.655 pseudo-references 1 -13.655 out-of-shortlist 1 -13.655 maite taboada, 1 -13.655 (firth, 1 -13.655 developers. 1 -13.655 gamer 1 -13.655 republican 1 -13.655 consent. 1 -13.655 -0.041 1 -13.655 non-ne 1 -13.655 “unknown” 1 -13.655 go! 1 -13.655 “related” 1 -13.655 k35k20 k36k20 1 -13.655 speculation/negation 1 -13.655 dna methylationand 1 -13.655 “span” 1 -13.655 equally; 1 -13.655 bodies. 1 -13.655 hereafter. 1 -13.655 sentence-wise 1 -13.655 1.4. 1 -13.655 seq. 1 -13.655 impossibility 1 -13.655 advantageous. 1 -13.655 ‘goodness’ 1 -13.655 “u” 1 -13.655 `arquez, 1 -13.655 (i. 1 -13.655 t-test) 1 -13.655 demonstrations 1 -13.655 underresourced 1 -13.655 model-1. 1 -13.655 martins,smith,and 1 -13.655 espresso. 1 -13.655 hersh 1 -13.655 routledge. 1 -13.655 . 1 -13.655 (acc.) 1 -13.655 xij. 1 -13.655 problem”. 1 -13.655 sentiment-labeled 1 -13.655 vocabulary), 1 -13.655 en-adj-nn 1 -13.655 deal. 1 -13.655 d'melo, witherspon, 1 -13.655 ems 1 -13.655 .uk 1 -13.655 caption. 1 -13.655 (gurevich 1 -13.655 score? 1 -13.655 the”. 1 -13.655 function/auxiliary 1 -13.655 (5.0%) 1 -13.655 nthu, 1 -13.655 f). 1 -13.655 deruyter 1 -13.655 promoter. 1 -13.655 matches). 1 -13.655 ...). 1 -13.655 taxa, 1 -13.655 permutated 1 -13.655 plugged 1 -13.655 modest. 1 -13.655 (§5) 1 -13.655 il 61801 1 -13.655 1930). 1 -13.655 2(b). 1 -13.655 yk. 1 -13.655 (hildebrand 1 -13.655 clearer. 1 -13.655 sq. 1 -13.655 model1. 1 -13.655 ha. 1 -13.655 mooney@cs.utexas.edu 1 -13.655 deceptive. 1 -13.655 77.0% 1 -13.655 (6.3) 1 -13.655 gnppa. 1 -13.655 singulars/plurals, 1 -13.655 92%. 1 -13.655 dipper, 1 -13.655 symptom. 1 -13.655 (jing 1 -13.655 icassp, 1 -13.655 greater, 1 -13.655 gradients. 1 -13.655 bench. 1 -13.655 rijke. 1 -13.655 vbs 1 -13.655 murachver 1 -13.655 offtopic 1 -13.655 sumarizes 1 -13.655 hold; 1 -13.655 2 mentiondetection 1 -13.655 (n-grams) 1 -13.655 “external” 1 -13.655 [b1,b2] 1 -13.655 39.00 1 -13.655 1here, 1 -13.655 possessive. 1 -13.655 misunderstood. 1 -13.655 metamap. 1 -13.655 linguistics, ghent 1 -13.655 (-1.2, 1 -13.655 5.4.3 1 -13.655 syllable-level 1 -13.655 infinity. 1 -13.655 ?type1 1 -13.655 vino. 1 -13.655 nichols 1 -13.655 senmpqa. 1 -13.655 gracefully. 1 -13.655 reasons). 1 -13.655 1990s. 1 -13.655 non-identified 1 -13.655 inkb 1 -13.655 api1 1 -13.655 (cu) 1 -13.655 complexes, 1 -13.655 (erdmann 1 -13.655 asymmetrical 1 -13.655 answer synsets. 1 -13.655 blast, 1 -13.655 inflate 1 -13.655 structured. 1 -13.655 span [b,c] 1 -13.655 neighbors’ 1 -13.655 c.s. trp 1 -13.655 (typed) 1 -13.655 dictating 1 -13.655 dataset’s 1 -13.655 receive. 1 -13.655 feedback? 1 -13.655 subscripted 1 -13.655 .98 rsemh 1 -13.655 noun∈sn. 1 -13.655 l1; 1 -13.655 doubly-represented 1 -13.655 disease) 1 -13.655 identities. 1 -13.655 homepage. 1 -13.655 33.3%, 1 -13.655 chunk) 1 -13.655 aer. 1 -13.655 left-to-right. 1 -13.655 gesture. 1 -13.655 (beam width) 1 -13.655 1950). 1 -13.655 web1. 1 -13.655 munteanu, 1 -13.655 arthropods 1 -13.655 hfl .348 1 -13.655 wordpair 1 -13.655 interactively. 1 -13.655 logicaly 1 -13.655 substructure. 1 -13.655 5-grams. 1 -13.655 (0.07) 1 -13.655 profile 1 -13.655 28.0% 1 -13.655 krey`ol, 1 -13.655 boprime. 1 -13.655 english-finnish 1 -13.655 “consensus” 1 -13.655 pcfg-based 1 -13.655 class); 1 -13.655 overwhelms 1 -13.655 (§5), 1 -13.655 12note 1 -13.655 moderating 1 -13.655 crowd-sourced 1 -13.655 words/targets 1 -13.655 syntax+aug. 1 -13.655 sócrates (c1) 1 -13.655 implicated 1 -13.655 encouragement. 1 -13.655 n0. 1 -13.655 2.52 1 -13.655 2:(1)). 1 -13.655 value=1.0) 1 -13.655 pers.ind 1 -13.655  in  bread 1 -13.655 montgomery 1 -13.655 rivals. 1 -13.655 dataset1. 1 -13.655 subparts, 1 -13.655 rater’s 1 -13.655 lsi-based 1 -13.655 1-ter. 1 -13.655 misleading, 1 -13.655 r© 1 -13.655 bangjiao 1 -13.655 ϕ. 1 -13.655 2nd resp. 1 -13.655 bicycles). 1 -13.655 ,...,b 1 -13.655 mammal). 1 -13.655 damage. 1 -13.655 greeting. 1 -13.655 classifiers’ 1 -13.655 circumplex. 1 -13.655 sounds. 1 -13.655 resorts 1 -13.655 workshops, 1 -13.655 (burek 1 -13.655 cats... 1 -13.655 triangulation, 1 -13.655 rubrics 1 -13.655 groen 1 -13.655 (sch¨utze 1 -13.655 ja. 1 -13.655 spinning 1 -13.655 raghavendra udupa, 1 -13.655 graph), 1 -13.655 ongoing debate 1 -13.655 type’s 1 -13.655 throughput, 1 -13.655 consti. 1 -13.655 scorers. 1 -13.655 mountainbikes 1 -13.655 founded(jobs, apple) 1 -13.655 -100 1 -13.655 trofi (trope 1 -13.655 (194). 1 -13.655 ties). 1 -13.655 elusive 1 -13.655 using. 1 -13.655 overgeneration, 1 -13.655 malik, 1 -13.655 certainty. 1 -13.655 back-translation: ‘charter 1 -13.655 homonyms, 1 -13.655 portal. 1 -13.655 ‘personal translator’ 1 -13.655 non-professionals 1 -13.655 light) 1 -13.655 f′ 1 -13.655 etrees, 1 -13.655 dki ,dvi 1 -13.655 4-ref 1 -13.655 weld’s 1 -13.655 komodakis 1 -13.655 investment, 1 -13.655 relatumfrequency 1 -13.655 intonation. 1 -13.655 m-step. 1 -13.655 non-cooperative 1 -13.655 string). 1 -13.655 avatar. 1 -13.655 kevin dela 1 -13.655 58.8%. 1 -13.655 epi. 1 -13.655 7m 1 -13.655 disagrement 1 -13.655 (fagni 1 -13.655 (rich 1 -13.655 l-match. 1 -13.655 sfst-based 1 -13.655 growing. 1 -13.655 inter-pattern 1 -13.655 ], [np 1 -13.655 inline-ranker 1 -13.655 .04) 1 -13.655 genia. 1 -13.655 flawed. 1 -13.655 gravano. 1 -13.655 initialized, 1 -13.655 (wallach 1 -13.655 propose. 1 -13.655 section5. 1 -13.655 context-clustering 1 -13.655 negri@fbk.eu 1 -13.655 mansour 1 -13.655 verbosity, 1 -13.655 56. 1 -13.655 interested. 1 -13.655 scotland. 1 -13.655 obj, obj2 1 -13.655 pst. 1 -13.655 alopez@cs.jhu.edu 1 -13.655 them.) 1 -13.655 complex) 1 -13.655 (dai 1 -13.655 wmt11, 1 -13.655 whatsoever. 1 -13.655 wprime2 1 -13.655 stock. 1 -13.655 factorization. 1 -13.655 john benjamins, 1 -13.655 drawback, 1 -13.655 spot. 1 -13.655 mwes). 1 -13.655 rolespecific 1 -13.655 ,the 1 -13.655 in”. 1 -13.655 d4, 1 -13.655 transitive chains 1 -13.655 synonym-like 1 -13.655 (bin 1 -13.655 ravichandran. 1 -13.655 morphological, syntactic, 1 -13.655 initn 1 -13.655 tong, 1 -13.655 rt3 1 -13.655 0.6) 1 -13.655 -0.38 1 -13.655 c.d. 1 -13.655 williamson, 1 -13.655 mention). 1 -13.655 exponentially. 1 -13.655 ν. 1 -13.655 bar-hillel 1 -13.655 (mbr) 1 -13.655 clac and 1 -13.655 copies. 1 -13.655 (melli 1 -13.655 suspends 1 -13.655 reallife 1 -13.655 distractor. 1 -13.655 (deb 1 -13.655 s-patterns, 1 -13.655 pf-positive f-score(%), 1 -13.655 in§4. 1 -13.655 conditional, 1 -13.655 salience. 1 -13.655 opinionatedness, 1 -13.655 wall. 1 -13.655 cough. 1 -13.655 blast. 1 -13.655 (3a). 1 -13.655 (pevzner 1 -13.655 promotion) 1 -13.655 accelerating 1 -13.655 (f1). 1 -13.655 meaning” 1 -13.655 28%. 1 -13.655 28.89 1 -13.655 complimentary 1 -13.655 pl. 1 -13.655 english-spanish. 1 -13.655 unnoticed 1 -13.655 treewidth. 1 -13.655 pps, 1 -13.655 lda−300m−0.1−0.1 lda−200m−0.1−0.01 1 -13.655 multinomial. 1 -13.655 decomposability 1 -13.655 suite. 1 -13.655 summarizer. 1 -13.655 similar” 1 -13.655 entity-based ceaf 1 -13.655 [..] 1 -13.655 downhillsimplex 1 -13.655 submodules. 1 -13.655 fpr. 1 -13.655 meronyms. 1 -13.655 integrity. 1 -13.655 æj. 1 -13.655 misunderstandings. 1 -13.655 diplomat 1 -13.655 (napoles 1 -13.655 ��'wt�luk�d�. 1 -13.655 she has never 1 -13.655 (191). 1 -13.655 ph.d. 1 -13.655 2010c) 1 -13.655 (orkin 1 -13.655 markedly. 1 -13.655 german-accented 1 -13.655 format.1 1 -13.655 da). 1 -13.655 en–de 1 -13.655 corazza 1 -13.655 (+1.3) 1 -13.655 lmll 1 -13.655 microand macro- 1 -13.655 m3+stanford 1 -13.655 mst. 1 -13.655 wordnet.pt, 1 -13.655 select. 1 -13.655 cleared. 1 -13.655 sfp +bow 1 -13.655 incomplete, 1 -13.655 clda. 1 -13.655 (dis)similarities 1 -13.655 inits, 1 -13.655 set.) 1 -13.655 task1. 1 -13.655 transversal 1 -13.655 therein). 1 -13.655 st’09, 1 -13.655 paskin, 1 -13.655 14.28 1 -13.655 order), 1 -13.655 well-formed. 1 -13.655 participating. 1 -13.655 forms). 1 -13.655 y-axis. 1 -13.655 maise, 1 -13.655 respectively—connectingthe 1 -13.655 cascade. 1 -13.655 well-documented 1 -13.655 ca 90094 1 -13.655 (s). 1 -13.655 arsl. 1 -13.655 ntc-7 1 -13.655 (eqs. 1 -13.655 svms). 1 -13.655 discourse: 1 -13.655 parse-tree 1 -13.655 med. 1 -13.655 instinctively 1 -13.655 dual-task 1 -13.655 directions). 1 -13.655 v, entitle] 1 -13.655 197). 1 -13.655 respiratory symptoms, 1 -13.655 8-grams 1 -13.655 teacher. 1 -13.655 files: 1 -13.655 scarcity. 1 -13.655 +1. 1 -13.655 algorithms) 1 -13.655 clas. 1 -13.655 infix. 1 -13.655 (zechner 1 -13.655 (gs) 1 -13.655 name). 1 -13.655 rater. 1 -13.655 durme, 1 -13.655 bing xiang, 1 -13.655 multi-headed 1 -13.655 (schmid 1 -13.655 excising 1 -13.655 brackets, 1 -13.655 .280 1 -13.655 standards, total 1 -13.655 (scfg). 1 -13.655 mityc 1 -13.655 hypernym. 1 -13.655 decode. 1 -13.655 mochihashi 1 -13.655 giver 1 -13.655 feelings. 1 -13.655 shorthand. 1 -13.655 (liscombe 1 -13.655 ∏ w∈q 1 -13.655 0.92. 1 -13.655 demographics, 1 -13.655 “sales 1 -13.655 words w/o. 1 -13.655 german-english. 1 -13.655 variables). 1 -13.655 set.3 1 -13.655 junk. 1 -13.655 milne, 1 -13.655 scheible, 1 -13.655 (qg) 1 -13.655 d5. 1 -13.655 (.95) 1 -13.655 maxlanguage. 1 -13.655 automaticallyextracted 1 -13.655 kldivergence 1 -13.655 (giza++) 1 -13.655 westbrook 1 -13.655 unlemmatised 1 -13.655 verification. 1 -13.655 adjoining. 1 -13.655 observations; 1 -13.655 (w); 1 -13.655 hardware. 1 -13.655 pausing 1 -13.655 nonmwes. 1 -13.655 2.0. 1 -13.655 ji. 1 -13.655 (supervised) 1 -13.655 effect) 1 -13.655 popularly 1 -13.655 mle. 1 -13.655 (arabic) 1 -13.655 culties inherent 1 -13.655 …]. 1 -13.655 paulo portas 1 -13.655 50.00% 1 -13.655 pragmatics. 1 -13.655 parses) 1 -13.655 ,...,t 1 -13.655 pickering 1 -13.655 n.pron.indef 1 -13.655 74.29% 1 -13.655 organization). 1 -13.655 burgess, 1 -13.655 subrequest. 1 -13.655 apposition. 1 -13.655 (are) 1 -13.655 (2011a) 1 -13.655 grammars). 1 -13.655 "  1 -13.655 101–109, 1 -13.655 halts 1 -13.655 ml08 1 -13.655 alphabets. 1 -13.655 negative-opinionated 1 -13.655 controlled, 1 -13.655 expressions) 1 -13.655 turney. 1 -13.655 discussions (aawd) 1 -13.655 (itg) 1 -13.655 post-evaluation 1 -13.655 3http://www.statmt.org/moses/ 1 -13.655 a1a3a2 a3a0a2 1 -13.655 pre-test score) 1 -13.655 25.86 1 -13.655 chalenge, 1 -13.655 phrasings. 1 -13.655 domain-matched 1 -13.655 upon. 1 -13.655 exclusively, 1 -13.655 slaves. 1 -13.655 il. 1 -13.655 pearson’s spearman’scorrelation 1 -13.655 osu 1 -13.655 mattress.” 1 -13.655 “ground truth” 1 -13.655 speculations, 1 -13.655 rte. 1 -13.655 think?” 1 -13.655 per semeval2010 1 -13.655 meier, 1 -13.655 np-hard. 1 -13.655 under-investigated 1 -13.655 re-running 1 -13.655 costa-juss`a, 1 -13.655 (yang, 1 -13.655 chicago: 1 -13.655 2(a). 1 -13.655 queryproduct 1 -13.655 (shorter 1 -13.655 defer this 1 -13.655 resolution). 1 -13.655 pulman. 1 -13.655 pmultiple 1 -13.655 name? 1 -13.655 +* 1 -13.655 extrinsically. 1 -13.655 eventmine 1 -13.655 kenneth church 1 -13.655 63.00 1 -13.655 allophony. 1 -13.655 skt 1 -13.655 substitute. 1 -13.655 ying-kuei 1 -13.655 implicative 1 -13.655 fairy tale 1 -13.655 1http://www.hcrc.ed.ac.uk/comic/ 1 -13.655 house.’ 1 -13.655 insensitive. 1 -13.655 j.-r. 1 -13.655 t1,...,tn 1 -13.655 tees 1 -13.655 syntactic. 1 -13.655 word/state 1 -13.655 (1.3) 1 -13.655 minimization. 1 -13.655 parsim. 1 -13.655 partially-complete 1 -13.655 merger. 1 -13.655 strengths, 1 -13.655 lippincott 1 -13.655 evaluee. 1 -13.655 (wmt11). 1 -13.655 ‘seen 1 -13.655 ncls. 1 -13.655 succesfully 1 -13.655 subcategory 1 -13.655 word-predictor. 1 -13.655 electromag. 1 -13.655 balkanet 1 -13.655 work), 1 -13.655 dk. 1 -13.655 (ddt). 1 -13.655 40.84 1 -13.655 0.29. 1 -13.655 opencyc 1 -13.655 q-type: 1 -13.655 +ept〉(0022222),〈0,=dv〉(20,21,23),〈1,v=>+ept〉(22)) 1 -13.655 debugging. 1 -13.655 cordinate 1 -13.655 postposition.no 1 -13.655 cij. 1 -13.655 completely) 1 -13.655 1991. 1 -13.655 f(si) 1 -13.655 syntactic-based 1 -13.655 harmonize 1 -13.655 2001)). 1 -13.655 inquiry. 1 -13.655 query-aggr. 1 -13.655 goldberger 1 -13.655 histogram. 1 -13.655 examination. 1 -13.655 .05. 1 -13.655 constraint). 1 -13.655   1 -13.655 copyright. 1 -13.655 81.57% 1 -13.655 grass. 1 -13.655 choice! 1 -13.655 bi-parsing. 1 -13.655 strictness 1 -13.655 c1. 1 -13.655 pitmanyor 1 -13.655 w0. 1 -13.655 mirroring 1 -13.655 +. 1 -13.655 template schemas, 1 -13.655 st, st〉→〈[s 1 -13.655 erased 1 -13.655 (〈creator〉v ∨〈genowner〉v)))) 1 -13.655 vogel. 1 -13.655 (mcintosh 1 -13.655 ben-gurion 1 -13.655 (1989), 1 -13.655 agrawal 1 -13.655 71%. 1 -13.655 “relative 1 -13.655 halved 1 -13.655 gu+hv 1 -13.655 bi-grams. 1 -13.655 linearly. 1 -13.655 source-conditional 1 -13.655 φ(s), 1 -13.655 (ldc2003e14). 1 -13.655 rlevy@ucsd.edu 1 -13.655 provoke 1 -13.655 female). 1 -13.655 related’. 1 -13.655 bi-text. 1 -13.655 (hr) 1 -13.655 (levin 1 -13.655 downside, 1 -13.655 randomly). 1 -13.655 that-choice 1 -13.655 (flat)” 1 -13.655 8.2% 1 -13.655 centroid: 1 -13.655 gb, 1 -13.655 attendant 1 -13.655 interlinear glosser 1 -13.655 hapax, 1 -13.655 “and”. 1 -13.655 developer’s 1 -13.655 (tsuboi et 1 -13.655 conceived bsoa, 1 -13.655 mel-cepstrum 1 -13.655 arbitrarily, 1 -13.655 regina}@csail.mit.edu 1 -13.655 temperley, 1 -13.655 book”. 1 -13.655 cprimen 1 -13.655 montecarlo 1 -13.655 randomized. 1 -13.655 eid=''e1'' ... 1 -13.655 d[i] 1 -13.655 pichuan 1 -13.655 93.7% 1 -13.655 (mds) 1 -13.655 intersection: 1 -13.655 geobase 1 -13.655 realizes 1 -13.655 (unknown) 1 -13.655 2006)) 1 -13.655 myspace. 1 -13.655 intersective 1 -13.655 lj. 1 -13.655 (tr). 1 -13.655 walk, do 1 -13.655 context-free. 1 -13.655 rampant 1 -13.655 elicited imitation 1 -13.655 intelligent. 1 -13.655 over”. 1 -13.655 compact. 1 -13.655 non-response. 1 -13.655 (dca) 1 -13.655 fires. 1 -13.655 lowercased. 1 -13.655 duplications, 1 -13.655 impairments, 1 -13.655 0.531 1 -13.655 kukich, 1 -13.655 ally 1 -13.655 mary’s 1 -13.655 opportunity, 1 -13.655 subdomains. 1 -13.655 cant 1 -13.655 neuhold, 1 -13.655 mochales 1 -13.655 occupational 1 -13.655 (dear daughter), 1 -13.655 ji;jj 1 -13.655 determiner “a” 1 -13.655 actor. 1 -13.655 geometrically-motivated 1 -13.655 evoke. 1 -13.655 excellence. 1 -13.655 #iis-0811974. 1 -13.655 bnc) 1 -13.655 27.0% 1 -13.655 distinctive, 1 -13.655 updatecandidatearcs(bestarc, candidatearcs, 1 -13.655 numerator. 1 -13.655 (-1.5, 1 -13.655 11/2007. 1 -13.655 null2null(man’(null), 1 -13.655 phonetically-encoded 1 -13.655 field). 1 -13.655 hearer’s 1 -13.655 9k 1 -13.655 shown: 1 -13.655 (ve) 1 -13.655 (n-gram 1 -13.655 (gibbs, 1 -13.655 phylogeny. 1 -13.655 subclass. 1 -13.655 38.18% 1 -13.655 hiero’s 1 -13.655 general; 1 -13.655 hundreds, 1 -13.655 ferres 1 -13.655 m13 9pl, 1 -13.655 urbana-champaign urbana, 1 -13.655 r(e). 1 -13.655 clasifiers. 1 -13.655 ¨ofver svenska 1 -13.655 baseline+both 1 -13.655 began. 1 -13.655 51.73 1 -13.655 entries). 1 -13.655 (s)cdp 1 -13.655 (v´eronis, 1 -13.655 time-based 1 -13.655 process? 1 -13.655 fk-bestθ (xi) 1 -13.655 rmap. 1 -13.655 jagged 1 -13.655 overlap-based 1 -13.655 ellipses. 1 -13.655 extents. 1 -13.655 word/sub-word 1 -13.655 stems). 1 -13.655 valentina bartalesi 1 -13.655 its. 1 -13.655 ;ij ;a/ 1 -13.655 (api). 1 -13.655 birke-sarkar 1 -13.655 sdp. 1 -13.655 nes, 1 -13.655 gfqs 1 -13.655 dialogues). 1 -13.655 193). 1 -13.655 ballgame 1 -13.655 blywn (eng. 1 -13.655 cse, university 1 -13.655 predicators, 1 -13.655 co¨occurring 1 -13.655 feat. 1 -13.655 mio. 1 -13.655 worsens 1 -13.655 wikinyt-107. 1 -13.655 (milne08) 1 -13.655 imageable 1 -13.655 78.74 1 -13.655 leass, 1 -13.655 879,807 1 -13.655 simple) 1 -13.655 discrepancy, 1 -13.655 accuracy/speed tradeoff 1 -13.655 suppressed. 1 -13.655 lately, 1 -13.655 herbst. 1 -13.655 = [d+n1 1 -13.655 =2. 1 -13.655 71.84 1 -13.655 brittleness 1 -13.655 impl. 1 -13.655 lis. 1 -13.655 data-set. 1 -13.655 editorials, 1 -13.655 stacking-based 1 -13.655 (federmann 1 -13.655 ??, 1 -13.655 explosion, 1 -13.655 terminates, 1 -13.655 sparsity-inducing 1 -13.655 discussionand 1 -13.655 ,...,f 1 -13.655 equaled 1 -13.655 1995; chen 1 -13.655 sub-categories 1 -13.655 stanzas. 1 -13.655 domaineditor 1 -13.655 long-tailed 1 -13.655 verbocean. 1 -13.655 12%. 1 -13.655 long-form. 1 -13.655 rickford, 1 -13.655 gibbs, 1 -13.655 co-referent. 1 -13.655 78–86, 1 -13.655 ssel 1 -13.655 preprocessed. 1 -13.655 fillmore’s 1 -13.655 besser 1 -13.655 memorising 1 -13.655 descent (sgd) 1 -13.655 efectively 1 -13.655 toplevel 1 -13.655 (ps). 1 -13.655 kinds. 1 -13.655 (2.5) 1 -13.655 mrl 1 -13.655 uppsala, sweden 1 -13.655 0.65. 1 -13.655 proper. 1 -13.655 sfr 1 -13.655 daughter. 1 -13.655 hyperplane, 1 -13.655 zhiyi 1 -13.655 papers). 1 -13.655 sporleder, 1 -13.655 goodnes-of-fit 1 -13.655 slu. 1 -13.655 consecutive +proper 1 -13.655 daughters. 1 -13.655 74.16% 1 -13.655 send-pst-dec 1 -13.655 synthesizer. 1 -13.655 invalid. 1 -13.655 telephony 1 -13.655 hgresolver 1 -13.655 inducer. 1 -13.655 (hardmeier 1 -13.655 likert scale. 1 -13.655 woepos, 1 -13.655 (task b) 1 -13.655 2-tailed 1 -13.655 amount. 1 -13.655 conceivable. 1 -13.655 turkdev. 1 -13.655 speech), 1 -13.655 (mcclosky, 1 -13.655 diversifying 1 -13.655 resynthesized 1 -13.655 open-test 1 -13.655 jj, jjr, 1 -13.655 automated. 1 -13.655 medlda. 1 -13.655 lezen. 1 -13.655 montoyo. 1 -13.655 consumer’s 1 -13.655 (ben-david 1 -13.655 (huhns 1 -13.655 literate 1 -13.655 perfect). 1 -13.655 firms. 1 -13.655 distant, 1 -13.655 maxdifference 1 -13.655 rejection. 1 -13.655 action”. 1 -13.655 governors. 1 -13.655 provides. 1 -13.655 p≤k(w) = 1 -13.655 exiting 1 -13.655 words! 1 -13.655 experimental psychology: 1 -13.655 f-score). 1 -13.655 acquired, 1 -13.655 nowson 1 -13.655 (4.7) 1 -13.655 gt. 1 -13.655 wasn’t until 1 -13.655 deployment. 1 -13.655 sortal, 1 -13.655 ! 1 -13.655 tv234@cam.ac.uk 1 -13.655 aql. 1 -13.655 newspapers. 1 -13.655 etzioni. 1 -13.655 (e→f) 1 -13.655 85.19 84.60 1 -13.655 ∗f(⃗w,⃗xj(1)qi −⃗xj(2)qi 1 -13.655 27. 1 -13.655 gathered. 1 -13.655 weiss, 1 -13.655 51.5 ( 1 -13.655 restrictive. 1 -13.655 notational, 1 -13.655 ces. 1 -13.655 (f→e) 1 -13.655 tasks). 1 -13.655 interventions 1 -13.655 kazawa, 1 -13.655 section 5 concludes. 1 -13.655 pieraccini, 1 -13.655 4.1); 1 -13.655 (hearst, 1997) 1 -13.655 yaser arafat 1 -13.655 unto 1 -13.655 toolkit4 1 -13.655 concept-to-text 1 -13.655 trainingdata. 1 -13.655 pir2, 1 -13.655 *although 1 -13.655 re-tweet 1 -13.655 #hold/0 1 -13.655 rapp’s 1 -13.655 functionalities. 1 -13.655 concreteness. 1 -13.655 regrettably 1 -13.655 conceived. 1 -13.655 thirdorder 1 -13.655 accessible, 1 -13.655 lmwt 1 -13.655 pp-modification 1 -13.655 frontiers. 1 -13.655 autre 1 -13.655 54, 309, 1 -13.655 cross. 1 -13.655 comma-separated 1 -13.655 jennifer foster, 1 -13.655 1and 2-way 1 -13.655 .109 1 -13.655 acount. 1 -13.655 (x-axis). 1 -13.655 stewart 1 -13.655 94.91 1 -13.655 restoring 1 -13.655 diagnosis. 1 -13.655 concerns. 1 -13.655 informatics. 1 -13.655 (lus) 1 -13.655 enthusiasm, elation, 1 -13.655 america. 1 -13.655 put. 1 -13.655 .5, 1 -13.655 uncertainties 1 -13.655 recal. 1 -13.655 deadline. 1 -13.655 acknowledgement. 1 -13.655 band. 1 -13.655 visualized. 1 -13.655 unite 1 -13.655 50%). 1 -13.655 both), 1 -13.655 varied, 1 -13.655 females. 1 -13.655 [1]. 1 -13.655 null u11(x,y,di,p) 1 -13.655 imposition, 1 -13.655 economies 1 -13.655 chapter. 1 -13.655 .5. 1 -13.655 (asts). 1 -13.655 “coherence” 1 -13.655 attitudinal 1 -13.655 41.73 1 -13.655 sharedtask. 1 -13.655 elson 1 -13.655 λx1 1 -13.655 patrick pantel 1 -13.655 ancestor. 1 -13.655 representation). 1 -13.655 #beat/0 1 -13.655 a↓ 1 -13.655 sid=''s1'' 1 -13.655 broad, topical 1 -13.655 (srl). 1 -13.655 steps). 1 -13.655 scribed 1 -13.655 2007]. 1 -13.655 xj. 1 -13.655 aist 1 -13.655 quantity%1:23:00 1 -13.655 ˆe, 1 -13.655 (2.7%) 1 -13.655 rβ 1 -13.655 u1. 1 -13.655 (18.a-b) 1 -13.655 know”. 1 -13.655 ‘our+’ 1 -13.655 is/not 1 -13.655 reordering: 1 -13.655 discrepancy.n 1 -13.655 filling. 1 -13.655 hk10. 1 -13.655 donald metzler 1 -13.655 ki-ya 1 -13.655 818,741 1 -13.655 interactivity. 1 -13.655 closest-first 1 -13.655 user-specified 1 -13.655 472 1 -13.655 ! 1 -13.655 root) 1 -13.655 granularity: 1 -13.655 reference). 1 -13.655 10−4. 1 -13.655 sib:parse 1 -13.655 cocited 1 -13.655 staff. 1 -13.655 best). 1 -13.655 w911nf-09-2-0053, 1 -13.655 sigkdd explorations, 1 -13.655 randomizations 1 -13.655 fever. 1 -13.655 slider. 1 -13.655 x(sprime)∧xprime(s) < 1 -13.655 recombination. 1 -13.655 acceleration. 1 -13.655 acquisition (sla) 1 -13.655 pnd 1 -13.655 tors 1 -13.655 14,200 1 -13.655 mrr, 1 -13.655 jer-iel 1 -13.655 browse. 1 -13.655 transliteration) 1 -13.655 haifa, 31905 1 -13.655 (hgfc) 1 -13.655 ers 1 -13.655 combined). 1 -13.655 flattened. 1 -13.655 sun. 1 -13.655 (193). 1 -13.655 sense(s) 1 -13.655 features? 1 -13.655 confidences 1 -13.655 costing. 1 -13.655 nivre, joakim; 1 -13.655 natural. 1 -13.655 nvda. 1 -13.655 (schein, 1993), 1 -13.655 300-best 1 -13.655 ��'wt�luk��d�. 1 -13.655 jail, 1 -13.655 duan, zhao, 1 -13.655 adjective’s 1 -13.655 sentimentlabeled 1 -13.655 iit bombay, 1 -13.655 homonymy. 1 -13.655 krey`ol. 1 -13.655 spreyer 1 -13.655 information? 1 -13.655 (wordnet) 1 -13.655 subcomponents 1 -13.655 euro. 1 -13.655 re-evaluate 1 -13.655 disputants, 1 -13.655 lmf mrd 1 -13.655 08.4 1 -13.655 link¨oping, 1 -13.655 low-content 1 -13.655 d1,...,d4 1 -13.655 hubness 1 -13.655 eq.8 1 -13.655 (meta-) 1 -13.655 iso 9241-11 1 -13.655 years), 1 -13.655 enhanced. 1 -13.655 non-aligned 1 -13.655 naacl. 1 -13.655 tagger). 1 -13.655 11.8% 1 -13.655 prop-dep. 1 -13.655 lasso, 1 -13.655 lumping 1 -13.655 measured, 1 -13.655 guaranteed. 1 -13.655 (gordon, 1 -13.655 place). 1 -13.655 tagsets. 1 -13.655 node-external 1 -13.655 affirmed, negated, 1 -13.655 causedmotion 1 -13.655 ision 1 -13.655 cabrio 1 -13.655 [conceived bsoa,intention 1 -13.655 references). 1 -13.655 unscoped. 1 -13.655 subtreelets. 1 -13.655 zi,jp 1 -13.655 season. 1 -13.655 σprime. 1 -13.655 (focusing 1 -13.655 3%) 1 -13.655 morphosyntactically 1 -13.655 153.0 1 -13.655 jj? 1 -13.655 bio. 1 -13.655 nominating 1 -13.655 rewordings 1 -13.655 (γ 1 -13.655 action-state 1 -13.655 wcnn 1 -13.655 annotator). 1 -13.655 liu lefei 1 -13.655 (mother, oblige, 1 -13.655 cs-tutoring 1 -13.655 d6. 1 -13.655 (:* ont::event-time-rel 1 -13.655 changing. 1 -13.655 classification? 1 -13.655 zd,n. 1 -13.655 translated) 1 -13.655 (p<0.01) 1 -13.655 (2.4) 1 -13.655 adverb) 1 -13.655 clean. 1 -13.655 pca. 1 -13.655 customer reviews. 1 -13.655 “irrelevant”. 1 -13.655 15.4% 1 -13.655 aprime. 1 -13.655 tailoring 1 -13.655 p!h! 1 -13.655 (hasegawa, 1 -13.655 film. 1 -13.655 (rao 1 -13.655 fd. 1 -13.655 1087 1 -13.655 i=1..k 1 -13.655 habe ihn 1 -13.655 ak. 1 -13.655 (em). 1 -13.655 (0.08) 1 -13.655 miss, 1 -13.655 exploring. 1 -13.655 module). 1 -13.655 6%. 1 -13.655 puerto rico, 1 -13.655 legomena 1 -13.655 (nouns 1 -13.655 poor). 1 -13.655 wordby-word 1 -13.655 video recordings 1 -13.655 semi-supervised. 1 -13.655 triggered. 1 -13.655 folder 1 -13.655 tile. 1 -13.655 tom kwiatkowski, 1 -13.655 desk. 1 -13.655 (e.g.) 1 -13.655 57,441 1 -13.655 type/token 1 -13.655 4374 1 -13.655 web1t. 1 -13.655 mehdad@fbk.eu 1 -13.655 adler 1 -13.655 k=10. 1 -13.655 41.8% 1 -13.655 co-document frequency 1 -13.655 fixed-size 1 -13.655 broadcoverage 1 -13.655 pascal-voc 1 -13.655 peerspeak 1 -13.655 subjects: 1 -13.655 nvda, 1 -13.655 (es). 1 -13.655 schematically 1 -13.655 crflm 1 -13.655 18159k 39.36 1 -13.655 town. 1 -13.655 fr1,fr2 1 -13.655 inaccuracies 1 -13.655 parsing? 1 -13.655 proximity, 1 -13.655 virus. 1 -13.655 homologene. 1 -13.655 per’s 1 -13.655 exists). 1 -13.655 contextvector 1 -13.655 quickest 1 -13.655 event-argument 1 -13.655 84.95 84.92 1 -13.655 umls. 1 -13.655 smalland 1 -13.655 decision-making. 1 -13.655 investigating. 1 -13.655 lp. 1 -13.655 break. 1 -13.655 makedon 1 -13.655 list2. 1 -13.655 “?” 1 -13.655 4see 1 -13.655 negations. 1 -13.655 cumulative. 0.5 -13.655 yuv 0.5 -13.655 s32 0.205882 -13.6554 we will refer 0.428571 -13.6555 lagus, 0.5 -13.6555 (bonneau-maynard 0.5 -13.6555 affect. 0.555556 -13.6556 1,2 0.555556 -13.6556 this process. 0.75 -13.6559 (zhu, 0.75 -13.6559 stanford university. 0.75 -13.6559 (dst), 0.75 -13.6559 84%. 0.75 -13.6559 target pl 0.75 -13.6559 computationally infeasible 0.75 -13.6559 huang. 0.75 -13.6559 2005b; 0.75 -13.6559 theorem. 0.75 -13.6559 (airola 0.75 -13.6559 contains loud 0.75 -13.6559 tensor. 0.75 -13.6559 u(v) 0.75 -13.6559 ctb2tp. 0.75 -13.6559 lacking. 0.75 -13.6559 (ast-covb) 0.75 -13.6559 x2,x3 0.75 -13.6559 per-subject 0.75 -13.6559 wilson. 0.75 -13.6559 tc. 0.75 -13.6559 confusionnetwork. 0.75 -13.6559 0}. 0.75 -13.6559 microtask 0.75 -13.6559 car. 0.75 -13.6559 negatively. 0.75 -13.6559 submitted. 0.75 -13.6559 66.14 0.75 -13.6559 when. 0.75 -13.6559 inappropriately 0.75 -13.6559 ciaramita, 0.75 -13.6559 1000) 0.75 -13.6559 scarcely 0.75 -13.6559 r4). 0.75 -13.6559 dataset). 0.75 -13.6559 (coifman 0.75 -13.6559 (jing, 0.75 -13.6559 keyword, 0.75 -13.6559 0.827 0.75 -13.6559 message). 0.75 -13.6559 edus. 0.75 -13.6559 1990. 0.75 -13.6559 amber-nl 0.75 -13.6559 deputy 0.75 -13.6559 generation). 0.75 -13.6559 conditioning. 0.75 -13.6559 crosscovariance 0.75 -13.6559 higgins, 0.75 -13.6559 – hurting 0.75 -13.6559 penn (2000). 0.75 -13.6559 selector. 0.75 -13.6559 undergraduates 0.75 -13.6559 aggregation, 0.75 -13.6559 (boyer 0.75 -13.6559 optimize. 0.75 -13.6559 iteration). 0.75 -13.6559 ^p 0.75 -13.6559 anosmia, 0.75 -13.6559 night. 0.75 -13.6559 tb1.2 0.75 -13.6559 unless specified 0.75 -13.6559 “complete” 0.75 -13.6559 front-end 0.75 -13.6559 believed. 0.75 -13.6559 maximality conditions 0.75 -13.6559 gold-annotated 0.75 -13.6559 ubiquitous, 0.75 -13.6559 taskspecific 0.75 -13.6559 pr vn 0.75 -13.6559 inverse. 0.75 -13.6559 λs.λx.λy.λdi.λp.u1(x,y,di,p,s) 0.75 -13.6559 fl. 0.75 -13.6559 created automatically, 0.75 -13.6559 developmentdata. 0.75 -13.6559 (core) 0.75 -13.6559 year). 0.75 -13.6559 mt; 0.75 -13.6559 unreliability 0.75 -13.6559 notion. 0.75 -13.6559 minimum-risk 0.75 -13.6559 subdivide 0.75 -13.6559 smh. 0.75 -13.6559 (grammatical 0.75 -13.6559 (0.4) 0.75 -13.6559 g(x). 0.75 -13.6559 rise. 0.75 -13.6559 sadrzadeh 0.75 -13.6559 occasional 0.75 -13.6559 utilization, 0.75 -13.6559 91.05 0.75 -13.6559 36.0% 0.75 -13.6559 cleanser. 0.75 -13.6559 wm. 0.75 -13.6559 0.15. 0.75 -13.6559 (fes) 0.75 -13.6559 .ngramm 0.75 -13.6559 pound 0.75 -13.6559 phone recognizer 0.75 -13.6559 excess 0.75 -13.6559 aq 0.75 -13.6559 literatures. 0.75 -13.6559 identifinder. 0.75 -13.6559 92.99 0.75 -13.6559 football player 0.75 -13.6559 representative. 0.75 -13.6559 u.s., 0.75 -13.6559 1023 0.75 -13.6559 2,214 0.75 -13.6559 interferences 0.75 -13.6559 0.657 0.75 -13.6559 mira. 0.75 -13.6559 rajkumar, 0.75 -13.6559 candidatearcs. 0.75 -13.6559 buildings. 0.75 -13.6559 interaction). 0.75 -13.6559 ead metadata 0.75 -13.6559 adl 0.75 -13.6559 out-of-context 0.75 -13.6559 burges 0.75 -13.6559 australia. 0.75 -13.6559 “general” 0.75 -13.6559 study). 0.75 -13.6559 gi. 0.75 -13.6559 stopword removal 0.75 -13.6559 cognates. 0.75 -13.6559 (marsi 0.75 -13.6559 prevent overfitting, 0.75 -13.6559 uima pipelines 0.75 -13.6559 release. 0.75 -13.6559 orthographies 0.75 -13.6559 (lsa), 0.75 -13.6559 pagerank. 0.75 -13.6559 unfortunate 0.75 -13.6559 90.00 0.75 -13.6559 45.83 0.75 -13.6559 acceptable dependent 0.75 -13.6559 chunkers, 0.75 -13.6559 (1976). 0.75 -13.6559 mixed initiative 0.75 -13.6559 baseball. 0.75 -13.6559 macro-ranking 0.75 -13.6559 13,13. 0.75 -13.6559 -0.200 0.75 -13.6559 s-factor. 0.75 -13.6559 predictor: 0.75 -13.6559 sector 0.75 -13.6559 accent. 0.75 -13.6559 computing. 0.75 -13.6559 auxiliary. 0.75 -13.6559 0.811 0.75 -13.6559 wordnet affect. 0.75 -13.6559 nowhere 0.75 -13.6559 proces, 0.75 -13.6559 (mcintosh, 0.75 -13.6559 1.4m 0.75 -13.6559 coeff. 0.75 -13.6559 similarity). 0.75 -13.6559 fourth i2b2/va 0.75 -13.6559 – par. 0.75 -13.6559 ¯c. 0.75 -13.6559 underway. 0.75 -13.6559 client. 0.75 -13.6559 e. banchs 0.75 -13.6559 markov order-2 0.75 -13.6559 chakrabarti 0.75 -13.6559 p(e) κ 0.75 -13.6559 (kobus 0.75 -13.6559 -dep 0.75 -13.6559 (§3.1), 0.75 -13.6559 scoretype-1 0.75 -13.6559 socioeconomic 0.75 -13.6559 0.549 0.75 -13.6559 experimentwith 0.75 -13.6559 feedforward 0.75 -13.6559 prs 0.75 -13.6559 self-learning clustering 0.75 -13.6559 0.721 0.75 -13.6559 capitalizing 0.75 -13.6559 scheme-set. 0.75 -13.6559 syntax-directed 0.75 -13.6559 localization. 0.75 -13.6559 predicates). 0.75 -13.6559 category). 0.75 -13.6559 (*). 0.75 -13.6559 sk. 0.75 -13.6559 yasuda 0.75 -13.6559 0.0001). 0.75 -13.6559 well-formed, 0.75 -13.6559 adjective-noun phrases, 0.75 -13.6559 cmnt1 0.75 -13.6559 detailed. 0.75 -13.6559 classify. 0.75 -13.6559 modality. 0.75 -13.6559 precision/recall/f-measure 0.75 -13.6559 (mrs) 0.75 -13.6559 gaz. 0.75 -13.6559 3 pre-test 0.75 -13.6559 # bleu ter 0.75 -13.6559 32.82 0.75 -13.6559 fragile. 0.75 -13.6559 our notation. 0.75 -13.6559 valtst 0.75 -13.6559 fun. 0.75 -13.6559 word-forms 0.75 -13.6559 tibshirani, 0.75 -13.6559 194). 0.75 -13.6559 error! 0.75 -13.6559 explanation: 0.75 -13.6559 magdy 0.75 -13.6567 vignettes. 0.75 -13.6567 embedding propositions 0.75 -13.6567 sentimen 0.75 -13.6567 david traum. 0.75 -13.6567 crawled2 0.75 -13.6567 1987. 0.75 -13.6567 lda) 0.75 -13.6567 eduard hovy, 0.75 -13.6567 t* 0.75 -13.6567 underlying lp 0.75 -13.6567 background: 0.75 -13.6567 28.67 0.75 -13.6567 messaging. 0.75 -13.6567 commonly studied 0.75 -13.6567 nc nl 0.75 -13.6567 nullth 0.75 -13.6567 84.73 0.75 -13.6567 (l2) 0.75 -13.6567 conflates 0.75 -13.6567 heterogeneity vs. 0.75 -13.6567 i’d like 0.75 -13.6567 50.89 0.75 -13.6567 (bhagat 0.75 -13.6567 topic drift 0.75 -13.6567 q5 0.75 -13.6567 channels. 0.75 -13.6567 feeling, 0.75 -13.6567 ji aajiji 0.75 -13.6567 22. 0.75 -13.6567 first responders 0.75 -13.6567 (candidate 0.75 -13.6567 (lda; 0.75 -13.6567 1954). 0.75 -13.6567 espresso (pantel 0.75 -13.6567 non-polysemous 0.75 -13.6567 45.08 0.75 -13.6567 b&l, 0.75 -13.6567 newspapers). 0.75 -13.6567 94.17 0.75 -13.6567 lexicographic resource 0.75 -13.6567 range) 0.75 -13.6575 3003 0.75 -13.6575 subjective judgments 0.75 -13.6575 † indicates 0.75 -13.6575 brier score, 0.75 -13.6575 vp)) 0.75 -13.6575 freund 0.75 -13.6575 sgml/xml 0.75 -13.6575 1st international 0.75 -13.6575 zou haar 0.75 -13.6575 48.62 0.75 -13.6575 da cunha 0.75 -13.6575 (in particular, 0.75 -13.6575 common-sense knowledge, 0.75 -13.6575 (curran, 0.75 -13.6575 typedm, 0.75 -13.6575 1−ter 0.75 -13.6575 * significant 0.75 -13.6575 powerful, 0.75 -13.6575 5-6 0.75 -13.6575 502 0.75 -13.6575 (hartung 0.75 -13.6575 ∪ ?x 0.75 -13.6575 q22 0.75 -13.6575 66.7%. 0.75 -13.6575 63.72 0.75 -13.6575 n-gram cooccurrence 0.75 -13.6575 w(u,v) 0.428571 -13.6579 artists 0.666667 -13.658 cachepipe 0.75 -13.6584 discussed below, 0.75 -13.6584 obituary 0.75 -13.6584 §2, 0.75 -13.6584 209, 0.75 -13.6584 0.524 0.75 -13.6584 ed” 0.75 -13.6584 amphitheatre parkway 0.75 -13.6584 reading research 0.75 -13.6584 done against 0.75 -13.6584 specially, 0.75 -13.6584 what matters 0.75 -13.6584 save space. 0.75 -13.6584 lessen 0.75 -13.6584 (feature-edges, 0.75 -13.6584 gene databases. 0.255814 -13.6585 lemon 0.555556 -13.6587 (she 0.666667 -13.6588 memorable 0.666667 -13.6588 positives: 0.666667 -13.6588 336 0.75 -13.6592 linguistic processors 0.75 -13.6592 state sprimei, 0.75 -13.6592 misinterpreted 0.75 -13.6592 sign test, 0.75 -13.6592 often lacks 0.75 -13.6592 stop lists 0.75 -13.6592 correct cognate 0.75 -13.6592 f1) 0.75 -13.6592 multiclas 0.75 -13.6592 helps explain 0.75 -13.6592 notwithstanding any 0.75 -13.6592 bigram clustering. 0.75 -13.6592 wiki pages 0.75 -13.6592 (brin 0.555556 -13.6594 ukraine 0.666667 -13.6595 { ?s 0.666667 -13.6595 mcccj+sd 0.75 -13.66 21.0% 0.75 -13.66 english (nucle), 0.75 -13.66 0.828 0.75 -13.66 converged, 0.75 -13.66 expressed differently 0.75 -13.66 part-ofspeech tagging 0.75 -13.66 2,420 0.75 -13.66 near perfect 0.75 -13.66 probs 0.75 -13.66 0.903 0.125 -13.6603 appearing 0.244898 -13.6605 instructor 0.75 -13.6609 cucerzan, 0.75 -13.6609 j48 algorithm 0.75 -13.6609 (doddington 0.75 -13.6609 kintsch 0.75 -13.6609 entropy pruning 0.75 -13.6609 “when was 0.75 -13.6609 (primary) 0.75 -13.6609 msdr+ 0.75 -13.6609 strongly tied 0.75 -13.6609 context intrusion, 0.666667 -13.6617 technology-assisted 0.666667 -13.6617 sim(q,x) 0.75 -13.6617 “and”-splice 0.75 -13.6617 0.896 0.75 -13.6617 parse tree) 0.75 -13.6617 coercion detection, 0.75 -13.6617 lexical database, 0.75 -13.6617 morphological inflections 0.75 -13.6617 essential part 0.75 -13.6617 invertebrates 0.75 -13.6617 collisions 0.75 -13.6617 generalized dependencies, 0.5 -13.6622 ;a/ ‘ 0.5 -13.6622 thus cannot 0.666667 -13.6625 monday, 0.75 -13.6626 one-class 0.75 -13.6626 beam search decoder 0.75 -13.6626 seven-word 0.75 -13.6626 85.20 0.75 -13.6626 bottom-up manner 0.75 -13.6626 natural sounding 0.5 -13.6633 pisa 0.75 -13.6634 69.3% 0.75 -13.6634 (pasca 0.75 -13.6634 pharmaceutical 0.75 -13.6634 different granularities 0.75 -13.6634 besides having 0.75 -13.6634 kbp 2010 0.110638 -13.6634 1). 0.555556 -13.6637 21.1 0.24 -13.6638 answer. 0.75 -13.6642 rand-base 0.75 -13.6642 3.4.1 0.75 -13.6642 taliban 0.75 -13.6642 repeated labeling 0.75 -13.6642 folded into 0.75 -13.6642 toy corpus, 0.75 -13.6642 4.45 0.555556 -13.665 higher bleu score 0.75 -13.6651 unfactored 0.75 -13.6651 label li 0.75 -13.6651 ∪ ^x 0.75 -13.6651 rule-set 0.75 -13.6651 dss 0.75 -13.6651 (pn) 0.666667 -13.6655 586 0.666667 -13.6655 ∗d(fn) 0.666667 -13.6655 decidido 0.666667 -13.6655 error-tagged 0.666667 -13.6655 similar: 0.213115 -13.6655 labeled training 0.5 -13.6656 conclude, 0.5 -13.6656 bojar, 0.5 -13.6656 guard 0.75 -13.6659 5-best 0.75 -13.6659 do triangleright 0.75 -13.6659 paul et al. 0.75 -13.6659 wilks, 0.154472 -13.6661 we experimented 0.5 -13.6661 catalogue 0.666667 -13.6662 j-qry 0.2 -13.6662 verbs are 0.461538 -13.6664 velikovich 0.363636 -13.6664 two groups 0.75 -13.6668 (nullnull @null) 0.75 -13.6668 0.820 0.75 -13.6668 timemachine 0.75 -13.6668 las tasas 0.75 -13.6668 0.727 0.75 -13.6668 (baseline), 0.75 -13.6668 who contributed 0.75 -13.6668 older, 0.75 -13.6668 kdd-d 0.75 -13.6668 ogi data 0.75 -13.6668 moilanen 0.75 -13.6668 latent-variable grammar 0.75 -13.6668 0.798 0.363636 -13.6668 documents) 0.666667 -13.667 ot 0.666667 -13.667 interchange 0.268293 -13.6672 vogel, 0.75 -13.6676 open mind 0.75 -13.6676 eskenazi, m. 0.75 -13.6676 vp1 0.75 -13.6676 principle: 0.666667 -13.6677 kaufmann 0.666667 -13.6677 hpsg parser 0.666667 -13.6677 aly 0.666667 -13.6677 “some 0.461538 -13.6679 gaps are 0.666667 -13.6685 899 0.3 -13.6685 fast and 0.75 -13.6685 (cafarella 0.75 -13.6685 henderson, 2006; 0.75 -13.6685 monday. 0.75 -13.6685 dfa 0.75 -13.6685 ∊ 0.75 -13.6685 (plate, 0.75 -13.6685 ekman, 0.277778 -13.6687 both systems 0.75 -13.6693 0.944 0.75 -13.6693 perfectly, 0.75 -13.6693 nevertheles, 0.75 -13.6693 buckley 0.333333 -13.6696 opted 0.75 -13.6701 functional morphological 0.75 -13.6701 rt03+dev04f 0.75 -13.6701 (2005)) 0.75 -13.6701 like tea 0.75 -13.6701 tij 0.75 -13.6701 popularized by 0.75 -13.6701 998 0.75 -13.6701 abstracting over 0.75 -13.6701 confers 0.75 -13.6701 budapestacad 0.411765 -13.6701 mail 0.277778 -13.6705 newsgroups 0.666667 -13.6707 kick 0.666667 -13.6707 filename 0.428571 -13.6709 event occurrences 0.75 -13.671 temp 0.75 -13.671 without any further 0.75 -13.671 domains) 0.75 -13.671 or sentence) 0.75 -13.671 sudo 0.461538 -13.6711 mailing 0.310345 -13.6714 efficacy 0.666667 -13.6715 lvcs, 0.666667 -13.6715 conditioning on 0.0609137 -13.6717 instance, 0.75 -13.6718 unavailability 0.75 -13.6718 “t” 0.75 -13.6718 table 6: example 0.75 -13.6718 uryupina 0.75 -13.6718 “all years” 0.75 -13.6718 0.762 0.75 -13.6718 rijke 0.555556 -13.6718 ubiub 0.156522 -13.6722 summary. 0.666667 -13.6722 baby 0.666667 -13.6722 (2008): 0.75 -13.6727 828 0.75 -13.6727 ws-353 0.75 -13.6727 0.444 0.75 -13.6727 soleil 0.75 -13.6727 209) 0.75 -13.6727 shows promise 0.75 -13.6727 clear(agenda) 0.75 -13.6727 nullnullnullnull nullnullnullnull 0.75 -13.6727 hoy, 0.75 -13.6727 very-non 0.75 -13.6727 damsl annotation 0.27027 -13.6732 a generalization 0.75 -13.6735 hubmorph 0.75 -13.6735 complete-link 0.75 -13.6735 odp 0.75 -13.6735 0.774 0.75 -13.6735 communicative function, 0.75 -13.6735 qualitative, 0.75 -13.6735 then-gram 0.5 -13.6735 translation (mt). 0.666667 -13.6737 99.9 0.666667 -13.6737 1250 0.0827423 -13.6739 2004; 0.220339 -13.674 manner, 0.333333 -13.6741 ny, 0.27027 -13.6741 these include 0.75 -13.6744 schulte 0.75 -13.6744 1580 0.75 -13.6744 post-coreference 0.75 -13.6744 model-generated 0.75 -13.6744 78.2% 0.75 -13.6744 (voorhees 0.75 -13.6744 (munro, 0.666667 -13.6745 ¨ 0.162162 -13.6749 extraction from 0.0548654 -13.675 1 introduction 0.113636 -13.6751 translation system 0.75 -13.6752 0.658 0.75 -13.6752 vms, 0.75 -13.6752 five-year 0.75 -13.6752 disprefer 0.75 -13.6752 algorithms operate 0.75 -13.6752 +gw 0.75 -13.6752 sorting, 0.75 -13.6752 rosé, c. 0.75 -13.6752 topicality 0.461538 -13.6752 2005)). 0.5 -13.6752 (?; 0.666667 -13.6753 21.79 0.666667 -13.6753 1050 0.461538 -13.6758 subrequest 0.666667 -13.676 profitable 0.388889 -13.6761 seed set 0.75 -13.6761 section: 0.75 -13.6761 -0.12 0.75 -13.6761 basing on 0.75 -13.6761 duluth-wsi 0.75 -13.6761 truncation, 0.75 -13.6761 an opensource 0.75 -13.6761 re-run 0.75 -13.6761 15.0 0.75 -13.6761 562 0.75 -13.6761 15.9 0.75 -13.6761 2.36 0.75 -13.6761 subontologies 0.75 -13.6761 modeling morphological 0.75 -13.6761 (nie 0.75 -13.6761 diverged 0.666667 -13.6768 assets 0.75 -13.6769 northern 0.75 -13.6769 (larsson 0.75 -13.6769 cube, 0.75 -13.6769 72.1% 0.75 -13.6769 (vossen, 0.75 -13.6769 sarmento 0.75 -13.6769 indexes, 0.75 -13.6769 reviewing, 0.75 -13.6769 committee, 0.75 -13.6769 attended 0.75 -13.6769 (butnariu 0.666667 -13.6775 currier 0.555556 -13.6775 electricity 0.363636 -13.6775 lines indicate 0.194805 -13.6778 argued 0.75 -13.6778 grounds, 0.75 -13.6778 rarely, 0.75 -13.6778 any bootstrapping. 0.75 -13.6778 lm-fst 0.0725953 -13.6779 evaluation of 0.5 -13.678 be preserved 0.428571 -13.6781 alon 0.428571 -13.6781 .29 0.666667 -13.6783 tcov 0.411765 -13.6785 methods, such 0.75 -13.6786 88.03 0.75 -13.6786 (1968), 0.75 -13.6786 wordspace, 0.75 -13.6786 pseudo-projective 0.75 -13.6786 changes - 0.75 -13.6786 +token-pos 0.75 -13.6786 upon closer 0.75 -13.6786 encapsulates 0.75 -13.6786 italian: 0.75 -13.6786 bit, 0.75 -13.6786 higginbotham 0.75 -13.6786 after convergence, 0.119403 -13.679 above. 0.666667 -13.679 0.129 0.411765 -13.679 (xue 0.75 -13.6795 distortions 0.75 -13.6795 [inverted, 0.75 -13.6795 abbreviated, 0.75 -13.6795 advice, 0.75 -13.6795 matsuzaki 0.75 -13.6795 h&s10 0.75 -13.6795 emotions evoked 0.75 -13.6795 07] 0.75 -13.6795 ih 0.75 -13.6795 cooccurence 0.235294 -13.6799 parsing models 0.555556 -13.68 pad 0.27027 -13.6802 rieser 0.5 -13.6803 regularize 0.5 -13.6803 road, 0.75 -13.6804 drag 0.75 -13.6804 decrement 0.75 -13.6804 kulesza 0.75 -13.6804 plural forms 0.75 -13.6804 qis 0.75 -13.6804 these phenomena. 0.75 -13.6804 likelihood l 0.75 -13.6804 album, 0.75 -13.6804 categorically 0.75 -13.6804 smells 0.666667 -13.6806 indexer 0.310345 -13.6809 liu and 0.461538 -13.681 episode 0.75 -13.6812 gibson 0.75 -13.6812 (ye 0.75 -13.6812 organ 0.75 -13.6812 0.462 0.75 -13.6812 edition) 0.75 -13.6812 svr-base 0.75 -13.6812 tat-seng 0.75 -13.6812 1.02 0.75 -13.6812 hostpart 0.666667 -13.6813 inthe 0.5 -13.6814 nakazawa 0.321429 -13.6818 our solution 0.117073 -13.682 it may 0.244444 -13.682 can thus 0.285714 -13.6821 and wintner 0.666667 -13.6821 passive sentences 0.75 -13.6821 cash-desk 0.75 -13.6821 90.9% 0.75 -13.6821 415 0.75 -13.6821 right-endoc 0.75 -13.6821 decouple 0.75 -13.6821 compressing 0.75 -13.6821 ق 0.75 -13.6821 20.83 0.75 -13.6821 1968), 0.75 -13.6821 non-lexicalised 0.75 -13.6821 94.8% 0.75 -13.6821 (temporal 0.75 -13.6821 (utiyama 0.555556 -13.6826 surdeanu et 0.461538 -13.6826 sky 0.75 -13.6829 twenty different 0.75 -13.6829 around 20% 0.75 -13.6829 charlie 0.75 -13.6829 lack, 0.75 -13.6829 (sahlgren, 0.75 -13.6829 stabilized 0.75 -13.6829 pres) 0.75 -13.6829 insulting 0.75 -13.6829 llc 0.75 -13.6829 unattested but 0.75 -13.6829 thinking, 0.75 -13.6829 (tables 1 0.555556 -13.6832 v6 0.666667 -13.6836 µ(ν) is 0.666667 -13.6836 auf 0.461538 -13.6836 chart-based 0.75 -13.6838 protagonist’s 0.75 -13.6838 dredze, 2010). 0.75 -13.6838 hw’s 0.75 -13.6838 −pp] 0.75 -13.6838 term) 0.75 -13.6838 roleset, 0.75 -13.6838 x’s 0.666667 -13.6844 (4): 0.555556 -13.6845 “bag 0.75 -13.6846 items; 0.75 -13.6846 f1%rte-5 0.75 -13.6846 (38) 0.75 -13.6846 recipients, 0.75 -13.6846 dhpb 0.666667 -13.6851 devel 0.461538 -13.6852 unimodal 0.209677 -13.6854 opportunity 0.75 -13.6855 (fewer 0.75 -13.6855 foes 0.75 -13.6855 19.1 0.75 -13.6855 mp34ibm1 0.75 -13.6855 exceptionally 0.75 -13.6855 approximately, 0.75 -13.6855 measurements, 0.75 -13.6855 news), 0.75 -13.6855 1.92 0.75 -13.6855 nber data, 0.428571 -13.6855 marks. 0.277778 -13.6857 search results 0.666667 -13.6859 is) 0.411765 -13.686 (r), 0.75 -13.6864 liwc (pennebaker 0.75 -13.6864 actions), 0.75 -13.6864 (rifkin 0.75 -13.6864 duce 0.75 -13.6864 diacritics, 0.75 -13.6864 559 0.75 -13.6864 forest: 0.75 -13.6864 underperforming 0.75 -13.6864 1012 0.75 -13.6864 grow-diag 0.75 -13.6864 centred 0.428571 -13.6865 (have 0.310345 -13.6866 bayes classifier 0.666667 -13.6867 analyzed texts 0.75 -13.6872 0.72, 0.75 -13.6872 represents, 0.75 -13.6872 apple’ 0.75 -13.6872 agarwal’s 0.75 -13.6872 equal size. 0.75 -13.6872 15.82 0.75 -13.6872 high-degree 0.75 -13.6872 oseo 0.666667 -13.6874 under development 0.666667 -13.6874 feature-augmented 0.666667 -13.6874 wikification 0.666667 -13.6874 attacked 0.666667 -13.6874 ct(w) 0.555556 -13.6877 flemish 0.75 -13.6881 85.28 0.75 -13.6881 f0, 0.75 -13.6881 12,245 0.75 -13.6881 0.787 0.75 -13.6881 13.8 0.75 -13.6881 68.12 0.75 -13.6881 gender: 0.75 -13.6881 lopez, 0.75 -13.6881 decoded, 0.75 -13.6881 4.1), 0.461538 -13.6884 final feature 0.75 -13.6889 warning 0.75 -13.6889 27.5% 0.75 -13.6889 0.650 0.75 -13.6889 model” 0.75 -13.6889 fulfilling 0.75 -13.6889 s−1 0.75 -13.6889 svt 0.75 -13.6889 (kulkarni09) 0.75 -13.6889 off-theshelf 0.75 -13.6889 my name 0.75 -13.6889 langmead 0.461538 -13.6889 have addressed 0.428571 -13.6894 p(d 0.75 -13.6898 yessenalina 0.75 -13.6898 messages; 0.75 -13.6898 n[pi] 0.75 -13.6898 cross-market 0.75 -13.6898 kinyon 0.00330014 -13.6902 the 0.150794 -13.6902 result. 0.428571 -13.6904 term-label 0.428571 -13.6904 interval. 0.75 -13.6907 0.674 0.75 -13.6907 (mizuta 0.75 -13.6907 less) 0.75 -13.6907 butter 0.75 -13.6907 indoeuropean 0.75 -13.6907 italics) 0.75 -13.6907 analysers 0.75 -13.6907 deposition sentences, 0.75 -13.6907 o(n4) 0.75 -13.6907 conjugations 0.290323 -13.6912 (surdeanu 0.75 -13.6915 ahmad 0.75 -13.6915 f15g3 0.75 -13.6915 decay.” 0.75 -13.6915 detecting entailment 0.75 -13.6915 12.28 0.75 -13.6915 adjpagesim 0.75 -13.6915 substrate 0.75 -13.6915 auxiliary-verb 0.75 -13.6915 secondary, 0.75 -13.6915 likelihood) 0.75 -13.6915 mitamura 0.255814 -13.6915 matsumoto, 0.5 -13.6917 this representation, 0.0252716 -13.6919 will 0.064194 -13.6919 size of 0.666667 -13.692 cn-based 0.666667 -13.692 0.250 0.388889 -13.6922 consonant 0.75 -13.6924 d-tuna 0.75 -13.6924 difer 0.75 -13.6924 curved 0.75 -13.6924 transgenes 0.75 -13.6924 (wilson, 0.75 -13.6924 sangati et al. 0.75 -13.6924 clause) 0.75 -13.6924 founding 0.75 -13.6924 query tree, 0.75 -13.6924 permanent 0.75 -13.6924 more concretely, 0.75 -13.6924 400,000 0.75 -13.6924 nanba 0.75 -13.6924 (mostly) 0.75 -13.6924 about” 0.75 -13.6924 topicsum 0.75 -13.6924 clause; 0.0402723 -13.6925 part 0.555556 -13.6928 0.760 0.75 -13.6933 60.26 0.75 -13.6933 removing noisy 0.75 -13.6933 β) 0.75 -13.6933 71.7% 0.75 -13.6933 dynamics, 0.75 -13.6933 bost 0.75 -13.6933 consciousness 0.75 -13.6933 28.52 0.75 -13.6933 1979, 0.75 -13.6933 under-represented 0.75 -13.6933 hdp, 0.75 -13.6933 rule-based, 0.5 -13.6935 norwegian 0.666667 -13.6936 (average) 0.75 -13.6941 siteof 0.75 -13.6941 sensible, 0.75 -13.6941 elements) 0.75 -13.6941 activated, 0.75 -13.6941 (is) 0.75 -13.6941 1977, 0.75 -13.6941 myspace 0.75 -13.6941 17.51 0.75 -13.6941 hands, 0.75 -13.6941 valuation 0.75 -13.6941 (verb) 0.75 -13.6941 tree-banked 0.75 -13.6941 snr 0.75 -13.6941 overtraining 0.75 -13.6941 in§5 0.75 -13.6941 ghahramani, 0.75 -13.6941 (finlayson 0.666667 -13.6943 narasimhan 0.666667 -13.6943 (± 0.666667 -13.6943 oil* 0.666667 -13.6943 preferential 0.014458 -13.6946 not 0.5 -13.6946 ‘in 0.555556 -13.6947 mpnc 0.75 -13.695 (chinchor, 0.75 -13.695 foreign-foreign 0.75 -13.695 misalignment 0.75 -13.695 count) 0.75 -13.695 left-endoc 0.75 -13.695 cross-validating 0.75 -13.695 cohyponyms 0.75 -13.695 ,p 0.75 -13.695 shuttlecock 0.75 -13.695 klautau, 0.75 -13.695 lgc 0.75 -13.695 29.3% 0.75 -13.695 (takamura 0.75 -13.695 desiring 0.75 -13.695 v3i 0.75 -13.695 37.2% 0.75 -13.695 resilient 0.75 -13.695 (vanlehn 0.75 -13.695 discontinuity 0.75 -13.695 clueweb09 0.75 -13.695 1270 0.666667 -13.6951 1/0 0.5 -13.6958 schwartz 0.5 -13.6958 punyakanok 0.666667 -13.6959 these challenges. 0.666667 -13.6959 1979; 0.666667 -13.6959 399 0.75 -13.6959 tx(y) 0.75 -13.6959 (nouns), 0.75 -13.6959 xn 0.75 -13.6959 landscape 0.75 -13.6959 situate 0.75 -13.6959 91.0% 0.75 -13.6959 upenn 0.75 -13.6959 discrete, 0.75 -13.6959 signers 0.75 -13.6959 gong 0.75 -13.6959 in-degree 0.75 -13.6959 bi-gram, 0.75 -13.6959 ads, 0.75 -13.6959 product recommendation 0.75 -13.6959 ideal, 0.75 -13.6959 foltz, 0.75 -13.6959 desk 0.75 -13.6959 (gruber 0.461538 -13.6964 in our corpus, 0.0989761 -13.6965 al., 2007). 0.75 -13.6967 weightings 0.75 -13.6967 p<.01 0.75 -13.6967 ark 0.75 -13.6967 p(qis 0.75 -13.6967 not-continue, 0.75 -13.6967 fazly 0.75 -13.6967 31.9% 0.75 -13.6967 0.722 0.75 -13.6967 wick 0.75 -13.6967 competencies 0.75 -13.6967 7it 0.75 -13.6967 pos-4gram 0.75 -13.6967 −→ni 0.75 -13.6967 really, 0.75 -13.6967 bleu-case 0.75 -13.6967 w2) 0.75 -13.6967 seventeen 0.666667 -13.6974 holes 0.75 -13.6976 1-way 0.75 -13.6976 σi 0.75 -13.6976 negation) 0.75 -13.6976 left-toright 0.75 -13.6976 iran $325 0.75 -13.6976 single-trial 0.75 -13.6976 schroeder (2007) 0.75 -13.6976 84.5% 0.75 -13.6976 hop 0.75 -13.6976 caseli 0.75 -13.6976 extlesk 0.75 -13.6976 webber, 0.75 -13.6976 (particularly 0.75 -13.6976 rater-comment 0.75 -13.6976 recourse 0.75 -13.6976 lays 0.75 -13.6976 sr. 0.666667 -13.6982 588 0.666667 -13.6982 log-normal 0.75 -13.6985 linguistic, 0.75 -13.6985 utterances; 0.75 -13.6985 (bikel, 0.75 -13.6985 earn 0.75 -13.6985 written, 0.75 -13.6985 interfaces, 0.75 -13.6985 related fields, 0.75 -13.6985 sprimel 0.75 -13.6985 are warranted 0.75 -13.6985 comprehender 0.75 -13.6985 39.23 0.75 -13.6985 teachers, 0.75 -13.6985 0.496 0.75 -13.6985 valid intersection 0.75 -13.6985 (mmr) 0.0626741 -13.6986 2010) 0.5 -13.6987 logit 0.428571 -13.6988 tall 0.75 -13.6993 807 0.75 -13.6993 spm 0.75 -13.6993 19.4% 0.75 -13.6993 (bernstein 0.75 -13.6993 moretti 0.75 -13.6993 (opinion 0.75 -13.6993 (cherry, 0.75 -13.6993 “core” 0.75 -13.6993 joanis, 0.75 -13.6993 brin, 0.75 -13.6993 assaulted 0.75 -13.6993 rennr¨ader” 0.666667 -13.6997 s-factors, 0.75 -13.7002 2.75 0.75 -13.7002 topicalization 0.75 -13.7002 machine-based 0.75 -13.7002 medt 0.75 -13.7002 shaped 0.75 -13.7002 fier 0.75 -13.7002 746 0.75 -13.7002 (biber, 0.75 -13.7002 42.31 0.75 -13.7002 broad-coverage, 0.75 -13.7002 membership, 0.180723 -13.7006 and mooney, 0.75 -13.7011 oprd 0.75 -13.7011 depth-one 0.75 -13.7011 modulated 0.75 -13.7011 (future 0.75 -13.7011 975 0.75 -13.7011 sub-span 0.75 -13.7011 xts 0.75 -13.7011 robots, 0.75 -13.7011 logarithms 0.75 -13.7011 (experiment 0.75 -13.7011 shafran 0.75 -13.7011 3.23 0.75 -13.7011 (seginer, 0.75 -13.7011 (vroomen 0.75 -13.7011 punctuation; 0.666667 -13.7013 p(a)−p(e)1−p(e) 0.388889 -13.7016 entity disambiguation 0.290323 -13.7016 surprising, 0.555556 -13.7018 philippe 0.555556 -13.7018  0.75 -13.7019 14.1% 0.75 -13.7019 new proteins, 0.75 -13.7019 (cruse, 0.75 -13.7019 full: 0.75 -13.7019 unselected 0.75 -13.7019 negotiation, 0.75 -13.7019 target language’s 0.75 -13.7019 eck 0.75 -13.7019 crop 0.75 -13.7019 objects; 0.75 -13.7019 2.18 0.75 -13.7019 4.8% 0.666667 -13.7021 turing 0.216667 -13.7023 scale. 0.115385 -13.7027 framework. 0.666667 -13.7028 formidable 0.666667 -13.7028 exp( 0.75 -13.7028 earley 0.75 -13.7028 92.5% 0.75 -13.7028 variable random 0.75 -13.7028 underpinnings 0.75 -13.7028 “so” 0.75 -13.7028 93.79 0.75 -13.7028 520 0.75 -13.7028 reddy 0.75 -13.7028 un-reordered 0.75 -13.7028 contexts seed 0.75 -13.7028 selfridge 0.75 -13.7028 recombination 0.75 -13.7028 (finding 0.75 -13.7028 (r, 0.75 -13.7028 3.3), 0.75 -13.7028 (ptb 0.75 -13.7028 machines) 0.461538 -13.7028 data points. 0.428571 -13.7028 section 5 concludes 0.3 -13.7032 bleu nist 0.5 -13.7033 marcello 0.145038 -13.7036 is thus 0.666667 -13.7036 bytes/ 0.75 -13.7037 over-fit 0.75 -13.7037 qgstec-2010 0.75 -13.7037 isolates 0.75 -13.7037 njalsgade 0.75 -13.7037 wellner, 0.75 -13.7037 54.44 0.75 -13.7037 countless 0.75 -13.7037 inaccuracy 0.75 -13.7037 size ≥ 0.75 -13.7037 impression, 0.75 -13.7037 enju, 0.75 -13.7037 (sridhar, 0.75 -13.7037 non-expert swsd 0.75 -13.7037 water) 0.75 -13.7037 uncompressed 0.428571 -13.7043 hierarchical topic 0.571429 -13.7045 times). 0.571429 -13.7045 randomly-selected 0.571429 -13.7045 gwb20 0.571429 -13.7045 ¨a 0.571429 -13.7045 news-commentary, 0.134615 -13.7045 category. 0.5 -13.7045 most efficient 0.75 -13.7046 composed-rule 0.75 -13.7046 'coffee' 0.75 -13.7046 history-length 0.75 -13.7046 (chinese 0.75 -13.7046 entities” 0.75 -13.7046 thesaurus: 0.75 -13.7046 denmark 0.75 -13.7046 aliases, 0.75 -13.7046 intense 0.75 -13.7046 (entity) 0.75 -13.7046 elicits 0.75 -13.7046 (tt) 0.75 -13.7046 osteoporosis 0.75 -13.7054 0.671 0.75 -13.7054 focusses 0.75 -13.7054 53.31 0.75 -13.7054 out-of-domain data, 0.75 -13.7054 77.4% 0.75 -13.7054 967 0.75 -13.7054 “h 0.75 -13.7054 0.224 0.75 -13.7054 library’s 0.75 -13.7054 (newstest2010) 0.75 -13.7054 geolocating 0.75 -13.7054 simpler) 0.75 -13.7054 fce 0.75 -13.7054 jackknife 0.75 -13.7054 exists: 0.75 -13.7054 disjoint, 0.75 -13.7054 tated 0.666667 -13.706 prime) 0.666667 -13.706 cn-lvc 0.75 -13.7063 guess, 0.75 -13.7063 corpus.1 0.75 -13.7063 (kessler 0.75 -13.7063 dvd, 0.75 -13.7063 (xiao 0.75 -13.7063 0.254 0.75 -13.7063 samples-per-sentence 0.75 -13.7063 0.127 0.75 -13.7063 polarity-dissimilar 0.75 -13.7063 pendency 0.75 -13.7063 interchangeable, 0.75 -13.7063 (reg) 0.75 -13.7063 ‘o’ 0.75 -13.7063 association measure. 0.666667 -13.7067 amigos 0.666667 -13.7067 encipherment 0.75 -13.7072 noisy; 0.75 -13.7072 conll-2003 0.75 -13.7072 notated 0.75 -13.7072 articulation 0.75 -13.7072 subtle but 0.75 -13.7072 94.02 0.75 -13.7072 unlimited) 0.75 -13.7072 (wei 0.75 -13.7072 conroy 0.75 -13.7072 ky 0.5 -13.7074 (upper 0.666667 -13.7075 ¸ 0.555556 -13.7076 sang 0.571429 -13.7078 dual form 0.571429 -13.7078 accept or 0.5 -13.708 ships 0.75 -13.7081 85.18 0.75 -13.7081 “2” 0.75 -13.7081 0.083 0.75 -13.7081 two-pronged 0.75 -13.7081 anonymous reviewer 0.75 -13.7081 lowerlevel 0.75 -13.7081 ibm-baseline 0.75 -13.7081 lazily 0.75 -13.7081 verification, 0.75 -13.7081 canas, 0.75 -13.7081 synset’s 0.75 -13.7081 tversky 0.75 -13.7081 schabes 0.75 -13.7081 productivity, 0.75 -13.7081 1479 0.75 -13.7081 ha-2 0.75 -13.7081 replying 0.75 -13.7081 hindrance 0.75 -13.7081 pseudoword 0.75 -13.7081 -d 0.75 -13.7081 (associated 0.75 -13.7081 mention: 0.75 -13.7081 alignmenta 0.75 -13.7081 n’th 0.75 -13.7081 (binomial 0.75 -13.7081 encouragement 0.75 -13.7081 network) 0.263158 -13.7082 handful of 0.555556 -13.7083 sigir 0.571429 -13.7085 jump back 0.277778 -13.7086 expected. 0.5 -13.7086 roser 0.75 -13.709 leuski 0.75 -13.709 petersen 0.75 -13.709 hiragana 0.75 -13.709 chandrasekar 0.75 -13.709 epps, 0.75 -13.709 -loc 0.75 -13.709 we: 0.75 -13.709 [6 0.75 -13.709 systemdescription 0.75 -13.709 -alo 0.75 -13.709 english-foreign 0.75 -13.709 0.788 0.75 -13.709 adequacy-oriented 0.75 -13.709 renewed 0.75 -13.709 highly, 0.75 -13.709 28.63 0.75 -13.709 94.6% 0.75 -13.709 character’s 0.75 -13.709 eula 0.75 -13.709 99.2% 0.666667 -13.7091 q∈q 0.571429 -13.7091 (he, 0.571429 -13.7091 reflexes 0.571429 -13.7091 wmtoverview 0.571429 -13.7091 steinberger, 0.5 -13.7092 lopez 0.555556 -13.7096 (fung, 0.168421 -13.7096 bleu ter 0.27027 -13.7096 in contrast, our 0.75 -13.7098 (harabagiu 0.75 -13.7098 turn-internal 0.75 -13.7098 intra-category 0.75 -13.7098 slot-filling 0.75 -13.7098 (document) 0.75 -13.7098 “fill 0.75 -13.7098 goodman’s 0.75 -13.7098 0.225 0.75 -13.7098 chi-ho 0.75 -13.7098 (matsumoto 0.75 -13.7098 sakyo-ku, 0.75 -13.7098 extraversion, 0.75 -13.7098 894 0.571429 -13.7098 331 0.571429 -13.7098 every string 0.666667 -13.7099 easily incorporate 0.3 -13.7099 (baker 0.555556 -13.7102 281 0.555556 -13.7102 q1. 0.347826 -13.7105 α1 0.571429 -13.7105 redistribution 0.571429 -13.7105 22nd 0.75 -13.7107 berkeleyparser, 0.75 -13.7107 carberry, 0.75 -13.7107 0.899 0.75 -13.7107 0.172 0.75 -13.7107 “good”) 0.75 -13.7107 byrne 0.75 -13.7107 (butt, 0.75 -13.7107 fep 0.75 -13.7107 0.484 0.75 -13.7107 12,449 0.75 -13.7107 hiccup 0.75 -13.7107 gual 0.75 -13.7107 oflazer, 0.3 -13.7109 (church 0.571429 -13.7111 expand on 0.555556 -13.7115 avg/max 0.75 -13.7116 0.873 0.75 -13.7116 slots) 0.75 -13.7116 sub-lattice 0.75 -13.7116 rivals 0.75 -13.7116 rel-type = 0.75 -13.7116 hits) 0.75 -13.7116 (approved 0.75 -13.7116 signifying 0.75 -13.7116 removal, 0.75 -13.7116 hear, 0.75 -13.7116 wordnet (narayan 0.75 -13.7116 prétexte 0.75 -13.7116 examiner’s 0.75 -13.7116 adjudicators 0.75 -13.7116 wsj-trained 0.75 -13.7116 lebanon, 0.75 -13.7116 adjectives), 0.75 -13.7116 55.16 0.75 -13.7116 traditional bag-of-words 0.75 -13.7116 entity-pair 0.368421 -13.7118 a. smith 0.0852713 -13.712 problem. 0.555556 -13.7122 ntcir-8 0.103448 -13.7123 they can 0.75 -13.7125 ortega, 0.75 -13.7125 #1, 0.75 -13.7125 λ; 0.75 -13.7125 higgins 0.75 -13.7125 scorers, 0.75 -13.7125 88.28 0.75 -13.7125 0.061 0.75 -13.7125 forj 0.75 -13.7125 0.533 0.75 -13.7125 macroaveraged 0.75 -13.7125 0.060 0.75 -13.7125 win2 0.0435657 -13.7125 compared 0.461538 -13.7125 songs 0.5 -13.7127 hai 0.666667 -13.713 double entendre 0.75 -13.7134 ‘noisy’ 0.75 -13.7134  breading 0.75 -13.7134 (abdul-rauf 0.75 -13.7134 684 0.75 -13.7134 koster 0.75 -13.7134 ped06 0.75 -13.7134 40.63 0.75 -13.7134 speech-based 0.75 -13.7134 perceiving 0.75 -13.7134 advertisement 0.75 -13.7134 calea 0.75 -13.7134 masi 0.75 -13.7134 accumulative 0.75 -13.7134 marx, 0.75 -13.7134 laisr, 0.75 -13.7134 (novak 0.75 -13.7134 26.95 0.75 -13.7134 discovered, 0.75 -13.7134 authors; 0.75 -13.7134 ocr/hr 0.75 -13.7134 sugar 0.75 -13.7134 changes; 0.666667 -13.7138 〈 0.571429 -13.7138 atlas 0.555556 -13.7141 cooked 0.75 -13.7143 grammaticality; 0.75 -13.7143 reporters 0.75 -13.7143 44.59 0.75 -13.7143 “no” 0.75 -13.7143 2004b; 0.75 -13.7143 nat 0.75 -13.7143 grandparents 0.75 -13.7143 regression equation 0.75 -13.7143 commas, 0.75 -13.7143 extremes: 0.75 -13.7143 0.87, 0.75 -13.7143 ourc 0.75 -13.7143 turkdev 0.75 -13.7143 turkish) 0.75 -13.7143 pus 0.75 -13.7143 ˇa 0.555556 -13.7148 (schler et 0.23913 -13.7148 turn. 0.75 -13.7151 hierarchy) 0.75 -13.7151 rudimentary 0.75 -13.7151 mainly, 0.75 -13.7151 vnp-a 0.75 -13.7151 26.70 0.75 -13.7151 ephemeral 0.75 -13.7151 sep 0.75 -13.7151 orthogonal, 0.75 -13.7151 paths; 0.75 -13.7151 94.44 0.75 -13.7151 4.55 0.75 -13.7151 “coordination” 0.75 -13.7151 (wsd), 0.75 -13.7151 answer) 0.75 -13.7151 needed), 0.75 -13.7151 3-grams, 0.571429 -13.7152 semeval2007 0.290323 -13.7156 (vp 0.235294 -13.7156 — — 0.226415 -13.7157 have high 0.428571 -13.7159 all levels 0.571429 -13.7159 some authors 0.75 -13.716 [1,8] 0.75 -13.716 estimated via 0.75 -13.716 28.44 0.75 -13.716 0.241 0.75 -13.716 11k 0.75 -13.716 far below 0.75 -13.716 (aronson, 0.75 -13.716 replied 0.75 -13.716 reduce, 0.75 -13.716 (keeping 0.75 -13.716 collocation error correction 0.75 -13.716 17.0 0.75 -13.716 syntax+feat. 0.75 -13.716 brent’s 0.75 -13.716 (allen 0.75 -13.716 pre-annotation 0.75 -13.716 events; 0.75 -13.716 preemptive 0.666667 -13.7161 φ′ 0.666667 -13.7161 1994, 0.5 -13.7163 brier 0.571429 -13.7165 414-letter 0.189189 -13.7168 every word 0.5 -13.7168 in” 0.75 -13.7169 registered, 0.75 -13.7169 samuel 0.75 -13.7169 (joint) 0.75 -13.7169 ravenclaw 0.75 -13.7169 280k 0.75 -13.7169 significant) 0.75 -13.7169 performance) 0.75 -13.7169 mpqa, 0.75 -13.7169 teleportation 0.75 -13.7169 post-editor 0.75 -13.7169 view; 0.75 -13.7169 1995); 0.75 -13.7169 0.844 0.75 -13.7169 53.5% 0.75 -13.7169 endemic 0.75 -13.7169 ramage 0.75 -13.7169 alteration 0.75 -13.7169 privileged 0.75 -13.7169 compact(s) 0.75 -13.7169 sub-problems 0.75 -13.7169 socialization 0.172043 -13.7172 annotated as 0.666667 -13.7177 s(y) 0.75 -13.7178 t-table 0.75 -13.7178 postagging, 0.75 -13.7178 (bunt, 0.75 -13.7178 lhpa 0.75 -13.7178 sieve, 0.75 -13.7178 abusing 0.75 -13.7178 (wang, 0.75 -13.7178 toolkit2 0.75 -13.7178 ‘f’ 0.75 -13.7178 (outside 0.75 -13.7178 scored, 0.75 -13.7178 townshend, 0.75 -13.7178 0.269 0.75 -13.7178 23,222 0.75 -13.7178 escaping 0.75 -13.7178 make2 0.75 -13.7178 conect 0.571429 -13.7179 0.264 0.32 -13.7179 jun’ichi 0.461538 -13.718 necessary because 0.666667 -13.7185 dg, 0.153846 -13.7185 klein, 0.571429 -13.7186 company) 0.75 -13.7187 flow, 0.75 -13.7187 immediately, 0.75 -13.7187 supervised: 0.75 -13.7187 intersects 0.75 -13.7187 regression-based 0.75 -13.7187 biand 0.75 -13.7187 3.22 0.75 -13.7187 mounted 0.75 -13.7187 tems 0.75 -13.7187 virtual–virtual 0.75 -13.7187 words”, 0.75 -13.7187 31.52 0.75 -13.7187 (immediate) 0.75 -13.7187 carry, 0.75 -13.7187 “look” 0.75 -13.7187 native arabic 0.75 -13.7187 (ambati 0.75 -13.7187 “line” 0.75 -13.7187 present[verbal] 0.310345 -13.7191 derivations for 0.310345 -13.7191 relative position 0.666667 -13.7193 〈a, 0.75 -13.7196 (aac) 0.75 -13.7196 bio-crf-h 0.75 -13.7196 sciknowmine 0.75 -13.7196 eduardo 0.75 -13.7196 adult, 0.75 -13.7196 necessitates 0.75 -13.7196 vignettes, 0.75 -13.7196 bi-parsing 0.75 -13.7196 pan, 0.75 -13.7196 tag), 0.75 -13.7196 scoping, 0.75 -13.7196 (barrachina 0.75 -13.7196 mwadv 0.75 -13.7196 lowresource 0.75 -13.7196 penalty), 0.75 -13.7196 q(x), 0.75 -13.7196 segment) 0.75 -13.7205 80%, 0.75 -13.7205 mbytes 0.75 -13.7205 t” 0.75 -13.7205 “local” 0.75 -13.7205 freitag 0.75 -13.7205 been omitted, 0.75 -13.7205 encouragingly, 0.75 -13.7205 conjugacy 0.75 -13.7205 perspectives: 0.75 -13.7205 0.512 0.75 -13.7205 press) 0.75 -13.7205 66.20 0.75 -13.7205 (shawetaylor 0.75 -13.7205 altlex 0.75 -13.7205 “within 0.75 -13.7205 daunting 0.75 -13.7205 0.478 0.75 -13.7205 (k), 0.75 -13.7205 retweeted 0.75 -13.7205 invention 0.75 -13.7205 816 0.75 -13.7214 (1973) 0.75 -13.7214 iran, 0.75 -13.7214 “bad”) 0.75 -13.7214 roque 0.75 -13.7214 unavoidable 0.75 -13.7214 decoder(s) 0.75 -13.7214 seller 0.75 -13.7214 khephi 0.75 -13.7214 0.998 0.75 -13.7214 ‘merge’ 0.75 -13.7214 cns, 0.75 -13.7214 acyclic, 0.75 -13.7214 29.06 0.75 -13.7214 429 0.75 -13.7214 exclusive, 0.555556 -13.722 anderson, 0.571429 -13.722 (2b), 0.75 -13.7222 rel-lda, 0.75 -13.7222 j.d. 0.75 -13.7222 digest 0.75 -13.7222 zhong 0.75 -13.7222 sew 0.75 -13.7222 ddts 0.75 -13.7222 sandy 0.75 -13.7222 hedging, 0.75 -13.7222 0.881 0.75 -13.7222 pullum, 0.75 -13.7222 joshua, 0.75 -13.7222 multi-category 0.75 -13.7222 uper 0.75 -13.7222 males, 0.75 -13.7222 0.448 0.75 -13.7222 number), 0.75 -13.7222 complementizers 0.75 -13.7222 reformulation, 0.75 -13.7222 stimulating 0.5 -13.7222 heat 0.5 -13.7222 gd(s) 0.666667 -13.7225 p(yi;λ) 0.571429 -13.7226 (haveliwala, 0.555556 -13.7227 professionally 0.347826 -13.7228 (mwe 2011), 0.5 -13.7228 d3 0.226415 -13.7229 calculated by 0.388889 -13.7231 ceafe 0.75 -13.7231 59.6% 0.75 -13.7231 deprel 0.75 -13.7231 two-types 0.75 -13.7231 rearrange 0.75 -13.7231 performers 0.75 -13.7231 fledged 0.75 -13.7231 beam settings 0.75 -13.7231 hostage 0.75 -13.7231 nuggets, 0.75 -13.7231 86.02 0.75 -13.7231 aziz 0.75 -13.7231 (2007): 0.75 -13.7231 forensic 0.75 -13.7231 reconciled 0.75 -13.7231 7-point 0.75 -13.7231 news-c, 0.75 -13.7231 come 0.75 -13.7231 non-monotonic 0.75 -13.7231 conroy, 0.75 -13.7231 47.47 0.571429 -13.7233 controversy 0.5 -13.7234 (monolingual) 0.461538 -13.7235 es, 0.666667 -13.724 m), 0.75 -13.724 boussidan 0.75 -13.724 0.387 0.75 -13.724 normal, 0.75 -13.724 4-point 0.75 -13.724 1536 0.75 -13.724 (pearson’s 0.75 -13.724 26.99 0.75 -13.724 0.126 0.75 -13.724 1,865 0.75 -13.724 (ic) 0.75 -13.724 60.13 0.75 -13.724 765 0.5 -13.724  in 0.75 -13.7249 32.8% 0.75 -13.7249 doublet 0.75 -13.7249 text-messages 0.75 -13.7249 dimension-reduction 0.75 -13.7249 140-character 0.75 -13.7249 mistaking 0.75 -13.7249 0.152 0.75 -13.7249 mem 0.75 -13.7249 anticipation, 0.75 -13.7249 blair 0.75 -13.7249 genders, 0.75 -13.7249 part: 0.75 -13.7249 (rimell 0.75 -13.7249 standing, 0.75 -13.7249 layouts 0.310345 -13.725 s-factor 0.040493 -13.7252 obtained 0.571429 -13.7254 brothers grimm 0.571429 -13.7254 r(β) 0.666667 -13.7256 sabine 0.75 -13.7258 0.935 0.75 -13.7258 face) 0.75 -13.7258 district, 0.75 -13.7258 82.48 0.75 -13.7258 elections, 0.75 -13.7258 timeml, 0.75 -13.7258 p1234ibm1 0.75 -13.7258 biologists 0.75 -13.7258 shalow 0.75 -13.7258 unchanged, 0.75 -13.7258 pre-specify 0.75 -13.7258 fails when 0.75 -13.7258 26.09 0.75 -13.7258 underwent 0.75 -13.7258 hm_wc 0.75 -13.7258 60.17 0.75 -13.7258 v-obj 0.368421 -13.7262 three possible 0.666667 -13.7264 776 0.0935484 -13.7266 are given 0.00569443 -13.7266 a 0.75 -13.7267 97.8% 0.75 -13.7267 sound, 0.75 -13.7267 dienes 0.75 -13.7267 wasp 0.75 -13.7267 sequence-labeling 0.75 -13.7267 (a: 0.75 -13.7267 table7: 0.75 -13.7267 0.559 0.75 -13.7267 3.18 0.75 -13.7267 4.5, 0.75 -13.7267 toral 0.75 -13.7267 tablet 0.75 -13.7267 pdf, 0.75 -13.7267 created; 0.75 -13.7267 top-best 0.75 -13.7267 detokenization 0.75 -13.7267 guidelines: 0.35 -13.7269 (dyer et al., 0.666667 -13.7272 interrupting 0.125 -13.7275 we note 0.75 -13.7276 minima 0.75 -13.7276 (belonging 0.75 -13.7276 26.17 0.75 -13.7276 edges: 0.75 -13.7276 (correlation 0.75 -13.7276 2-10 0.75 -13.7276 1while 0.75 -13.7276 raters, 0.75 -13.7276 abelson, 0.75 -13.7276 sense-targeted 0.75 -13.7276 1.71 0.75 -13.7276 ill-defined 0.75 -13.7276 zerogram 0.75 -13.7276 right” 0.75 -13.7276 idan 0.75 -13.7276 (reiter 0.75 -13.7276 cssdb 0.75 -13.7276 (dop) 0.75 -13.7276 82.25 0.75 -13.7276 occ(t1) 0.75 -13.7276 prendinger, 0.106996 -13.7277 comes 0.388889 -13.7278 cross-event 0.213115 -13.7279 b = 0.75 -13.7285 (kaplan 0.75 -13.7285 intra-date 0.75 -13.7285 esa 0.75 -13.7285 offshore 0.75 -13.7285 rescale 0.75 -13.7285 warrants 0.75 -13.7285 #iis-0811745 0.75 -13.7285 sexual 0.75 -13.7285 alignment) 0.75 -13.7285 (ip) 0.75 -13.7285 (chomsky, 0.75 -13.7285 clitic-segmented 0.75 -13.7285 provision 0.75 -13.7285 task-dependent 0.75 -13.7285 panther 0.75 -13.7285 daille 0.75 -13.7285 dynamical 0.75 -13.7285 (osborne 0.75 -13.7285 textual, 0.75 -13.7285 non-normalized 0.75 -13.7285 concretely 0.75 -13.7285 0r6 0.0307272 -13.7287 the first 0.0944625 -13.7292 examine 0.555556 -13.7293 +β 0.5 -13.7293 (p1,p2) 0.3 -13.7294 to favor 0.75 -13.7294 naacl-hlt 0.75 -13.7294 reviewers’ 0.75 -13.7294 optimum, 0.75 -13.7294 accomplishes 0.75 -13.7294 alike, 0.75 -13.7294 choudhury 0.75 -13.7294 32.6% 0.75 -13.7294 subspans 0.75 -13.7294 edges; 0.75 -13.7294 << 0.75 -13.7294 possessor 0.75 -13.7294 type-2 0.75 -13.7294 matlab 0.75 -13.7294 plc 0.75 -13.7294 22.0 0.75 -13.7294 (bosco 0.75 -13.7294 58.38 0.75 -13.7294 realignment 0.75 -13.7294 43.88 0.75 -13.7294 (mcdonald,crammer,and 0.75 -13.7294 (chu 0.75 -13.7294 satisfying, 0.75 -13.7294 tricky 0.75 -13.7294 n− 0.75 -13.7294 29.65 0.75 -13.7294 mteval 0.75 -13.7294 excerpts, 0.75 -13.7294 (lang 0.571429 -13.7294 sysptm 0.388889 -13.7296 liblinear 0.140741 -13.7298 it is also 0.461538 -13.7301 peking 0.571429 -13.7301 percent points 0.75 -13.7303 locative, 0.75 -13.7303 cosinesimilarity 0.75 -13.7303 campbell 0.75 -13.7303 gender/number 0.75 -13.7303 bahasa 0.75 -13.7303 monologue/dialogue 0.75 -13.7303 0.75 -13.7303 nulllog(p) 0.75 -13.7303 sequence’ 0.75 -13.7303 tager 0.75 -13.7303 web-crawled 0.75 -13.7303 hypothesizes 0.75 -13.7303 59.9% 0.75 -13.7303 570 0.75 -13.7303 occurrences) 0.75 -13.7303 hour, 0.75 -13.7303 semimarkov 0.75 -13.7303 hs03 0.75 -13.7303 schmidt 0.5 -13.7305 jun 0.555556 -13.7307 ng. 0.0932476 -13.7307 sentences from 0.428571 -13.7307 features describing 0.571429 -13.7308 46.0 0.5 -13.7311 pairwise agreement 0.75 -13.7312 15.7% 0.75 -13.7312 via direct 0.75 -13.7312 3.51 0.75 -13.7312 p; 0.75 -13.7312 languagedependent 0.75 -13.7312 nonconcatenative 0.75 -13.7312 sentences’ 0.75 -13.7312 fairness 0.75 -13.7312 1(a)) 0.75 -13.7312 self-confidence 0.75 -13.7312 567 0.75 -13.7312 wh-word 0.75 -13.7312 20.3 0.571429 -13.7315 (grefenstette, 0.555556 -13.732 empirically-derived 0.555556 -13.732 koo, 0.75 -13.7321 0.022 0.75 -13.7321 26.67 0.75 -13.7321 3.36 0.75 -13.7321 english–chinese 0.75 -13.7321 carriers 0.75 -13.7321 cuts, 0.75 -13.7321 edison 0.75 -13.7321 occured 0.75 -13.7321 ang 0.75 -13.7321 (partly 0.75 -13.7321 25) 0.75 -13.7321 non-eos 0.75 -13.7321 phonemically 0.75 -13.7321 system-produced 0.4 -13.7321 zweigenbaum, 0.571429 -13.7322 het 0.666667 -13.7328 bus, 0.75 -13.733 (many 0.75 -13.733 (abney, 0.75 -13.733 asahara 0.75 -13.733 62.16 0.75 -13.733 parliament, 0.75 -13.733 “list 0.75 -13.733 caller’s 0.75 -13.733 (type 0.75 -13.733 scarce, 0.75 -13.733 class-level 0.75 -13.733 want, 0.75 -13.733 consensual 0.75 -13.733 coalesce 0.75 -13.733 share, 0.75 -13.733 kernels: 0.75 -13.733 reanalyze 0.75 -13.733 (read, 0.75 -13.733 t:±0 0.75 -13.733 84.2% 0.75 -13.733 you”, 0.333333 -13.7332 experimental results, 0.5 -13.7335 that-less 0.666667 -13.7336 diana 0.666667 -13.7336 recalling 0.571429 -13.7336 (katz 0.571429 -13.7336 time vs. 0.75 -13.7339 w; 0.75 -13.7339 (knn) 0.75 -13.7339 25.1 0.75 -13.7339 haque 0.75 -13.7339 sahlgren, 0.75 -13.7339 event-related 0.75 -13.7339 confinement, 0.75 -13.7339 irstlm’s 0.75 -13.7339 10.1% 0.75 -13.7339 non-letter 0.75 -13.7339 abed, 0.75 -13.7339 humanwritten 0.571429 -13.7342 =summationdisplay 0.203125 -13.7343 been successfully 0.666667 -13.7344 458 0.666667 -13.7344 dersc 0.461538 -13.7345 linking. 0.75 -13.7348 (patb) 0.75 -13.7348 (gr) 0.75 -13.7348 “stable” 0.75 -13.7348 de: 0.75 -13.7348 trained: 0.75 -13.7348 mpl 0.75 -13.7348 unmanageable 0.75 -13.7348 19.2m 0.75 -13.7348 (nsf) 0.75 -13.7348 (european 0.75 -13.7348 r” 0.75 -13.7348 long-forms, 0.75 -13.7348 synsets), 0.75 -13.7348 issues; 0.75 -13.7348 uzuner 0.75 -13.7348 vector; 0.75 -13.7348 internally, 0.75 -13.7348 giza-weak 0.75 -13.7348 csomai, 0.75 -13.7348 fom, 0.75 -13.7348 flavour 0.571429 -13.7349 339 0.35 -13.7349 valitutti, 0.388889 -13.7357 (1998). 0.75 -13.7357 attractive, 0.75 -13.7357 reimer 0.75 -13.7357 treebank: 0.75 -13.7357 rij 0.75 -13.7357 human-translated 0.75 -13.7357 tree-adjoining 0.75 -13.7357 in(in) 0.75 -13.7357 54.35% 0.75 -13.7357 hewavitharana 0.75 -13.7357 normalizer, 0.75 -13.7357 (subject) 0.75 -13.7357 det) 0.75 -13.7357 patches 0.75 -13.7357 38.82 0.75 -13.7357 dependents, 0.75 -13.7357 (subsection 0.75 -13.7357 timeline, 0.571429 -13.7363 genie 0.5 -13.7365 .92 0.75 -13.7366 1875 0.75 -13.7366 ak, 0.75 -13.7366 1970; 0.75 -13.7366 age-of-acquisition 0.75 -13.7366 52.28 0.75 -13.7366 word-dependent 0.75 -13.7366 (levinson, 0.75 -13.7366 morin, 0.75 -13.7366 subcategorisation 0.75 -13.7366 51.1 0.75 -13.7366 phrase-pair, 0.75 -13.7366 t:±1” 0.75 -13.7366 drug, 0.75 -13.7366 lang=[eng, 0.75 -13.7366 multi-stack 0.75 -13.7366 mori 0.571429 -13.737 0.861 0.571429 -13.737 test11 0.75 -13.7375 langford 0.75 -13.7375 (in) 0.75 -13.7375 posted, 0.75 -13.7375 tsur’s 0.75 -13.7375 b(y) 0.75 -13.7375 circumplex 0.75 -13.7375 2.26 0.75 -13.7375 0.359 0.75 -13.7375 65.50 0.75 -13.7375 capability, 0.75 -13.7375 decomposition: 0.75 -13.7375 (nakazawa 0.75 -13.7375 jaeger 0.75 -13.7375 (54%) 0.75 -13.7375 55.11 0.75 -13.7375 lvms 0.75 -13.7375 (taken 0.75 -13.7375 (schrijver, 0.75 -13.7375 disputant‟s 0.75 -13.7375 bergmann, 0.75 -13.7375 micro-accuracy 0.75 -13.7375 sexiness 0.75 -13.7375 18.58 0.75 -13.7375 5s 0.75 -13.7375 14.81 0.75 -13.7375 lin-kernighan 0.0979021 -13.7375 so we 0.666667 -13.7376 0.924 0.666667 -13.7376 much bigger 0.571429 -13.7377 dang 0.571429 -13.7377 0.913 0.368421 -13.7377 words (e.g. 0.5 -13.7383 ψi 0.75 -13.7384 936 0.75 -13.7384 ntp 0.75 -13.7384 reversed, 0.75 -13.7384 client-server 0.75 -13.7384 pof 0.75 -13.7384 prh 0.75 -13.7384 “closed 0.75 -13.7384 unregistered 0.75 -13.7384 negri 0.75 -13.7384 others; 0.75 -13.7384 preserver 0.75 -13.7384 es+pc 0.75 -13.7384 reserving 0.75 -13.7384 word-by-document 0.75 -13.7384 rivers, 0.75 -13.7384 counts; 0.75 -13.7384 herein, 0.75 -13.7384 (ringger 0.75 -13.7384 insights, 0.75 -13.7384 (diab 0.75 -13.7384 tommaso 0.75 -13.7384 quant 0.75 -13.7384 (floyd 0.75 -13.7384 92.55 0.75 -13.7384 discretizing 0.571429 -13.7384 0.018 0.5 -13.739 rec% 0.75 -13.7393 xml, 0.75 -13.7393 subexpression 0.75 -13.7393 iron 0.75 -13.7393 doing, 0.75 -13.7393 truth, 0.75 -13.7393 62.72 0.75 -13.7393 bouamor 0.75 -13.7393 suporting 0.75 -13.7393 liakata 0.75 -13.7393 newcastle 0.75 -13.7393 2our 0.75 -13.7393 svm-all 0.75 -13.7393 anotating 0.75 -13.7393 perplexities, 0.75 -13.7393 pre-written 0.75 -13.7393 turbo 0.75 -13.7393 microbloggers 0.75 -13.7393 (dd) 0.75 -13.7393 ∀i,j 0.75 -13.7393 1.45 0.75 -13.7393 argm-mns 0.75 -13.7393 synsem 0.388889 -13.7394 precedesequal 0.145161 -13.7394 (kim 0.5 -13.7396 further development 0.571429 -13.7398 48.2 0.32 -13.7401 were labeled 0.75 -13.7402 unrealized 0.75 -13.7402 (mlns) 0.75 -13.7402 unadapted 0.75 -13.7402 ei,t 0.75 -13.7402 0.581 0.75 -13.7402 (hpsg) 0.75 -13.7402 wilks 0.75 -13.7402 phraseological 0.75 -13.7402 androutsopoulos, 0.75 -13.7402 wordform, 0.75 -13.7402 connectionist 0.75 -13.7402 ccl, 0.75 -13.7402 unwilling 0.75 -13.7402 regularities, 0.75 -13.7402 (la) 0.75 -13.7402 2003)), 0.75 -13.7402 qs1 0.75 -13.7402 (french) 0.75 -13.7402 dafter 0.571429 -13.7405 ride 0.571429 -13.7405 jj) 0.75 -13.7411 plas 0.75 -13.7411 8as 0.75 -13.7411 one-time 0.75 -13.7411 ovs. 0.75 -13.7411 randomly chooses 0.75 -13.7411 jos 0.75 -13.7411 posttest 0.75 -13.7411 individualized 0.75 -13.7411 40.98 0.75 -13.7411 recomputed 0.75 -13.7411 corpora’s 0.75 -13.7411 maximal, 0.75 -13.7411 bitter 0.75 -13.7411 0.185 0.75 -13.7411 large-sized 0.75 -13.7411 z-normalized 0.75 -13.7411 ablative 0.75 -13.7411 americanos 0.75 -13.7411 14,000 0.75 -13.7411 turninitial 0.75 -13.7411 helps, 0.5 -13.7414 elementary structure 0.194444 -13.7415 a proper 0.666667 -13.7417 nist’08 0.666667 -13.7417 firing 0.75 -13.742 arnold, 0.75 -13.742 who, 0.75 -13.742 hardest 0.75 -13.742 occasion 0.75 -13.742 (sangati 0.75 -13.742 shao 0.75 -13.742 0.632 0.75 -13.742 parenthesed 0.75 -13.742 γ) 0.75 -13.742 explained, 0.75 -13.742 o-based 0.75 -13.742 np; 0.75 -13.742 four-way 0.75 -13.742 (shieber, 0.75 -13.742 language-noise 0.75 -13.742 4.47 0.75 -13.742 93.5% 0.75 -13.742 87.58 0.75 -13.742 graceful 0.75 -13.742 (3)), 0.75 -13.742 43.65 0.75 -13.742 features.3 0.75 -13.742 two-argument 0.75 -13.742 41.84 0.75 -13.742 andersen 0.75 -13.742 neither, 0.368421 -13.7426 translation output. 0.5 -13.7426 annotated name 0.75 -13.743 p < .001. 0.75 -13.743 nullnullnullnullnullnullnull, 0.75 -13.743 average), 0.75 -13.743 38.87 0.75 -13.743 (duda 0.75 -13.743 +3 0.75 -13.743 well-typedness 0.75 -13.743 11for 0.75 -13.743 raw, 0.75 -13.743 vectore 0.75 -13.743 rajeev 0.75 -13.743 contexts; 0.75 -13.743 transliterator 0.75 -13.743 48.36 0.75 -13.743 1a), 0.75 -13.743 darmstadt 0.75 -13.743 friends’ 0.75 -13.743 discourse-constrained 0.75 -13.743 +det+png+lmm 0.75 -13.743 haitian-cr´eole 0.75 -13.743 edge-factored 0.75 -13.743 languageindependent 0.571429 -13.7432 concordances 0.75 -13.7439 deemter 0.75 -13.7439 (2004)’s 0.75 -13.7439 particularly hard 0.75 -13.7439 (jurgens 0.75 -13.7439 0.177 0.75 -13.7439 batches, 0.75 -13.7439 non-tailored 0.75 -13.7439 schatzmann, 0.75 -13.7439 4.19 0.75 -13.7439 613 0.75 -13.7439 (worst) 0.75 -13.7439 principally 0.75 -13.7439 uploading 0.75 -13.7439 0.672 0.75 -13.7439 eruption 0.75 -13.7439 fluent, 0.75 -13.7439 i∈i 0.75 -13.7439 )), 0.75 -13.7448 (bg) 0.75 -13.7448 (gerber 0.75 -13.7448 gis 0.75 -13.7448 bilda 0.75 -13.7448 periphrastic 0.75 -13.7448 departments 0.75 -13.7448 207, 0.75 -13.7448 lowlevel 0.75 -13.7448 carries over 0.75 -13.7448 i-hmm 0.75 -13.7448 kneser-ney, 0.75 -13.7448 returns, 0.75 -13.7448 “location” 0.75 -13.7448 zufferey 0.75 -13.7448 jeju 0.75 -13.7448 0.281 0.75 -13.7448 minimum span 0.75 -13.7448 incredibly 0.75 -13.7448 arrived, 0.75 -13.7448 9.6% 0.666667 -13.7457 korhonen, 0.666667 -13.7457 dubbed 0.666667 -13.7457 length (in 0.75 -13.7457 sub-graph 0.75 -13.7457 (ratinov 0.75 -13.7457 (**) 0.75 -13.7457 matsuo 0.75 -13.7457 *c 0.75 -13.7457 file) 0.75 -13.7457 typed predicates. 0.75 -13.7457 hypertext 0.75 -13.7457 recal, 0.75 -13.7457 supertypes 0.75 -13.7457 syntactic: 0.75 -13.7457 1although 0.75 -13.7457 system.1 0.75 -13.7457 portray 0.75 -13.7457 p(f) 0.75 -13.7457 zi,j 0.428571 -13.7457 (hofmann, 0.333333 -13.7457 scenes. 0.368421 -13.7457 and john 0.571429 -13.746 (white, 0.75 -13.7466 2.70 0.75 -13.7466 cprimei 0.75 -13.7466 68.52 0.75 -13.7466 forst, 0.75 -13.7466 68.10 0.75 -13.7466 multi-point 0.75 -13.7466 suspicious 0.75 -13.7466 530 0.75 -13.7466 61.85 0.75 -13.7466 (1.0) 0.75 -13.7466 27.79 0.75 -13.7466 arsenal 0.75 -13.7466 denominal 0.75 -13.7466 unanticipated 0.75 -13.7466 en–fr 0.75 -13.7466 0.868 0.368421 -13.7466 *. 0.4 -13.7467 48.5 0.571429 -13.7467 flyslip 0.27027 -13.7468 task 1 0.0978261 -13.7473 translations. 0.571429 -13.7474 wikipedia policy 0.5 -13.7474 “n 0.75 -13.7475 nonparametrics 0.75 -13.7475 history2 0.75 -13.7475 alpha-beta 0.75 -13.7475 (45 0.75 -13.7475 (parton 0.75 -13.7475 fancy 0.75 -13.7475 3,072 0.75 -13.7475 centuries, 0.75 -13.7475 dimensionalities 0.75 -13.7475 synset) 0.75 -13.7475 (setup 0.75 -13.7475 reconsider 0.75 -13.7484 mk, 0.75 -13.7484 “hard” 0.75 -13.7484 (venkataraman, 0.75 -13.7484 (am) 0.75 -13.7484 (0.7) 0.75 -13.7484 bleu_w 0.75 -13.7484 mmd 0.75 -13.7484 confidence-based 0.75 -13.7484 binarizations 0.75 -13.7484 (varga 0.75 -13.7484 ghost 0.75 -13.7484 less-than-a-day 0.75 -13.7484 intensively 0.75 -13.7484 3.97 0.75 -13.7484 neglect 0.75 -13.7484 word-class 0.75 -13.7484 m-dimensional 0.75 -13.7484 (star) 0.75 -13.7484 sampling: 0.75 -13.7484 roughly, 0.75 -13.7484 cn-lvcs, 0.75 -13.7484 0.577 0.75 -13.7484 (log) 0.27027 -13.7484 chosen from 0.135135 -13.7489 figure 4 0.666667 -13.749 bucket, 0.666667 -13.749 campus 0.75 -13.7493 plussyn 0.75 -13.7493 co-ranking 0.75 -13.7493 (temporal, 0.75 -13.7493 log-counts 0.75 -13.7493 (kamps 0.75 -13.7493 non-ambiguous 0.75 -13.7493 mvm, 0.75 -13.7493 vectorz 0.75 -13.7493 (zou 0.75 -13.7493 hovering 0.75 -13.7493 (null2) 0.75 -13.7493 specific) 0.75 -13.7493 vectory 0.75 -13.7493 kedis 0.75 -13.7493 preventer 0.75 -13.7493 tag-based 0.75 -13.7493 underquantification 0.75 -13.7493 topdown 0.75 -13.7493 hypothesise 0.75 -13.7493 jacinto 0.5 -13.7493 (mle) 0.290323 -13.7494 x-axis 0.571429 -13.7495 transactional 0.571429 -13.7502 wallach 0.571429 -13.7502 y), 0.75 -13.7503 s), 0.75 -13.7503 scov 0.75 -13.7512 subordinated 0.75 -13.7512 light, 0.75 -13.7512 eshmm 0.571429 -13.7516 expected loss 0.5 -13.7517 ˆs 0.0434478 -13.7518 the second 0.75 -13.7521 intersectionsq 0.75 -13.7521 blowout 0.75 -13.7521 listed, 0.75 -13.7521 aesthetic 0.75 -13.7521 annotationof 0.75 -13.7521 lungs 0.75 -13.7521 (unless 0.75 -13.7521 825 0.75 -13.7521 install 0.75 -13.7521 3.83 0.75 -13.7521 aligners, 0.75 -13.7521 antei 0.75 -13.7521 61.40 0.75 -13.7521 chinese-toenglish 0.75 -13.7521 (standard 0.666667 -13.7522 insert no 0.571429 -13.7523 ie-based 0.226415 -13.7525 by simply 0.454545 -13.7526 point at 0.454545 -13.7526 f(%) 0.184211 -13.753 concludes 0.75 -13.753 state” 0.75 -13.753 661 0.75 -13.753 84.26 0.75 -13.753 predicate) 0.75 -13.753 kn-rnn-lim-all 0.75 -13.753 (verbs 0.75 -13.753 attracts 0.75 -13.753 44.44 0.75 -13.753 cronbach’s 0.75 -13.753 1-vs-all 0.75 -13.753 tonight 0.75 -13.753 manli’s 0.75 -13.753 pii 0.75 -13.753 eliminated, 0.75 -13.753 mpp 0.75 -13.753 errors), 0.75 -13.753 24.58 0.75 -13.753 charniak’s 0.75 -13.753 tentatively 0.75 -13.753 1+2+3+4+5 0.75 -13.753 97.8 0.571429 -13.753 compliment 0.571429 -13.753 attacking 0.3 -13.7534 range. 0.5 -13.7536 contextual classifier’s 0.5 -13.7536 urbana-champaign 0.454545 -13.7537 (κ) 0.571429 -13.7537 and schlangen, 0.666667 -13.7538 5.1), 0.28125 -13.7538 arising 0.75 -13.7539 dp, 0.75 -13.7539 •the 0.75 -13.7539 prevailing 0.75 -13.7539 0.975 0.75 -13.7539 sensory 0.75 -13.7539 levels; 0.75 -13.7539 10-15 0.75 -13.7539 4.000 0.75 -13.7539 (wfst) 0.75 -13.7539 unclear, 0.75 -13.7539 linearize 0.75 -13.7539 (hakkani-t¨ur 0.75 -13.7539 (neither 0.75 -13.7539 suppose, 0.75 -13.7539 t31 0.75 -13.7539 36.36 0.75 -13.7539 8k 0.75 -13.7539 side’s 0.75 -13.7539 nonfiction 0.75 -13.7539 419 0.75 -13.7539 notion, 0.75 -13.7539 “verb 0.75 -13.7539 gibson, 0.206349 -13.7544 these systems 0.428571 -13.7546 lost. 0.454545 -13.7548 70k 0.0879765 -13.7548 dataset. 0.75 -13.7549 pst) 0.75 -13.7549 gradual 0.75 -13.7549 neg, 0.75 -13.7549 null3, 0.75 -13.7549 eigen 0.75 -13.7549 unconventional 0.75 -13.7549 leader 0.75 -13.7549 manipulation, 0.75 -13.7549 bordering 0.75 -13.7549 marched 0.75 -13.7549 modalities, 0.75 -13.7549 (near 0.75 -13.7549 (thompson 0.75 -13.7549 semantics) 0.75 -13.7549 israel, 0.75 -13.7549 we regularize 0.75 -13.7549 83.9 0.75 -13.7549 (bleu) 0.75 -13.7549 straight, 0.75 -13.7549 “inverse 0.666667 -13.7555 table1 0.75 -13.7558 affordances 0.75 -13.7558 0.489 0.75 -13.7558 (taylor, 0.75 -13.7558 irelevant 0.75 -13.7558 tweeters 0.75 -13.7558 (kilgarriff 0.75 -13.7558 admittedly 0.75 -13.7558 3.5.2 0.75 -13.7558 0.9] 0.75 -13.7558 model-theoretic 0.75 -13.7558 garbage, 0.75 -13.7558 (anova) 0.75 -13.7558 kallmeyer 0.75 -13.7558 10−8 0.75 -13.7558 tagsets, 0.571429 -13.7558 p(zi 0.571429 -13.7558 48.7 0.571429 -13.7558 iterative procedure 0.666667 -13.7563 sampo pyysalo∗ 0.4 -13.7564 cluster-level 0.5 -13.7566 “left” 0.75 -13.7567 jin, 0.75 -13.7567 pantic, 0.75 -13.7567 abstraction%1:03:00 0.75 -13.7567 star-free 0.75 -13.7567 de→en 0.75 -13.7567 granger, 0.75 -13.7567 hastie 0.75 -13.7567 machine-translation 0.75 -13.7567 1092 0.75 -13.7567 memorial 0.75 -13.7567 appendix, 0.75 -13.7567 74.9% 0.75 -13.7567 objectives: 0.75 -13.7567 intentions, 0.75 -13.7567 87.82 0.307692 -13.7568 a constrained 0.666667 -13.7571 bundles 0.571429 -13.7572 40.4 0.571429 -13.7572 work assumes 0.75 -13.7576 1.06 0.75 -13.7576 0.791 0.75 -13.7576 thomson, 0.75 -13.7576 a1: 0.75 -13.7576 22), 0.75 -13.7576 rate; 0.75 -13.7576 sdp, 0.75 -13.7576 pair), 0.75 -13.7576 holonym 0.75 -13.7576 808 0.75 -13.7576 meteor: 0.75 -13.7576 60.6% 0.75 -13.7576 od 0.75 -13.7576 fair, 0.75 -13.7576 1263 0.75 -13.7576 36.88 0.75 -13.7576 96.2% 0.75 -13.7576 mix, 0.75 -13.7576 (bennett 0.666667 -13.7579 / 1.0 0.75 -13.7585 primary, 0.75 -13.7585 pov, 0.75 -13.7585 tt-svm 0.75 -13.7585 handicap 0.75 -13.7585 moulton 0.75 -13.7585 organized, 0.75 -13.7585 0.917 0.75 -13.7585 cassell 0.75 -13.7585 km, 0.75 -13.7585 marketing, 0.75 -13.7585 percussion 0.75 -13.7585 fla 0.75 -13.7585 zeng, 0.75 -13.7585 ripe 0.75 -13.7585 r3, 0.75 -13.7585 explosive 0.75 -13.7585 certainty, 0.75 -13.7585 ∩e+2 0.75 -13.7585 catastrophic 0.5 -13.7585 francisco, 0.571429 -13.7586 (noreen, 0.368421 -13.7588 figure 8 0.2 -13.7589 conveyed 0.5 -13.7591 (δ 0.571429 -13.7593 edmonton, 0.571429 -13.7593 galley, 0.571429 -13.7593 duc-06 0.75 -13.7595 spell-checking 0.75 -13.7595 pronoun: 0.75 -13.7595 *+ 0.75 -13.7595 overcame 0.75 -13.7595 67.56 0.75 -13.7595 64.18 0.75 -13.7595 cfgs, 0.75 -13.7595 reductions, 0.75 -13.7595 (resulting 0.75 -13.7595 86.16 0.75 -13.7595 repeat, 0.75 -13.7595 twin 0.75 -13.7595 since: 0.75 -13.7595 non-understandings 0.75 -13.7595 reimplement 0.75 -13.7595 xyz 0.333333 -13.7595 for obtaining 0.666667 -13.7596 interrelated 0.666667 -13.7596 gold-standard part-of-speech 0.571429 -13.76 where η 0.107296 -13.7603 respectively, 0.5 -13.7603 84.6 0.75 -13.7604 atsp 0.75 -13.7604 representatives, 0.75 -13.7604 (combinations 0.75 -13.7604 splitted 0.75 -13.7604 (ratkiewicz 0.75 -13.7604 plants, 0.75 -13.7604 grs 0.75 -13.7604 non-identical 0.75 -13.7604 prover 0.571429 -13.7607 monopats 0.428571 -13.7609 al., 2010, 0.666667 -13.7612 m-step, 0.666667 -13.7612 (do 0.75 -13.7613 fraction, 0.75 -13.7613 scope-disambiguated 0.75 -13.7613 91.95 0.75 -13.7613 snowball: 0.75 -13.7613 torgomorph 0.75 -13.7613 side: 0.75 -13.7613 965 0.75 -13.7613 500-best 0.75 -13.7613 8.9% 0.75 -13.7613 och’s 0.75 -13.7613 proverbial 0.75 -13.7613 0.988 0.75 -13.7613 chineseto-english 0.75 -13.7613 (ci) 0.75 -13.7613 preprocessors 0.75 -13.7613 +0.23 0.75 -13.7613 svetlana 0.75 -13.7613 naturallyoccurring 0.0391029 -13.7614 proceedings of 0.229167 -13.7617 rules were 0.571429 -13.7621 a; 0.75 -13.7622 1.50 0.75 -13.7622 intended, 0.75 -13.7622 discipline 0.75 -13.7622 train-test 0.75 -13.7622 (cherry 0.75 -13.7622 pir 0.75 -13.7622 bullishness 0.75 -13.7622 precision@5 0.75 -13.7622 mirenda 0.75 -13.7622 1960 0.75 -13.7622 multidisciplinary 0.75 -13.7622 (tillman, 0.75 -13.7622 aka 0.75 -13.7622 23.89 0.75 -13.7622 elkan, 0.75 -13.7622 many-tomany 0.75 -13.7622 (deb, 0.75 -13.7622 “word” 0.75 -13.7622 f10 0.75 -13.7622 event-by-document 0.4 -13.7624 understood by 0.290323 -13.7629 transactions 0.368421 -13.7629 null2, 0.454545 -13.7632 token’s 0.75 -13.7632 users: 0.75 -13.7632 (girju, 0.75 -13.7632 [5/ 0.75 -13.7632 basic italian 0.75 -13.7632 -6 0.75 -13.7632 “like” 0.75 -13.7632 (unweighted) 0.75 -13.7632 familiarize 0.75 -13.7632 presently, 0.75 -13.7632 credits 0.75 -13.7632 83.95 0.75 -13.7632 rcmod 0.75 -13.7632 bionlp: 0.75 -13.7632 indo-aryan 0.75 -13.7632 prompt, 0.75 -13.7632 selftrained 0.75 -13.7632 khudanpur 0.75 -13.7632 31.22 0.75 -13.7632 32.79 0.666667 -13.7637 ergative 0.454545 -13.7638 department university 0.75 -13.7641 each session, 0.75 -13.7641 (shah 0.75 -13.7641 holz 0.75 -13.7641 (quadrianto 0.75 -13.7641 questions), 0.75 -13.7641 asign 0.75 -13.7641 man, 0.75 -13.7641 unusable 0.75 -13.7641 (young 0.75 -13.7641 pedagogically 0.75 -13.7641 choses 0.75 -13.7641 (§6) 0.75 -13.7641 k=10 0.571429 -13.7642 0.363 0.571429 -13.7642 flowsthrough 0.454545 -13.7649 markus 0.120879 -13.7649 hope 0.75 -13.765 p(pi2) 0.75 -13.765 jmax 0.75 -13.765 gegessen 0.75 -13.765 다 0.75 -13.765 unaligned, 0.75 -13.765 re-scoring, 0.75 -13.765 throut 0.75 -13.765 wechsler, 0.75 -13.765 tier’s 0.75 -13.765 (mfs) 0.75 -13.765 22) 0.75 -13.765 attest 0.75 -13.765 2although 0.75 -13.765 (lakoff, 0.75 -13.765 pos(w 0.75 -13.765 hybridation 0.75 -13.765 philosophical 0.571429 -13.7657 0.014 0.571429 -13.7657 step) 0.75 -13.766 (samt) 0.75 -13.766 27.2% 0.75 -13.766 moseschart 0.75 -13.766 ushahidi 0.75 -13.766 conll-style 0.75 -13.766 pearson, 0.75 -13.766 together: 0.75 -13.766 high-accuracy 0.75 -13.766 topline 0.75 -13.766 8.1% 0.75 -13.766 viz., 0.75 -13.766 innovation, 0.75 -13.766 rater-author 0.75 -13.766 application-oriented 0.428571 -13.7662 rochester, 0.75 -13.7669 39.07 0.75 -13.7669 metaphor, 0.75 -13.7669 575 0.75 -13.7669 argm-adv 0.75 -13.7669 (adjp 0.75 -13.7669 backbone, 0.75 -13.7669 10.00 0.75 -13.7669 confine 0.75 -13.7669 avalanche 0.75 -13.7669 length restriction 0.75 -13.7669 4.43 0.75 -13.7669 post-filtering 0.75 -13.7669 (bhatia 0.75 -13.7669 simplify notation, 0.75 -13.7669 (mitchell, 0.666667 -13.767 stroke 0.666667 -13.767 recall rate 0.571429 -13.7671 (≤40) 0.0814249 -13.7676 data from 0.141732 -13.7676 by using a 0.75 -13.7678 (good, 0.75 -13.7678 meta-analysis 0.75 -13.7678 0.261 0.75 -13.7678 listnet 0.75 -13.7678 containing: 0.75 -13.7678 kullbackleibler 0.75 -13.7678 n-foil, 0.75 -13.7678 0.976 0.75 -13.7678 (tatu 0.75 -13.7678 ferguson 0.75 -13.7678 similar; 0.75 -13.7678 “no”, 0.75 -13.7678 (plutchik, 0.5 -13.7678 h1,1 0.571429 -13.7685 room, 0.75 -13.7688 salton, 0.75 -13.7688 parte 0.75 -13.7688 (sentence) 0.75 -13.7688 symmetrizing 0.75 -13.7688 ∈{jj} 0.75 -13.7688 taipei, 0.75 -13.7688 r3) 0.75 -13.7688 touch, 0.75 -13.7688 1450 0.75 -13.7688 she’s 0.75 -13.7688 (ananiadou 0.75 -13.7688 ,… 0.75 -13.7688 qc, 0.75 -13.7688 delicate 0.75 -13.7688 morphologically, 0.75 -13.7688 positive; 0.75 -13.7688 noh, 0.75 -13.7688 590 0.75 -13.7688 +5 0.75 -13.7688 effects; 0.229167 -13.769 a naive 0.571429 -13.7692 hoa 0.181818 -13.7693 a rule-based 0.75 -13.7697 unemployment_rate 0.75 -13.7697 ’) 0.75 -13.7697 “compliment” 0.75 -13.7697 27.6 0.75 -13.7697 utt 0.75 -13.7697 error-specific 0.75 -13.7697 fmeasures 0.75 -13.7697 (bow 0.75 -13.7697 replayed 0.75 -13.7697 inch 0.75 -13.7697 batch, 0.75 -13.7697 entity-attribute 0.75 -13.7697 reprints 0.75 -13.7697 2003b) 0.75 -13.7697 hsueh 0.75 -13.7697 cues) 0.571429 -13.7699 srilm’s 0.571429 -13.7699 jordi 0.571429 -13.7699 dekang 0.75 -13.7706 cond, 0.75 -13.7706 d* 0.75 -13.7706 66.70 0.75 -13.7706 74.40 0.75 -13.7706 un, 0.75 -13.7706 relation), 0.75 -13.7706 (95% 0.75 -13.7706 63.6% 0.75 -13.7706 0.702 0.75 -13.7706 magnified 0.75 -13.7706 pbl 0.75 -13.7706 intracellular 0.75 -13.7706 sub-system, 0.75 -13.7706 rapidly, 0.147541 -13.7706 be made 0.666667 -13.7712 gezien 0.28125 -13.7712 formula: 0.75 -13.7716 dealers 0.75 -13.7716 48.81 0.75 -13.7716 interval) 0.75 -13.7716 president] 0.75 -13.7716 kuhn-munkres 0.75 -13.7716 utterances), 0.75 -13.7716 4-grams, 0.75 -13.7716 ‘hidden’ 0.75 -13.7716 shift-reduce-based 0.75 -13.7716 nullmod 0.75 -13.7716 33.2% 0.75 -13.7716 itg, 0.75 -13.7716 2005] 0.454545 -13.7717 moses: 0.0637636 -13.7725 account 0.75 -13.7725 sousa 0.75 -13.7725 classed 0.75 -13.7725 weasels 0.75 -13.7725 敯 0.75 -13.7725 3.4% 0.75 -13.7725 hurst, 0.75 -13.7725 retrievable 0.75 -13.7725 city), 0.75 -13.7725 59.33 0.75 -13.7725 1962) 0.75 -13.7725 tomorrow, 0.75 -13.7725 usernames, 0.75 -13.7725 49.38 0.666667 -13.7728 gn(eprime) 0.666667 -13.7728 commenting 0.666667 -13.7728 896 0.666667 -13.7728 arrowdblbothv 0.666667 -13.7728 poly 0.666667 -13.7728 4gram 0.666667 -13.7728 md. 0.666667 -13.7728 verb‟s 0.666667 -13.7728 (casacuberta 0.666667 -13.7728 taylor, 0.666667 -13.7728 lengthpenalty 0.666667 -13.7728 collobert 0.666667 -13.7728 mead, 0.666667 -13.7728 motion. 0.666667 -13.7728 utterances). 0.666667 -13.7728 0.25. 0.666667 -13.7728 volition 0.666667 -13.7728 source-parsed 0.666667 -13.7728 poverty 0.666667 -13.7728 structured, 0.666667 -13.7728 kipper, 0.666667 -13.7728 catch-all 0.666667 -13.7728 matched. 0.666667 -13.7728 translation). 0.666667 -13.7728 knn, 0.666667 -13.7728 authors) 0.666667 -13.7728 unlikely. 0.666667 -13.7728 1.00. 0.666667 -13.7728 @e(be 0.666667 -13.7728 re-scoring. 0.666667 -13.7728 onehop 0.666667 -13.7728 342 0.666667 -13.7728 pane 0.666667 -13.7728 explorations 0.666667 -13.7728 surprisal. 0.666667 -13.7728 -null-aalign 0.666667 -13.7728 sumaries. 0.666667 -13.7728 co-clustering 0.666667 -13.7728 52.83 0.666667 -13.7728 (svo) 0.666667 -13.7728 tsuji. 0.666667 -13.7728 pittsburgh. 0.571429 -13.7728 ‘vaalaa’ 0.166667 -13.773 them into 0.428571 -13.7732 indicators’ 0.75 -13.7734 40.28 0.75 -13.7734 threading 0.75 -13.7734 0.133 0.75 -13.7734 sample: 0.75 -13.7734 rlscriptk 0.75 -13.7734 0.235 0.75 -13.7734 compatible, 0.75 -13.7734 32.34 0.75 -13.7734 semi 0.75 -13.7734 amber(1,4,6) 0.75 -13.7734 39.90 0.75 -13.7734 8.12 0.75 -13.7734 zeitung 0.75 -13.7734 purposes: 0.75 -13.7734 ter points. 0.0666667 -13.7737 is, 0.571429 -13.7742 rel, 0.75 -13.7744 1.2, 0.75 -13.7744 (joy, 0.75 -13.7744 observes, 0.75 -13.7744 wikipedia's 0.75 -13.7744 40.31 0.75 -13.7744 2.79 0.75 -13.7744 persuasion, 0.75 -13.7744 subproblems, 0.75 -13.7744 561 0.75 -13.7744 probabilities; 0.75 -13.7744 packing, 0.75 -13.7744 hasse 0.75 -13.7744 good: 0.75 -13.7744 ofice 0.75 -13.7744 operationalized 0.75 -13.7744 factors) 0.75 -13.7753 sent back 0.75 -13.7753 (ibm 0.75 -13.7753 1.87 0.75 -13.7753 constituent-based 0.75 -13.7753 argument: 0.75 -13.7753 commands, 0.75 -13.7753 85.90 0.75 -13.7753 “semantically 0.75 -13.7753 spence 0.75 -13.7753 1.15 0.75 -13.7753 hoffmann 0.75 -13.7753 11.76 0.75 -13.7753 %), 0.75 -13.7753 tell, 0.75 -13.7753 86.9% 0.4 -13.7753 promoted 0.16 -13.7756 the middle 0.571429 -13.7756 14.1 0.75 -13.7763 hourly 0.75 -13.7763 48.44 0.75 -13.7763 disjuncture 0.75 -13.7763 +fb 0.75 -13.7763 homographic 0.75 -13.7763 106, 0.75 -13.7763 (bayesian 0.75 -13.7763 marie-francine 0.75 -13.7763 delhi 0.75 -13.7763 o2, 0.75 -13.7763 86.34 0.75 -13.7763 barrier, 0.75 -13.7763 “play”, 0.571429 -13.7764 stimulate 0.12987 -13.7765 on. 0.125 -13.7767 summationtext 0.454545 -13.7768 learnthigh 0.571429 -13.7771 exposures 0.75 -13.7772 (50% 0.75 -13.7772 conspicuous 0.75 -13.7772 whole-word 0.75 -13.7772 seq 0.75 -13.7772 coreference; 0.75 -13.7772 surveys, 0.75 -13.7772 mereology 0.75 -13.7772 p&g 0.75 -13.7772 “null” 0.75 -13.7772 (gildea, 0.333333 -13.7778 characteristics. 0.75 -13.7781 (adjectives), 0.75 -13.7781 (schwenk 0.75 -13.7781 (here) 0.75 -13.7781 additively 0.75 -13.7781 favours 0.75 -13.7781 59.7% 0.75 -13.7781 0.221 0.75 -13.7781 802 0.75 -13.7781 that-cc 0.75 -13.7781 (riaz 0.75 -13.7781 ambiguities: 0.75 -13.7781 s56s50 0.75 -13.7781 purse 0.75 -13.7781 666 0.75 -13.7781 segmenter’s 0.75 -13.7781 sliced 0.75 -13.7781 intern 0.75 -13.7781 ˆy1 0.75 -13.7781 0.226 0.454545 -13.7785 modeller 0.571429 -13.7785 craig 0.571429 -13.7785 (enright 0.75 -13.7791 51.57 0.75 -13.7791 territories 0.75 -13.7791 leipzig 0.75 -13.7791 (17), 0.75 -13.7791 pr(t,s) 0.75 -13.7791 syntacticsemantic 0.75 -13.7791 pots 0.75 -13.7791 burstiness 0.75 -13.7791 (dinu 0.75 -13.7791 it), 0.75 -13.7791 )} 0.75 -13.7791 r2aft 0.75 -13.7791 bi-texts 0.75 -13.7791 ixy 0.75 -13.7791 mnb/priors 0.571429 -13.7792 kanayama 0.571429 -13.7792 92.13 0.32 -13.7792 foundation. 0.296296 -13.7799 lu et 0.296296 -13.7799 details are 0.75 -13.78 fall, 0.75 -13.78 18.42 0.75 -13.78 herring, 0.75 -13.78 typologies, 0.75 -13.78 lrscore, 0.75 -13.78 growth rate 0.75 -13.78 ny 10598 0.75 -13.78 non-significant 0.75 -13.78 qiu, 0.75 -13.78 k1) 0.75 -13.78 phrase-pairs, 0.75 -13.78 vt 0.571429 -13.78 medstract 0.571429 -13.78 2011] 0.75 -13.781 prendinger 0.75 -13.781 32.33 0.75 -13.781 larger; 0.75 -13.781 (melamed, 0.75 -13.781 predictionscore 0.75 -13.781 nirenburg 0.75 -13.781 prashanth 0.75 -13.781 39.82 0.75 -13.781 (roles) 0.75 -13.781 “something 0.75 -13.781 15.85 0.75 -13.781 coarse-graining 0.75 -13.781 collin’s 0.75 -13.7819 diagnose 0.75 -13.7819 advisory 0.75 -13.7819 alshawi, 0.75 -13.7819 42.30 0.75 -13.7819 400, 0.75 -13.7819 29.54 0.75 -13.7819 contention, 0.75 -13.7819 (dotted 0.75 -13.7819 yamcha 0.75 -13.7819 firmly 0.75 -13.7819 16.57 0.75 -13.7819 7.00 0.75 -13.7819 olga 0.75 -13.7819 hillary 0.75 -13.7819 dps 0.75 -13.7819 8.8% 0.75 -13.7819 reduction-based 0.75 -13.7819 monolithic 0.75 -13.7819 surfing 0.75 -13.7819 bias: 0.75 -13.7819 0.474 0.75 -13.7819 high-ranking 0.28125 -13.7821 ip@null 0.5 -13.7822 singletons. 0.454545 -13.7825 54.9 0.234043 -13.7826 presented by 0.0637636 -13.7826 conference 0.571429 -13.7828 “excellent” 0.35 -13.7828 removed all 0.75 -13.7829 combinators, 0.75 -13.7829 rejoin 0.75 -13.7829 ‡department 0.75 -13.7829 two-character 0.75 -13.7829 surf 0.75 -13.7829 1353 0.75 -13.7829 0.517 0.75 -13.7829 1127 0.75 -13.7829 0.382 0.75 -13.7829 0.799 0.75 -13.7829 (bergmann 0.75 -13.7829 accomodate 0.75 -13.7829 60,000 0.75 -13.7829 theo 0.75 -13.7829 51.49 0.75 -13.7829 (probability 0.75 -13.7829 rhs, 0.454545 -13.7837 0.538 0.75 -13.7838 unanimously 0.75 -13.7838 [he] 0.75 -13.7838 modify, 0.75 -13.7838 numeral 0.75 -13.7838 burst 0.75 -13.7838 hs\it 0.75 -13.7838 lamar 0.75 -13.7838 lightly 0.75 -13.7838 “can” 0.75 -13.7838 956 0.75 -13.7838 bip-monods 0.75 -13.7838 bartlet, 0.75 -13.7838 predicate-arguments 0.75 -13.7838 simulations, 0.75 -13.7838 compact(s 0.75 -13.7838 buitelaar 0.75 -13.7838 pre-test, 0.75 -13.7838 np0:1 0.75 -13.7838 millions) 0.454545 -13.7843 and philip 0.75 -13.7848 detector-ii 0.75 -13.7848 english→japanese 0.75 -13.7848 upv 0.75 -13.7848 (sang 0.75 -13.7848 5-way 0.75 -13.7848 scoring: 0.75 -13.7848 pyp-1hmm 0.75 -13.7848 adjacent, 0.75 -13.7848 duction 0.75 -13.7848 687 0.75 -13.7848 gorn 0.75 -13.7848 (msa), 0.75 -13.7848 nested, 0.571429 -13.785 16.8 0.571429 -13.785 pizza 0.75 -13.7857 ˆs1 0.75 -13.7857 sentence-pairs 0.75 -13.7857 allen’s 0.75 -13.7857 52.57 0.75 -13.7857 haghighi, 0.75 -13.7857 stereotypes, 0.75 -13.7857 tweet; 0.75 -13.7857 32.81 0.75 -13.7857 “optimal” 0.75 -13.7857 offthe-shelf 0.75 -13.7857 memorybased 0.75 -13.7857 jm 0.75 -13.7857 aho 0.75 -13.7857 feels 0.75 -13.7857 sensitivity, 0.75 -13.7857 357 0.75 -13.7857 hypernymy, 0.75 -13.7857 0.607 0.571429 -13.7857 sec) 0.571429 -13.7864 systems r-1 0.571429 -13.7864 one-way 0.333333 -13.7866 all remaining 0.5 -13.7866 doc-avg 0.75 -13.7867 1968) 0.75 -13.7867 n\n 0.75 -13.7867 bremen 0.75 -13.7867 (significantly 0.75 -13.7867 b/c 0.75 -13.7867 ht-en 0.75 -13.7867 5if 0.75 -13.7867 manufactured 0.75 -13.7867 3.64 0.75 -13.7867 eanlu 0.75 -13.7867 (iarpa), 0.75 -13.7867 0.029 0.75 -13.7867 m&w 0.75 -13.7867 anew 0.75 -13.7867 12.92 0.0715706 -13.7869 problem of 0.454545 -13.7871 undefined 0.5 -13.7872 eθ 0.5 -13.7872 maxlength 0.75 -13.7876 contiguity 0.75 -13.7876 recreate 0.75 -13.7876 (how 0.75 -13.7876 newssyscombtest2011 0.75 -13.7876 (visual 0.75 -13.7876 aba 0.75 -13.7876 ngramm 0.75 -13.7876 sumbasic, 0.75 -13.7876 βs 0.75 -13.7876 misconceptions 0.75 -13.7876 2); 0.210526 -13.7879 bilingual language 0.75 -13.7886 56.58 0.75 -13.7886 46.88 0.75 -13.7886 state-splitting 0.75 -13.7886 mth 0.75 -13.7886 samples: 0.75 -13.7886 1spd 0.75 -13.7886 biasml 0.75 -13.7886 greedy local 0.75 -13.7886 1.08 0.571429 -13.7886 pollard 0.307692 -13.7887 event annotation 0.368421 -13.7888 def⇐⇒ 0.5 -13.7891 rank≥ 0.571429 -13.7893 0.376 0.75 -13.7895 (pustejovsky, 0.75 -13.7895 kreyol 0.75 -13.7895 country) 0.75 -13.7895 regulates 0.75 -13.7895 icwr: 0.75 -13.7895 bruce, 0.75 -13.7895 17.03 0.75 -13.7895 n3 0.75 -13.7895 26.15 0.75 -13.7895 enough” 0.75 -13.7895 sbar(q) 0.75 -13.7895 ye, 0.75 -13.7895 punctuation: 0.75 -13.7895 f2,f3,f4,f9 0.75 -13.7895 recursive, 0.75 -13.7895 (winston 0.75 -13.7895 test09 0.75 -13.7895 french– 0.75 -13.7895 nonsensical 0.75 -13.7895 56.87 0.75 -13.7895 97.5 0.128205 -13.7897 to give 0.19403 -13.7899 algorithms are 0.571429 -13.7901 naacl, 0.0957447 -13.7903 it would 0.5 -13.7904 0.8. 0.75 -13.7905 equal: 0.75 -13.7905 87–95, 0.75 -13.7905 (bilingual 0.75 -13.7905 4,567 0.75 -13.7905 42nd 0.75 -13.7905 endwhile 0.75 -13.7905 s�p16,p17 0.75 -13.7905 trouw 0.75 -13.7905 pbtas 0.75 -13.7905 lone 0.75 -13.7905 649 0.75 -13.7905 latexpbleu 0.75 -13.7905 uninformed 0.75 -13.7905 horizontally 0.75 -13.7905 1332 0.307692 -13.7907 offering 0.571429 -13.7908 undiacritized 0.368421 -13.7911 funding from 0.35 -13.7912 presented at 0.75 -13.7914 (3.4) 0.75 -13.7914 kx 0.75 -13.7914 patternbased 0.75 -13.7914 sov, 0.75 -13.7914 alleviating 0.75 -13.7914 (response 0.75 -13.7914 relatedness: 0.75 -13.7914 (protein 0.75 -13.7914 boolean-valued 0.75 -13.7914 prior) 0.75 -13.7914 once) 0.75 -13.7914 29.21 0.75 -13.7914 log-likelihod 0.75 -13.7914 weblm 0.75 -13.7914 qs3 0.75 -13.7914 modulo 0.75 -13.7914 mother’s 0.428571 -13.7915 plastic 0.454545 -13.7918 [e1 0.75 -13.7924 1975 0.75 -13.7924 0.486 0.75 -13.7924 76.1% 0.75 -13.7924 concreteness 0.75 -13.7924 annotating, 0.75 -13.7924 struggle 0.75 -13.7924 geolocate 0.75 -13.7924 feedbacktype 0.75 -13.7924 tβt 0.75 -13.7924 3-10 0.75 -13.7924 non-ace 0.75 -13.7924 63.11% 0.75 -13.7924 penalising 0.75 -13.7924 kato 0.75 -13.7924 waveforms 0.110577 -13.7924 here we 0.243902 -13.7928 positions. 0.571429 -13.793 chicken 0.571429 -13.793 a∗ 0.571429 -13.793 99.5 0.428571 -13.7932 spending 0.428571 -13.7932 extract events 0.75 -13.7933 1387 0.75 -13.7933 large: 0.75 -13.7933 p(pi1) 0.75 -13.7933 (lemon 0.454545 -13.7935 sms, 0.571429 -13.7937 almor 0.184211 -13.794 one way 0.5 -13.7942 valley 0.75 -13.7943 pronouns: 0.75 -13.7943 95.50 0.75 -13.7943 three: 0.75 -13.7943 trang 0.75 -13.7943 dialogue) 0.75 -13.7943 shortcomings, 0.75 -13.7943 research) 0.75 -13.7943 (skantze 0.75 -13.7943 cmnt2 0.35 -13.7943  0.571429 -13.7944 others), 0.4 -13.7945 (bagga 0.0330508 -13.7948 in our 0.75 -13.7953 1.94 0.75 -13.7953 i-1 0.75 -13.7953 topics” 0.75 -13.7953 “glue” 0.75 -13.7953 least-squares 0.75 -13.7953 neurodevelopmental 0.75 -13.7953 ‘correct’ 0.75 -13.7953 hindi: 0.75 -13.7953 sein 0.75 -13.7953 37.96 0.75 -13.7953 types” 0.75 -13.7953 ,f 0.75 -13.7953 staab, 0.75 -13.7953 sjl 0.75 -13.7953 interpolation, 0.5 -13.7955 62.6 0.264706 -13.7956 summed 0.28125 -13.7957 satellite 0.454545 -13.7958 ccg grammar 0.75 -13.7962 (unigram 0.75 -13.7962 issue: 0.75 -13.7962 (appendix 0.75 -13.7962 artery 0.75 -13.7962 54.05 0.75 -13.7962 +0.32 0.75 -13.7962 e-j 0.75 -13.7962 0.637 0.75 -13.7962 tional 0.75 -13.7962 0.251 0.75 -13.7962 {v 0.75 -13.7962 curcuma 0.75 -13.7962 “goodness” 0.75 -13.7962 english-to-japanese 0.75 -13.7962 logicalforms 0.75 -13.7962 loss-augmented 0.75 -13.7962 (similarly 0.163043 -13.7962 our analysis 0.454545 -13.7964 -10 0.454545 -13.7964 “cananea” 0.571429 -13.7966 across four 0.571429 -13.7966 (relatively 0.571429 -13.7966 16.79 0.454545 -13.797 xx 0.75 -13.7972 sisters 0.75 -13.7972 (murphy, 0.75 -13.7972 puzzling 0.75 -13.7972 1571 0.75 -13.7972 321 0.75 -13.7972 available), 0.75 -13.7972 gg 0.75 -13.7972 (frederking 0.75 -13.7972 0.402 0.75 -13.7972 (numbers 0.75 -13.7972 suffixed 0.75 -13.7972 0.455 0.75 -13.7972 test; 0.75 -13.7972 achieve, 0.75 -13.7972 unithood-based 0.75 -13.7972 simplifications, 0.75 -13.7972 preannotation 0.571429 -13.7974 tags; 0.12069 -13.7974 observed that 0.5 -13.7974 chin-yew 0.4 -13.7976 dt jj 0.307692 -13.7978 negative regulation 0.75 -13.7981 disappointing, 0.75 -13.7981 28.24 0.75 -13.7981 (full) 0.75 -13.7981 (fellowship 0.75 -13.7981 41.51 0.75 -13.7981 haitisurf 0.75 -13.7981 carletta, 0.75 -13.7981 (soft 0.75 -13.7981 2.4k 0.75 -13.7981 59.82 0.75 -13.7981 moore’s 0.75 -13.7981 93.51 0.75 -13.7981 occ(t2) 0.75 -13.7981 83k 0.75 -13.7981 any topic. 0.571429 -13.7981 nurses 0.571429 -13.7981 indefinite, 0.375 -13.7981 realizer 0.177215 -13.7982 approach has 0.428571 -13.7987 = 1...n 0.571429 -13.7988 np-obj 0.75 -13.7991 39.31 0.75 -13.7991 (b): 0.75 -13.7991 oliva, 0.75 -13.7991 28.1% 0.75 -13.7991 behaving 0.75 -13.7991 (kurohashi 0.75 -13.7991 universal, 0.75 -13.7991 c-per-i 0.75 -13.7991 acid, 0.75 -13.7991 individualistic 0.75 -13.7991 det, 0.75 -13.7991 a), 0.75 -13.7991 xue, 0.75 -13.7991 38.19 0.75 -13.7991 81.48 0.75 -13.7991 54.50 0.75 -13.7991 (k-κ 0.75 -13.7991 69.8% 0.140625 -13.7992 corpora are 0.571429 -13.7996 10-20 0.571429 -13.7996 505 0.4 -13.7997 arc. 0.5 -13.8 's 0.75 -13.8001 during, 0.75 -13.8001 wikulu, 0.75 -13.8001 reptile 0.75 -13.8001 soil, 0.75 -13.8001 80, 0.75 -13.8001 relations), 0.75 -13.8001 18.5% 0.75 -13.8001 finetocoarse 0.75 -13.8001 2computer 0.75 -13.8001 width, 0.75 -13.8001 department pomona 0.75 -13.8001 expression” 0.75 -13.8001 0.870 0.75 -13.8001 loqui 0.28125 -13.8004 translation models, 0.75 -13.801 depuis 0.75 -13.801 752 0.75 -13.801 objects’ 0.75 -13.801  meat 0.75 -13.801 wrede 0.75 -13.801 deacetylation 0.75 -13.801 1978; 0.75 -13.801 waits 0.75 -13.801 764 0.75 -13.801 joke 0.75 -13.801 b3) 0.75 -13.801 meal 0.75 -13.801 i-5 0.75 -13.801 bit-string 0.75 -13.801 inches 0.571429 -13.801 0s 0.428571 -13.8014 additional constraints 0.28125 -13.8015 7th 0.75 -13.802 outliers, 0.75 -13.802 2.5, 0.75 -13.802 tradition, 0.75 -13.802 -0.11 0.75 -13.802 fr→en 0.75 -13.802 lally, 0.75 -13.802 (referring 0.75 -13.802 perspective: 0.75 -13.802 :d)→x 0.75 -13.802 incremental, 0.75 -13.802 inv 0.75 -13.802 vw, 0.75 -13.803 investigated: 0.75 -13.803 censoring 0.75 -13.803 brilliant 0.75 -13.803 condition; 0.75 -13.803 40.68 0.75 -13.803 distinguished, 0.75 -13.803 boundaries), 0.75 -13.803 eventualities 0.75 -13.803 larval 0.75 -13.803 0.867 0.75 -13.803 s′ 0.75 -13.803 tag-set 0.75 -13.803 sociocultural 0.75 -13.803 version) 0.75 -13.803 drosophila 0.75 -13.803 properties) 0.75 -13.803 0.220 0.571429 -13.8032 1971; 0.32 -13.8032 provide some 0.75 -13.8039 dense, 0.75 -13.8039 0.005, 0.75 -13.8039 multi-pass 0.75 -13.8039 (navigli 0.75 -13.8039 mapped, 0.75 -13.8039 aspect: 0.75 -13.8039 definitely, 0.75 -13.8039 q+l 0.75 -13.8039 mohammad, 0.75 -13.8039 driver, 0.75 -13.8039 diaries 0.75 -13.8039 kpbl2 0.75 -13.8039 utterances: 0.75 -13.8039 “talk” 0.454545 -13.804 n nsummationdisplay 0.571429 -13.804 1,500 0.166667 -13.8046 behavior of 0.28125 -13.8047 bing 0.75 -13.8049 thenumberof 0.75 -13.8049 (met) 0.75 -13.8049 low-precision 0.75 -13.8049 noun/verb 0.75 -13.8049 cartoni 0.75 -13.8049 resembling 0.75 -13.8049 “children 0.75 -13.8049 decisionrelated 0.75 -13.8049 est(˚ι) 0.75 -13.8049 43.23 0.75 -13.8049 nullc 0.75 -13.8049 williams, 0.75 -13.8049 -7 0.75 -13.8049 (20), 0.75 -13.8049 finding, 0.454545 -13.8052 nivre’s 0.333333 -13.8054 stripping 0.163043 -13.8054 overview of the 0.571429 -13.8055 a-priori 0.375 -13.8058 any possible 0.375 -13.8058 clm 0.5 -13.8058 they achieved 0.75 -13.8059 rerankers 0.75 -13.8059 codl 0.75 -13.8059 mrrp 0.75 -13.8059 (x,r,y) 0.75 -13.8059 double-annotated 0.75 -13.8059 box) 0.75 -13.8059 transitively 0.75 -13.8059 endowed 0.75 -13.8059 (α) 0.75 -13.8059 centroids, 0.75 -13.8059 narrowed 0.75 -13.8059 paragraph) 0.75 -13.8059 overlapping, 0.75 -13.8059 88.37 0.75 -13.8059 spanmaps 0.75 -13.8059 ips, 0.571429 -13.8062 (mohammad et 0.333333 -13.8063 epigenetics 0.375 -13.8063 time). 0.428571 -13.8064 intervals, 0.75 -13.8068 96.2 0.75 -13.8068 operationally, 0.75 -13.8068 pra, 0.75 -13.8068 1,k] 0.75 -13.8068 boil 0.75 -13.8068 982 0.75 -13.8068 zong 0.75 -13.8068 syntbipmatched 0.75 -13.8068 (hwang 0.75 -13.8068 single-character 0.75 -13.8068 springer, 0.75 -13.8068 (every 0.75 -13.8068 3.5m 0.75 -13.8068 caroline 0.0432869 -13.8074 evaluate 0.75 -13.8078 child; 0.75 -13.8078 bonneau-maynard 0.75 -13.8078 !, 0.75 -13.8078 49.4% 0.75 -13.8078 companies) 0.75 -13.8078 “sentence” 0.75 -13.8078 striving 0.75 -13.8078 silence, 0.75 -13.8078 ods 0.75 -13.8078 look”, 0.75 -13.8078 t’th 0.75 -13.8078 sandrine 0.75 -13.8078 animated 0.28125 -13.808 paradigm. 0.139535 -13.808 gives a 0.454545 -13.8087 dass 0.75 -13.8088 strangers 0.75 -13.8088 an/vn 0.75 -13.8088 plagued 0.75 -13.8088 6.0% 0.75 -13.8088 up”, 0.75 -13.8088 workaround 0.75 -13.8088 announcement 0.75 -13.8088 48.97 0.75 -13.8088 shakespeare’s 0.75 -13.8088 synset–article 0.75 -13.8088 ‘new’ 0.75 -13.8088 (tur 0.75 -13.8088 re-alignment 0.75 -13.8088 exposition, 0.75 -13.8088 (3.7) 0.75 -13.8088 7.17 0.75 -13.8088 1048 0.22449 -13.809 (4) where 0.571429 -13.8091 ‘eric 0.571429 -13.8091 similarity scores, 0.428571 -13.8092 splitting. 0.75 -13.8097 counts: 0.75 -13.8097 entailment” 0.75 -13.8097 stimulus, 0.75 -13.8097 astray, 0.75 -13.8097 same; 0.75 -13.8097 csk 0.75 -13.8097 clc-base 0.75 -13.8097 watchstanders 0.75 -13.8097 identified) 0.75 -13.8097 89.9% 0.75 -13.8097 tac2010 0.75 -13.8097 re-estimating 0.75 -13.8097 morphism 0.75 -13.8097 optima, 0.75 -13.8097 gulf, 0.75 -13.8097 >.05 0.75 -13.8097 language(s) 0.571429 -13.8099 we’d 0.75 -13.8107 (dag) 0.75 -13.8107 null 0.75 -13.8107 agrovoc 0.75 -13.8107 extrapolating 0.75 -13.8107 (pt 0.75 -13.8107 retriever 0.75 -13.8107 striker 0.75 -13.8107 644 0.571429 -13.8114 airliner 0.5 -13.8116 p(x 0.75 -13.8117 thread-level 0.75 -13.8117 5.6% 0.75 -13.8117 flexible 0.75 -13.8117 946 0.75 -13.8117 bagga 0.75 -13.8117 orientated 0.75 -13.8117 grammars; 0.75 -13.8117 r-coh 0.75 -13.8117 framenet’s 0.75 -13.8117 non-technical 0.75 -13.8117 99.4% 0.75 -13.8117 underspecification, 0.75 -13.8117 passoneau’s 0.75 -13.8117 615 0.75 -13.8117 :a 0.75 -13.8117 x% 0.333333 -13.8119 usc 0.571429 -13.8121 sl, 0.571429 -13.8121 deviant’s 0.229167 -13.8122 set size 0.75 -13.8127 matchers 0.75 -13.8127 preterminal, 0.75 -13.8127 testers 0.75 -13.8127 interviewed 0.75 -13.8127 speeds, 0.75 -13.8127 aj, 0.75 -13.8127 “collapsed” 0.75 -13.8127 (duchi 0.75 -13.8127 12m 0.75 -13.8127 57.32 0.75 -13.8127 bcd 0.571429 -13.8129 wdt 0.571429 -13.8129 demonstrative, 0.75 -13.8136 passives, 0.75 -13.8136 non-viterbi 0.75 -13.8136 studio 0.75 -13.8136 transformedinto 0.75 -13.8136 mwv 0.75 -13.8136 mounting 0.75 -13.8136 h1n1 0.75 -13.8136 sentiwordnet, 0.75 -13.8136 subjectively 0.75 -13.8136 gonzalez 0.75 -13.8136 heart, 0.75 -13.8136 p.m. 0.75 -13.8136 lerman 0.75 -13.8136 harmonization 0.75 -13.8136 [i] 0.75 -13.8136 (tree 0.75 -13.8136 611 0.75 -13.8136 34.8 0.75 -13.8136 94.39 0.75 -13.8136 genomic 0.571429 -13.8136 85.92 0.0852273 -13.8139 portland, oregon, june 0.0182774 -13.8144 between 0.206897 -13.8144 12th 0.75 -13.8146 increase) 0.75 -13.8146 ann-gram 0.75 -13.8146 44.02 0.75 -13.8146 exchanged 0.75 -13.8146 kopp 0.75 -13.8146 contextualize 0.75 -13.8146 non-adapted 0.75 -13.8146 imposed, 0.75 -13.8146 (chapelle 0.75 -13.8146 note: 0.75 -13.8146 knowledge-rich 0.75 -13.8146 karamanis 0.4 -13.8147 crucial role 0.238095 -13.8151 results showed 0.571429 -13.8151 c8 0.571429 -13.8151 eisner, 2005; 0.75 -13.8156 entitybased 0.75 -13.8156 (bell 0.75 -13.8156 literally, 0.75 -13.8156 of’ 0.75 -13.8156 just, 0.75 -13.8156 headers, 0.75 -13.8156 kdimensional 0.75 -13.8156 adjuncts, 0.75 -13.8156 28.04 0.75 -13.8156 “using 0.75 -13.8156 above-described 0.75 -13.8156 first) 0.75 -13.8156 1.24 0.75 -13.8156 observed: 0.4 -13.8162 hirschberg 0.5 -13.8162 pc4 0.75 -13.8166 warn 0.75 -13.8166 crafting 0.75 -13.8166 4.00 0.75 -13.8166 learning) 0.75 -13.8166 vantage 0.75 -13.8166 centrally 0.75 -13.8166 (head 0.75 -13.8166 nature: 0.75 -13.8166 lester. 0.75 -13.8166 “don’t 0.75 -13.8166 (x-axis) 0.75 -13.8166 508 0.571429 -13.8166 figure2: 0.140625 -13.8167 lapata, 0.375 -13.8175 profit 0.75 -13.8176 damaging 0.75 -13.8176 top-20 0.75 -13.8176 icml 0.75 -13.8176 penalization 0.75 -13.8176 1–10, 0.75 -13.8176 headline, 0.75 -13.8176 corpus-based, 0.75 -13.8176 psycholinguistics, 0.75 -13.8176 car, 0.75 -13.8176 intsim(a→b) 0.75 -13.8176 thoughts 0.0735931 -13.8176 appears 0.28125 -13.8179 a feature vector 0.571429 -13.8181 forbes and 0.75 -13.8185 87.31 0.75 -13.8185 dualist, 0.75 -13.8185 legs 0.75 -13.8185 empty-element 0.75 -13.8185 college-level 0.75 -13.8185 67.97 0.75 -13.8185 single-language 0.75 -13.8185 (hand-crafted) 0.75 -13.8185 quantum 0.75 -13.8185 41.68 0.75 -13.8185 0.030 0.75 -13.8185 suppress 0.375 -13.8185 (pp 0.571429 -13.8188 umr 0.454545 -13.8194 regulon 0.75 -13.8195 909 0.75 -13.8195 89.7% 0.75 -13.8195 coarseness 0.75 -13.8195 learners) 0.75 -13.8195 regmt, 0.75 -13.8195 chatroom 0.75 -13.8195 parsimonious 0.75 -13.8195 item-type, 0.75 -13.8195 0.719 0.75 -13.8195 0.047 0.75 -13.8195 i∈p 0.75 -13.8195 comparatively, 0.75 -13.8195 post-tutoring 0.75 -13.8195 ss, 0.75 -13.8195 null’s 0.75 -13.8195 57.46 0.75 -13.8195 parent’s 0.571429 -13.8196 information technology, 0.571429 -13.8196 estar 0.257143 -13.8199 same, 0.4 -13.8204 it generates 0.75 -13.8205 ωgld 0.75 -13.8205 wmt’10 0.75 -13.8205 demos 0.75 -13.8205 33k 0.75 -13.8205 0.490 0.75 -13.8205 positively associated 0.75 -13.8205 causation, 0.75 -13.8205 m&e 0.75 -13.8205 906 0.75 -13.8205 subtypes: 0.75 -13.8205 5.9% 0.75 -13.8205 epstein 0.75 -13.8205 tractable inference. 0.75 -13.8205 grl 0.75 -13.8205 ittycheriah 0.75 -13.8205 rectly 0.75 -13.8205 sinks 0.454545 -13.8206 an sds 0.454545 -13.8211 case, where 0.571429 -13.8211 remarked 0.5 -13.8214 fields: 0.75 -13.8215 s�p16 0.75 -13.8215 twitter-lda 0.75 -13.8215 ckk, 0.75 -13.8215 asr-08-test 0.75 -13.8215 readme 0.75 -13.8215 oblique 0.75 -13.8215 pbmt(s,σ,a) 0.75 -13.8215 timeml: 0.75 -13.8215 harnesses 0.140625 -13.8216 than those 0.264706 -13.8217 national institute 0.75 -13.8225 japan’s 0.75 -13.8225 rms 0.75 -13.8225 pos-2 0.75 -13.8225 wn−11 0.75 -13.8225 513 0.75 -13.8225 (taboada 0.75 -13.8225 n’ 0.75 -13.8225 compose, 0.75 -13.8225 mohler 0.75 -13.8225 lost, 0.75 -13.8225 sentimatrix 0.75 -13.8225 chengxiang 0.75 -13.8225 psg 0.571429 -13.8226 concomp 0.571429 -13.8233 prerequisites 0.454545 -13.8235 hmms. 0.75 -13.8235 unlexicalized, 0.75 -13.8235 hell 0.75 -13.8235 uranium 0.75 -13.8235 +0.48 0.75 -13.8235 earnings, 0.75 -13.8235 “software” 0.75 -13.8235 (lemma) 0.75 -13.8235 66.06 0.75 -13.8235 simplex, 0.75 -13.8235 simultaneously: 0.75 -13.8235 ?, 0.75 -13.8235 paradox 0.75 -13.8235 abm*l 0.75 -13.8235 second-stage 0.75 -13.8235 tide 0.75 -13.8235 higherlevel 0.75 -13.8235 sampled, 0.75 -13.8235 67.17 0.75 -13.8235 economic, 0.318182 -13.8235 ], 0.121212 -13.8238 fraction 0.232558 -13.8238 rewritten 0.5 -13.824 word–emotion 0.571429 -13.8241 〈si,cij〉 0.75 -13.8244 module: 0.75 -13.8244 vaswani 0.75 -13.8244 kleene 0.75 -13.8244 94.82 0.75 -13.8244 (simple 0.75 -13.8244 aging 0.75 -13.8244 switches, 0.75 -13.8244 rdm’s 0.75 -13.8244 (kernighan 0.75 -13.8244 adams 0.75 -13.8244 groups; 0.133333 -13.825 relations are 0.75 -13.8254 turn: 0.75 -13.8254 54.55 0.75 -13.8254 -not 0.75 -13.8254 (c1, 0.75 -13.8254 pseudo-rules 0.75 -13.8254 recasing 0.75 -13.8254 76.03 0.75 -13.8254 bleu4 0.75 -13.8254 c7 0.75 -13.8254 schwartz et al. 0.75 -13.8254 macro-f 0.75 -13.8254 u (i;t 0.75 -13.8254 momentum 0.75 -13.8254 monolingual, 0.571429 -13.8256 user experience 0.571429 -13.8256 league 0.571429 -13.8256 leak 0.333333 -13.8258 r4 0.454545 -13.8259 0.0015 0.571429 -13.8263 h(µij) 0.75 -13.8264 c(i) 0.75 -13.8264 ps(x) 0.75 -13.8264 all), 0.75 -13.8264 comic-de 0.75 -13.8264 sonia 0.75 -13.8264 36.16 0.75 -13.8264 line] 0.75 -13.8264 speech-gesture 0.75 -13.8264 (2005b) 0.75 -13.8264 inducers 0.75 -13.8264 remind 0.75 -13.8264 (max, 0.75 -13.8264 texas, 0.75 -13.8264 (smo) 0.75 -13.8264 homologous 0.75 -13.8264 systematics 0.75 -13.8264 alternation: 0.75 -13.8264 sete 0.75 -13.8264 dzikovska, 0.571429 -13.8271 (macherey 0.75 -13.8274 alpage 0.75 -13.8274 homoplasy 0.75 -13.8274 obviate 0.75 -13.8274 e-research 0.75 -13.8274 k-tobi 0.75 -13.8274 tucson, 0.75 -13.8274 excessive news 0.75 -13.8274 pkl 0.75 -13.8274 syllable, 0.75 -13.8274 refered 0.75 -13.8274 neutralize 0.75 -13.8274 believable 0.75 -13.8274 simply, 0.75 -13.8274 28.57 0.5 -13.828 the song 0.75 -13.8284 diffuse 0.75 -13.8284 micr, 0.75 -13.8284 qsds 0.75 -13.8284 1021 0.75 -13.8284 stimulated 0.75 -13.8284 lmf, 0.75 -13.8284 53.19 0.75 -13.8284 behalf 0.75 -13.8284 register, 0.75 -13.8284 (51) 0.571429 -13.8286 wsj1 0.571429 -13.8286 lucia 0.5 -13.8287 /∈ 0.125 -13.829 class, 0.150943 -13.8293 wish to 0.375 -13.8293 trivial. 0.75 -13.8294 activities, 0.75 -13.8294 i-j 0.75 -13.8294 normal-gamma 0.75 -13.8294 26.53 0.75 -13.8294 cotton, 0.75 -13.8294 pivots 0.75 -13.8294 nullnullnull(null) 0.75 -13.8294 factorizing 0.75 -13.8294 (js) 0.75 -13.8294 designates 0.75 -13.8294 philosophers 0.75 -13.8294 pantic 0.75 -13.8294 0.248 0.571429 -13.8294 ses 0.571429 -13.8294 w2” 0.571429 -13.8294 correlation coefficient, 0.5 -13.83 ≥others 0.571429 -13.8301 zpar 0.75 -13.8304 45.35 0.75 -13.8304 staleness 0.75 -13.8304 soon, 0.75 -13.8304 pointed-to 0.75 -13.8304 20%) 0.75 -13.8304 per-corpus 0.75 -13.8304 scope: 0.75 -13.8304 19.37 0.75 -13.8304 (another 0.75 -13.8304 a(x) 0.75 -13.8304 compiled, 0.75 -13.8304 coach, 0.75 -13.8304 christopher, 0.75 -13.8304 65.58 0.75 -13.8304 locations) 0.307692 -13.8306 logarithm 0.0300616 -13.8309 because 0.571429 -13.8309 𝑍 0.571429 -13.8309 climate 0.75 -13.8314 fined 0.75 -13.8314 provides, 0.75 -13.8314 dom 0.75 -13.8314 paek 0.75 -13.8314 93.25 0.75 -13.8314 separated, 0.75 -13.8314 hyperparameters, 0.75 -13.8314 zoning 0.75 -13.8314 [b0,c0]squiggleright[b1,c1]...[bk,ck] 0.75 -13.8314 analog 0.75 -13.8314 34.23 0.75 -13.8314 secretion 0.571429 -13.8316 sentiment representation 0.571429 -13.8316 word-aligned bilingual 0.571429 -13.8316 98.4 0.1 -13.832 sentences with 0.5 -13.832 factoids. 0.75 -13.8324 (ge, 0.75 -13.8324 uš 0.75 -13.8324 splice 0.75 -13.8324 portmanteau 0.75 -13.8324 metabolic 0.75 -13.8324 “received” 0.75 -13.8324 existed, 0.75 -13.8324 sub-part 0.75 -13.8324 1417 0.75 -13.8324 0.491 0.75 -13.8324 i-computations 0.75 -13.8324 (attenberg et 0.75 -13.8324 mylonakis 0.375 -13.8328 t-de 0.454545 -13.8331 news-c 0.142857 -13.8333 stage, 0.75 -13.8334 indicated, 0.75 -13.8334 passe 0.75 -13.8334 0.367 0.75 -13.8334 n∗ 0.75 -13.8334 1016 0.75 -13.8334 tightened 0.75 -13.8334 (currently 0.75 -13.8334 62.23 0.318182 -13.8336 be explored 0.454545 -13.8337 disgust 0.318182 -13.834 active dual 0.454545 -13.8344 .1 0.75 -13.8344 (mbr-sc) 0.75 -13.8344 traders 0.75 -13.8344 “play” 0.75 -13.8344 sa: 0.75 -13.8344 shock 0.75 -13.8344 37.50 0.75 -13.8344 (n=20) 0.75 -13.8344 62.62 0.318182 -13.8344 word structures 0.0649573 -13.8347 2007; 0.75 -13.8354 nite 0.75 -13.8354 mary” 0.75 -13.8354 60.71 0.75 -13.8354 encyclopedia, 0.75 -13.8354 multisource 0.75 -13.8354 transliteration-transcription 0.75 -13.8354 bm, 0.75 -13.8354 deyi 0.75 -13.8354 (parts 0.75 -13.8354 (morphological 0.75 -13.8354 non-redundancy 0.75 -13.8354 atayalic 0.75 -13.8354 abstractive, 0.75 -13.8364 nuclearity 0.75 -13.8364 slangs, 0.75 -13.8364 ten-folds 0.75 -13.8364 edge-flips, 0.75 -13.8364 narow 0.75 -13.8364 popu 0.75 -13.8364 eighty 0.75 -13.8364 false; 0.75 -13.8364 0.597 0.75 -13.8364 s-space 0.75 -13.8364 94.46 0.75 -13.8364 0.035 0.75 -13.8364 ¿ 0.75 -13.8364 head-final 0.75 -13.8364 (esl) 0.75 -13.8364 start-up 0.75 -13.8364 organisms, 0.75 -13.8364 classifierto 0.5 -13.8366 and fabrizio 0.571429 -13.8369 .07 0.5 -13.8373   0.5 -13.8373 0.000 0.000 0.5 -13.8373 farkas 0.75 -13.8374 nicknames, 0.75 -13.8374 -0.1 0.75 -13.8374 from” 0.75 -13.8374 txt 0.75 -13.8374 48, 0.75 -13.8374 (names 0.75 -13.8374 0.252 0.75 -13.8374 pearl 0.75 -13.8374 l2-loss 0.75 -13.8374 ari 0.75 -13.8374 outlining 0.75 -13.8374 j(i) 0.75 -13.8374 articles), 0.75 -13.8374 ωq(w) 0.238095 -13.838 in order to make 0.183099 -13.8381 millions 0.75 -13.8384 khemphwuthe-lul 0.75 -13.8384 0.058 0.75 -13.8384 871 0.75 -13.8384 /u/ 0.75 -13.8384 spokesman 0.75 -13.8384 espresso, 0.75 -13.8384 ˜τ31 0.75 -13.8384 thanks, 0.75 -13.8384 +0.7 0.75 -13.8384 tem-full 0.75 -13.8384 86.37 0.75 -13.8384 ever, 0.75 -13.8384 twostep 0.75 -13.8384 wide-spread 0.168675 -13.8386 intend to 0.416667 -13.839 o(n 0.318182 -13.8391  0.571429 -13.8392 q1, 0.257143 -13.8393 been explored 0.75 -13.8394 xavier 0.75 -13.8394 debates, 0.75 -13.8394 passivization, 0.75 -13.8394 validations 0.75 -13.8394 stmn 0.75 -13.8394 audience, 0.75 -13.8394 “arrange 0.75 -13.8394 true-cased 0.75 -13.8394 parenleftbign2parenrightbig 0.75 -13.8394 memetube, 0.75 -13.8394 chess 0.5 -13.84 bracketrightbig + 0.5 -13.84 ci∈g 0.196721 -13.8401 technique is 0.25 -13.8403 axis 0.75 -13.8404 safe, 0.75 -13.8404 year”, 0.75 -13.8404 lm; 0.75 -13.8404 31.85 0.75 -13.8404 crysmann, 0.75 -13.8404 (schmitt 0.75 -13.8404 362 0.75 -13.8404 death, 0.75 -13.8404 extraposition 0.75 -13.8404 teaching, 0.75 -13.8404 [window]a1 0.75 -13.8404 distinctly 0.75 -13.8404 fact/emotion 0.75 -13.8404 uganda. 0.75 -13.8404 open-license 0.75 -13.8404 890 0.75 -13.8404 wikipedia) 0.75 -13.8404 ‘0’ 0.75 -13.8404 7520 0.257143 -13.841 mind, 0.257143 -13.841 a handful 0.75 -13.8414 purposeof 0.75 -13.8414 reception 0.75 -13.8414 27.55 0.75 -13.8414 s-model 0.75 -13.8414 ananthakrishnan 0.75 -13.8414 atr 0.75 -13.8414 threshold-10 0.75 -13.8414 nna72 0.75 -13.8414 verbnet‟s 0.180556 -13.8414 takes into 0.571429 -13.8415 illustrate, 0.318182 -13.8416 introduction, 0.5 -13.842 monolingually 0.75 -13.8424 74.29 0.75 -13.8424 609 0.75 -13.8424 co-occurrent 0.75 -13.8424 ijcnlp 0.75 -13.8424 92.45 0.75 -13.8424 203), 0.75 -13.8424 (bi) 0.75 -13.8424 isolated, 0.5 -13.8426 ∆, 0.75 -13.8434 “language 0.75 -13.8434 bursts 0.75 -13.8434 fra 0.75 -13.8434 damaged 0.75 -13.8434 celebrities, 0.75 -13.8434 526 0.75 -13.8434 0.635 0.571429 -13.8438 k-prefix 0.104545 -13.844 pair, 0.416667 -13.8441 vertically 0.35 -13.8443  0.75 -13.8444 80.18 0.75 -13.8444 27.23 0.75 -13.8444 single-sided 0.75 -13.8444 31.69 0.75 -13.8444 assembling 0.75 -13.8444 kit, 0.75 -13.8444 90%) 0.75 -13.8444 93.49 0.75 -13.8444 tation 0.75 -13.8444 3.50 0.75 -13.8444 hierachical 0.75 -13.8444 denero, 0.75 -13.8444 e.g: 0.75 -13.8444 o(q)j 0.75 -13.8444 amount, 0.75 -13.8444 f-value 0.75 -13.8444 intuitive: 0.75 -13.8444 6see 0.5 -13.8446 wiliam 0.4 -13.8448 furniture 0.416667 -13.8452 = 0.01 0.416667 -13.8452 especially true 0.5 -13.8453 (mt 0.75 -13.8454 τ: 0.75 -13.8454 vibrant 0.75 -13.8454 devset 0.75 -13.8454 participants: 0.75 -13.8454 hoste 0.75 -13.8454 (hopkins 0.75 -13.8454 speakers) 0.75 -13.8454 lip 0.75 -13.8454 fan-out, 0.75 -13.8454 1−α 0.75 -13.8454 parentheticals 0.75 -13.8454 es-en, 0.75 -13.8454 coast, 0.75 -13.8454 mterater, 0.75 -13.8454 fill-in-the-blank 0.5 -13.846 dpi, 0.75 -13.8464 wanna 0.75 -13.8464 49.70 0.75 -13.8464 84.77 0.75 -13.8464 45.63 0.75 -13.8464 comunications 0.75 -13.8464 over-segmentation 0.75 -13.8464 carbonell, 0.75 -13.8464 (considering 0.75 -13.8464 sub-section 0.75 -13.8464 2-21), 0.75 -13.8464 twentieth 0.454545 -13.8465 pro-best 0.454545 -13.8471 wd,n 0.178082 -13.8472 selected. 0.5 -13.8473 (context 0.75 -13.8474 incidents 0.75 -13.8474 photo 0.75 -13.8474 988 0.75 -13.8474 dp2 0.75 -13.8474 58.95 0.75 -13.8474 86.95 0.75 -13.8474 (exact) 0.75 -13.8474 cho 0.75 -13.8474 25th 0.75 -13.8474 dampening 0.75 -13.8474 cells; 0.75 -13.8474 g’s 0.307692 -13.8475 arc-eager 0.571429 -13.8477 greece, 0.454545 -13.8483 both tasks. 0.75 -13.8484 vectorspace 0.75 -13.8484 (topic 0.75 -13.8484 newspaper text. 0.75 -13.8484 14.29 0.75 -13.8484 dialogues: 0.75 -13.8484 “root” 0.75 -13.8484 exploring, 0.75 -13.8484 (karmarkar, 0.571429 -13.8484 r−1 0.188406 -13.8487 these rules 0.5 -13.8487 oliver 0.454545 -13.849 edward 0.571429 -13.8492 to accomplish this, 0.75 -13.8494 mi,j 0.75 -13.8494 1.79 0.75 -13.8494 buckley, 0.75 -13.8494 runs; 0.75 -13.8494 αsrc 0.75 -13.8494 path: 0.75 -13.8494 recognition: 0.75 -13.8494 (has 0.75 -13.8494 pedersen’s 0.5 -13.8494 menu-based 0.5 -13.8494 hr0011-06-c-0022 0.5 -13.8494 (allen, 0.5 -13.8494 many interesting 0.105991 -13.8496 basis of 0.5 -13.85 surprisal, 0.416667 -13.8502 label distributions 0.75 -13.8504 suggests, 0.75 -13.8504 politician 0.75 -13.8504 allfeatures 0.75 -13.8504 museums, 0.75 -13.8504 nest 0.75 -13.8504 α(v) 0.75 -13.8504 62.96 0.75 -13.8504 rsttool 0.75 -13.8504 stepped 0.75 -13.8504 dct, 0.75 -13.8504 44.75 0.75 -13.8504 39.58 0.352941 -13.8506 formula, 0.375 -13.8514 concordance 0.75 -13.8515 abnormal 0.75 -13.8515 2,400 0.75 -13.8515 theorem, 0.75 -13.8515 experts: 0.75 -13.8515 literature: 0.75 -13.8515 reviews) 0.75 -13.8515 lexicography 0.75 -13.8515 disparities 0.75 -13.8515 credentials 0.75 -13.8515 iv) 0.75 -13.8515 pt(z), 0.75 -13.8515 9.7 0.227273 -13.852 each word, 0.352941 -13.8521 (whether 0.185714 -13.8521 descriptions. 0.75 -13.8525 37m 0.75 -13.8525 gs, 0.75 -13.8525 eks 0.75 -13.8525 (default) 0.75 -13.8525 source), 0.75 -13.8525 stt 0.75 -13.8525 re-write 0.75 -13.8525 38.71 0.75 -13.8525 upwards 0.75 -13.8525 nap 0.75 -13.8525 minimization, 0.75 -13.8525 94.38 0.75 -13.8525 indonesian 0.75 -13.8525 language” 0.352941 -13.8526 were converted 0.571429 -13.8531 considerably better 0.75 -13.8535 tea, 0.75 -13.8535 nullnullnull1 0.75 -13.8535 bonds 0.75 -13.8535 nsubj, 0.75 -13.8535 metaphorical, 0.75 -13.8535 grefenstette, 0.75 -13.8535 1538 0.75 -13.8535 scientist 0.75 -13.8535 periphrasis 0.571429 -13.8538 (poesio 0.5 -13.8541 romania 0.75 -13.8545 1.82 0.75 -13.8545 “close 0.75 -13.8545 (yi 0.75 -13.8545 resolved, 0.75 -13.8545 broadcasts 0.75 -13.8545 0.830 0.75 -13.8545 re, 0.75 -13.8545 assortivity 0.75 -13.8545 abe 0.75 -13.8545 builder 0.75 -13.8545 84.37 0.75 -13.8545 counter, 0.75 -13.8545 xiaodong 0.75 -13.8555 tier-3 0.75 -13.8555 nara, 0.75 -13.8555 3.5.3 0.75 -13.8555 (contiguous) 0.75 -13.8555 ralf 0.75 -13.8555 (33 0.75 -13.8555 mats 0.75 -13.8555 nec 0.75 -13.8555 adjudicator 0.75 -13.8555 launch 0.75 -13.8555 476 0.75 -13.8555 srihari, 0.75 -13.8555 enclose 0.75 -13.8555 ixa taldea 0.75 -13.8555 θne+de 0.75 -13.8555 watkinson’s 0.75 -13.8555 usability, 0.75 -13.8555 (carlson, 0.257143 -13.8555 using just 0.75 -13.8565 #: 0.75 -13.8565 rh 0.75 -13.8565 cv-em 0.75 -13.8565 (8-best 0.75 -13.8565 (43 0.75 -13.8565 interjection 0.75 -13.8565 al$yx, 0.75 -13.8565 1,300 0.75 -13.8565 37.32 0.75 -13.8565 4.31 0.75 -13.8565 1.28 0.215686 -13.8565 tests. 0.5 -13.8568 x1, 0.5 -13.8568 finite-length 0.175676 -13.857 (2010) and 0.35 -13.8575 t-test. 0.416667 -13.8576 get: 0.75 -13.8576 map-based 0.75 -13.8576 cloze 0.75 -13.8576 “right 0.75 -13.8576 61.30 0.75 -13.8576 inflectionally 0.75 -13.8576 possessing 0.75 -13.8576 “somewhat 0.75 -13.8576 (las), 0.75 -13.8576 victim 0.571429 -13.8577 p[pi] 0.571429 -13.8577 anc 0.275862 -13.8581 ◦ 0.101266 -13.8581 we compute 0.454545 -13.8582 (you 0.318182 -13.8584 summationtextn 0.571429 -13.8585 0.352 0.75 -13.8586 wi) 0.75 -13.8586 passing, 0.75 -13.8586 ziff-davis 0.75 -13.8586 33.90 0.75 -13.8586 ×100% 0.75 -13.8586 13.1 0.75 -13.8586 “words 0.75 -13.8586 questionanswering 0.75 -13.8586 denominators 0.416667 -13.8587 frequent words, 0.454545 -13.8588 densities. 0.454545 -13.8588 community’s 0.0702479 -13.8589 5: 0.5 -13.8595 ubiquitination 0.5 -13.8595 n∑ 0.75 -13.8596 charniak-johnson 0.75 -13.8596 (scfgs) 0.75 -13.8596 33.3% 0.75 -13.8596 concept’s 0.75 -13.8596 gothenburg, 0.75 -13.8596 colston 0.75 -13.8596 rate: 0.75 -13.8596 (peer 0.75 -13.8596 hasn’t 0.375 -13.8596 noun vectors 0.275862 -13.8604 (habash 0.75 -13.8606 (vector 0.75 -13.8606 disfavored 0.75 -13.8606 mom 0.75 -13.8606 hofmann, 0.75 -13.8606 topic’s 0.75 -13.8606 “c” 0.264706 -13.8606 trigram language 0.571429 -13.8608 verhagen, 0.333333 -13.8612 annotators’ 0.75 -13.8616 (2-way 0.75 -13.8616 vernacular 0.75 -13.8616 english-italian 0.75 -13.8616 years: 0.75 -13.8616 beeferman 0.75 -13.8616 hendriks, 0.75 -13.8616 highperformance 0.5 -13.8616 ccc 0.5 -13.8623 sprime, 0.571429 -13.8624 vvf 0.0740741 -13.8626 finally, we 0.75 -13.8627 estonian, 0.75 -13.8627 unigrams: 0.75 -13.8627 wers 0.75 -13.8627 metaphoric 0.75 -13.8627 7.94 0.75 -13.8627 (komachi 0.571429 -13.8632 [] 0.571429 -13.8632 shankar 0.75 -13.8637 (comparable 0.75 -13.8637 left-pop 0.75 -13.8637 celebration 0.75 -13.8637 jamie 0.75 -13.8637 component: 0.75 -13.8637 nhr 0.75 -13.8637 knowledge-base 0.75 -13.8637 sys5 0.75 -13.8637 h2: 0.75 -13.8637 administer 0.75 -13.8637 out-performs 0.75 -13.8637 hadn’t 0.75 -13.8637 tripled 0.75 -13.8637 threats 0.185714 -13.864 portions 0.416667 -13.8644 ambiguous word. 0.416667 -13.8644 .4 0.75 -13.8647 65.92 0.75 -13.8647 i=n+1 0.75 -13.8647 katz-brown 0.75 -13.8647 utterance-based 0.75 -13.8647 renaming, 0.75 -13.8647 1963; 0.75 -13.8647 left-arc, 0.75 -13.8647 39.69 0.75 -13.8647 “get” 0.285714 -13.8652 1, where 0.75 -13.8658 446 0.75 -13.8658 (mwus) 0.75 -13.8658 fault 0.75 -13.8658 functions; 0.75 -13.8658 interlocutor’s 0.75 -13.8658 ∀1≤k≤d 0.375 -13.8667 adjpair 0.75 -13.8668 cut-offs 0.75 -13.8668 subscripts) 0.75 -13.8668 canonicalword 0.75 -13.8668 (2%) 0.75 -13.8668 (10% 0.75 -13.8668 moses-based 0.75 -13.8668 29.18 0.75 -13.8668 0.703 0.75 -13.8668 soegaard@hum.ku.dk 0.75 -13.8668 53.03 0.75 -13.8668 surveilance 0.227273 -13.8669 sections. 0.571429 -13.8671 delmas 0.5 -13.8671  0.416667 -13.8678 83.3 0.416667 -13.8678 ∀ 0.75 -13.8678 chronologically 0.75 -13.8678 +weights 0.75 -13.8678 {0,1} 0.75 -13.8678 (ozkan 0.75 -13.8678 profits 0.75 -13.8678 [and 0.75 -13.8678 1an 0.75 -13.8678 fluidity 0.75 -13.8678 expansion.conjunction 0.75 -13.8688 6.58 0.75 -13.8688 objects: 0.75 -13.8688 (bf) 0.75 -13.8688 lemmas) 0.75 -13.8688 ν∞ 0.75 -13.8688 spector 0.75 -13.8688 477 0.75 -13.8688 noun; 0.75 -13.8688 tk, 0.4 -13.8688 navigation, 0.5 -13.8691 swim 0.08 -13.8698 calculate 0.75 -13.8699 to), 0.75 -13.8699 53.42 0.75 -13.8699 4since 0.75 -13.8699 17.50 0.75 -13.8699 coals-based 0.75 -13.8699 clock 0.75 -13.8699 1.47 0.571429 -13.8702 20.2 0.318182 -13.8706 cup 0.75 -13.8709 2/1 0.75 -13.8709 γ0 0.75 -13.8709 converging 0.75 -13.8709 010 0.75 -13.8709 stack-based 0.75 -13.8709 1999); 0.571429 -13.871 fβ 0.0654545 -13.8715 words are 0.0562249 -13.8716 results are 0.257143 -13.8717 3.1, 0.571429 -13.8718 they mainly 0.75 -13.8719 σn 0.75 -13.8719 4.49 0.75 -13.8719 0.292 0.75 -13.8719 high-throughput 0.75 -13.8719 0.087 0.75 -13.8719 adjunct, 0.75 -13.8719 (λ1, 0.75 -13.8719 pt(x) 0.75 -13.8719 sociological 0.75 -13.8719 morrissey 0.157895 -13.872 collection. 0.296296 -13.8728 cloud 0.375 -13.8729 464 0.75 -13.873 likely) 0.75 -13.873 school” 0.75 -13.873 interior, 0.75 -13.873 hospitals 0.75 -13.873 1.73 0.75 -13.873 misled 0.75 -13.873 rmd, 0.75 -13.873 definites 0.75 -13.873 33.81 0.75 -13.873 27.11 0.75 -13.873 glossed 0.75 -13.873 λ: 0.75 -13.873 drawn, 0.5 -13.8733 all categories. 0.5 -13.8733 ondˇrej 0.5 -13.8733 coffee, 0.149533 -13.8736 aspect of 0.257143 -13.8739 this includes 0.116959 -13.8739 step. 0.75 -13.874 wikimedia 0.75 -13.874 invariants 0.75 -13.874 out-of-ten 0.75 -13.874 bangalore 0.75 -13.874 feat-ii 0.75 -13.874 (auc) 0.75 -13.874 position j, 0.75 -13.874 (namely, 0.75 -13.874 902 0.75 -13.874 dancing 0.571429 -13.8742 coreference: 0.168675 -13.8743 roth, 0.454545 -13.8744 an interlingual 0.227273 -13.8747 this way, we 0.75 -13.8751 17.5% 0.75 -13.8751 firms 0.75 -13.8751 rn, 0.75 -13.8751 to/from 0.75 -13.8751 wmt-07 0.75 -13.8751 palais 0.75 -13.8751 dvd-en 0.571429 -13.8758 sbml/celldesigner 0.416667 -13.8759 with: 0.5 -13.876 faith 0.5 -13.876 hum 0.5 -13.876 ana 0.75 -13.8761 (f-measure 0.75 -13.8761 (correctly) 0.75 -13.8761 wmpf’ 0.75 -13.8761 1304 0.75 -13.8761 long-form, 0.75 -13.8761 forwarded 0.75 -13.8761 soir 0.75 -13.8761 owes 0.75 -13.8761 anatomy 0.75 -13.8761 87.27 0.75 -13.8761 52.0% 0.75 -13.8761 mcdonald,lerman,and 0.416667 -13.877 moreover, they 0.75 -13.8771 jared 0.75 -13.8771 nixon 0.75 -13.8771 b1, 0.75 -13.8771 (2006) reported 0.75 -13.8771 42.20 0.75 -13.8771 1348 0.75 -13.8771 ltd 0.75 -13.8771 wprimeij 0.75 -13.8771 english-korean 0.75 -13.8771 match_count 0.75 -13.8771 ob 0.5 -13.8774 others 0.75 -13.8939 personnel 0.75 -13.8939 mme 0.129496 -13.8939 ranging 0.6 -13.8941 metrics; 0.6 -13.8941 (erp) 0.6 -13.8941 augmentation, 0.6 -13.8941 pr(w) 0.6 -13.8941 10s 0.5 -13.8942 restrictions: 0.5 -13.8942 optimality, 0.196721 -13.8944 all experiments 0.145455 -13.8947 1999). 0.571429 -13.8949 argentina, 0.6 -13.8949 per-soc 0.6 -13.8949 projection: 0.6 -13.8949 ngan 0.6 -13.8949 (semi-)structured 0.6 -13.8949 “overlap” 0.6 -13.8949 roi 0.6 -13.8949 routes 0.5 -13.8949 0.815 0.5 -13.8949 doc-pair 0.75 -13.895 1031 0.75 -13.895 radiology 0.75 -13.895 prince, 0.75 -13.895 illinois, 0.75 -13.895 n−i 0.333333 -13.895 (null 0.0752427 -13.8951 paper. 0.352941 -13.8955 negative examples, 0.5 -13.8956 “positive” 0.571429 -13.8958 y(q) 0.6 -13.8958 dice, 0.6 -13.8958 transitions: 0.6 -13.8958 ‘noisier’ 0.6 -13.8958 xxx 0.352941 -13.896 )= 0.75 -13.896 “let’s 0.75 -13.896 munged 0.75 -13.896 lh1(null) 0.75 -13.896 throut% 0.75 -13.896 bpt 0.75 -13.896 90.8% 0.75 -13.896 within-system 0.75 -13.896 positions: 0.333333 -13.8964 an electronic 0.571429 -13.8966 lperm 0.6 -13.8966 ˆ, 0.6 -13.8966 continuum, 0.416667 -13.8969 ext 0.75 -13.8971 modes: 0.75 -13.8971 dp: 0.75 -13.8971 enhancer 0.75 -13.8971 0.076 0.571429 -13.8974 hyponyms, 0.6 -13.8975 (pairs 0.6 -13.8975 ltd., 0.181818 -13.898 principle, 0.75 -13.8982 w/inference 0.75 -13.8982 ibm-4 0.75 -13.8982 721 0.75 -13.8982 zealand 0.75 -13.8982 optimizer, 0.75 -13.8982 buy/sell 0.75 -13.8982 modifier) 0.75 -13.8982 50.51 0.75 -13.8982 (actor, 0.571429 -13.8982 pos(t) 0.217391 -13.8983 ignored. 0.6 -13.8983 sox 0.196429 -13.8984 should also 0.571429 -13.899 sudan 0.416667 -13.8992 searchbench 0.75 -13.8992 state) 0.75 -13.8992 kruskal-wallis 0.75 -13.8992 25.23 0.75 -13.8992 statccg 0.75 -13.8992 19.78 0.75 -13.8992 acl-89 0.207547 -13.8992 high. 0.6 -13.8992 inria 0.6 -13.8992 (w1,w2) 0.6 -13.8992 acceptable, 0.6 -13.9 cons2str 0.75 -13.9003 brin 0.75 -13.9003 command, 0.75 -13.9003 predictably 0.75 -13.9003 governments 0.75 -13.9003 acquisition: 0.75 -13.9003 adik 0.75 -13.9003 present: 0.75 -13.9003 mismatched 0.75 -13.9003 789 0.454545 -13.9005 intelligence advanced 0.571429 -13.9006 beatrice 0.5 -13.9006 (good 0.6 -13.9009 mjk 0.75 -13.9014 t3: 0.75 -13.9014 drop, 0.75 -13.9014 re-estimate 0.571429 -13.9014 talkamatic 0.6 -13.9017 refuted 0.6 -13.9017 ones: 0.75 -13.9024 (par) 0.75 -13.9024 with1 0.75 -13.9024 “about 0.75 -13.9024 15k 0.75 -13.9024 wordlda, 0.75 -13.9024 lai, 0.75 -13.9024 amelioration 0.75 -13.9024 43m 0.75 -13.9024 918 0.0268601 -13.9024 would 0.454545 -13.9025 test this hypothesis, 0.6 -13.9026 93.28 0.6 -13.9026 workload 0.6 -13.9026 ce, 0.6 -13.9026 wg 0.6 -13.9026 5-10 0.5 -13.9027 absolutive 0.416667 -13.9034 partly supported 0.75 -13.9035 x-ray 0.75 -13.9035 nz 0.75 -13.9035 wrong; 0.110526 -13.9035 topic. 0.6 -13.9035 56.41 0.6 -13.9035 10027 0.6 -13.9035 brother 0.352941 -13.904 wie 0.6 -13.9043 links) 0.6 -13.9043 at my 0.75 -13.9046 documentspecific 0.75 -13.9046 (dct) 0.75 -13.9046 (pennacchiotti 0.75 -13.9046 k33k20 0.75 -13.9046 0.618 0.75 -13.9046 ‘many 0.75 -13.9046 executives 0.5 -13.9049 – draw 0.6 -13.9052 ,and 0.6 -13.9052 helen 0.75 -13.9056 (headden 0.75 -13.9056 (performance 0.75 -13.9056 (name 0.75 -13.9056 costa 0.75 -13.9056 listeners, 0.75 -13.9056 (psd) 0.75 -13.9056 adjudication, 0.75 -13.9056 b3,b2,b1 0.6 -13.906 “er 0.6 -13.906 silvia 0.6 -13.906 98.12 0.6 -13.906 interpreter 0.571429 -13.9063 si−1 0.416667 -13.9064 bone 0.75 -13.9067 punishes 0.75 -13.9067 0.495 0.75 -13.9067 “china 0.75 -13.9067 identifies 0.75 -13.9067 fully-lexicalized 0.75 -13.9067 speech) 0.75 -13.9067 lw 0.75 -13.9067 0.07, 0.75 -13.9067 domain-assisted 0.75 -13.9067 comprehenders’ 0.75 -13.9067 93.42 0.75 -13.9067 k−1 0.75 -13.9067 (yes/no) 0.0748792 -13.9069 plan 0.6 -13.9069 technically, 0.6 -13.9069 cs09 0.6 -13.9069 gerunds 0.416667 -13.907 interand 0.454545 -13.9077 carolyn 0.6 -13.9077 2.81 0.6 -13.9077 hierarchy: 0.6 -13.9077 8for 0.75 -13.9078 strings: 0.75 -13.9078 767 0.75 -13.9078 (implicit) 0.75 -13.9078 (ab), 0.75 -13.9078 88.59 0.75 -13.9078 confirmations, 0.75 -13.9078 prosodically 0.75 -13.9078 piek 0.75 -13.9078 roland 0.75 -13.9078 reordering patterns. 0.211538 -13.9081 (liang 0.217391 -13.9085 connection between 0.6 -13.9086 stepsize 0.6 -13.9086 538 0.6 -13.9086 runing 0.75 -13.9089 procedure outlined 0.75 -13.9089 sportsman 0.75 -13.9089 headquarters 0.75 -13.9089 52.95 0.75 -13.9089 φ insert 0.75 -13.9089 trinity 0.193548 -13.909 effort. 0.0902527 -13.909 to automatically 0.6 -13.9095 odd, 0.6 -13.9095 circumstance 0.571429 -13.9096 [conceiver 0.571429 -13.9096 de-en, 0.75 -13.9099 responsibilities 0.75 -13.9099 balls 0.75 -13.9099 decisions: 0.75 -13.9099 31.58 0.5 -13.9099 meyer 0.6 -13.9103 s083 0.6 -13.9103 (j) 0.6 -13.9103 ♦ 0.6 -13.9103 emerson 0.375 -13.9104 link, 0.111111 -13.9106 table 5 0.5 -13.9106 series) 0.75 -13.911 cnlvcs 0.75 -13.911 (michel 0.75 -13.911 zettair 0.75 -13.911 bengali, 0.75 -13.911 1420 0.571429 -13.9112 corr(me-seq) 0.6 -13.9112 2004; yang 0.6 -13.9112 lcfrss, 0.6 -13.9112 w(d) 0.6 -13.9112 trevor 0.5 -13.9113 percy 0.454545 -13.9116 (3a) 0.454545 -13.9116 fighter 0.571429 -13.912 four algorithms 0.6 -13.912 annie 0.6 -13.912 aue 0.6 -13.912 mrs, 0.75 -13.9121 38.02 0.75 -13.9121 “yes”, 0.75 -13.9121 semantizer 0.75 -13.9121 left-context 0.75 -13.9121 mturkswsd 0.75 -13.9121 substitutive 0.75 -13.9121 24.89 0.75 -13.9121 tectogrammatics 0.75 -13.9121 94.48 0.75 -13.9121 cai, 0.75 -13.9121 argadj 0.571429 -13.9129 uninterpretable 0.6 -13.9129 anonymization 0.6 -13.9129 tournament 0.6 -13.9129 1526 0.75 -13.9132 b=50 0.75 -13.9132 abilities, 0.75 -13.9132 semcor, 0.75 -13.9132 t-score, 0.75 -13.9132 2006b) 0.75 -13.9132 esr-1 0.454545 -13.9135 and richard 0.5 -13.9135 that-complementizer 0.6 -13.9138 431 0.6 -13.9138 suites 0.6 -13.9138 441 0.6 -13.9138 invested 0.6 -13.9138 previous findings 0.6 -13.9138 (base) 0.192982 -13.9142 (y 0.75 -13.9142 vlps 0.75 -13.9142 conformance 0.75 -13.9142 0.629 0.75 -13.9142 artm 0.125 -13.9142 documents are 0.5 -13.9142 (n-gram) 0.571429 -13.9145 0.682 0.571429 -13.9145 saarbr¨ucken, 0.571429 -13.9145 [s 0.6 -13.9146 puisque 0.6 -13.9146 land, 0.375 -13.9147 deduced 0.318182 -13.915 prob 0.75 -13.9153 swift 0.75 -13.9153 tables: 0.75 -13.9153 langue 0.75 -13.9153 (proposed 0.75 -13.9153 col. 0.75 -13.9153 inaccurate, 0.75 -13.9153 vo 0.75 -13.9153 3904 0.75 -13.9153 woody 0.75 -13.9153 type-raising 0.75 -13.9153 "* 0.75 -13.9153 θc 0.75 -13.9153 screenshots 0.571429 -13.9153 nakazawa et 0.571429 -13.9153 prefix. 0.375 -13.9153 wealth 0.6 -13.9155 24th 0.6 -13.9155 50.03 0.291667 -13.9155 comparable results 0.166667 -13.9157 string. 0.236842 -13.9157 message. 0.291667 -13.9159 may result 0.5 -13.9163 georgia 0.75 -13.9164 textrunner’s 0.75 -13.9164 regulonmember 0.75 -13.9164 579 0.75 -13.9164 career, 0.75 -13.9164 motion) 0.75 -13.9164 writen 0.75 -13.9164 ly 0.75 -13.9164 aide 0.6 -13.9164 performedon 0.6 -13.9164 12c 0.6 -13.9164 49.5% 0.6 -13.9164 λv 0.571429 -13.917 carpenter, 0.571429 -13.917 0.314 0.5 -13.9171 206; 0.6 -13.9172 cmu-heafield-combo 0.6 -13.9172 (synset 0.6 -13.9172 (start 0.6 -13.9172 11.89 0.6 -13.9172 blogger 0.75 -13.9175 fixed-length 0.75 -13.9175 1473 0.75 -13.9175 divorce 0.75 -13.9175 test08 0.75 -13.9175 drifted 0.75 -13.9175 men: 0.75 -13.9175 lh1(null). 0.75 -13.9175 tasks) 0.75 -13.9175 958 0.571429 -13.9178 ron 0.6 -13.9181 lists: 0.6 -13.9181 argm-cau 0.75 -13.9186 reducible 0.75 -13.9186 dramatically, 0.75 -13.9186 θe 0.75 -13.9186 “is” 0.75 -13.9186 aside, 0.75 -13.9186 (tokens) 0.75 -13.9186 cross-sections 0.75 -13.9186 @username 0.75 -13.9186 austria 0.318182 -13.9186 discuss some 0.6 -13.919 oberlander 0.6 -13.919 beam-search, 0.6 -13.919 twist 0.275862 -13.9193 two clusters 0.285714 -13.9195 copies 0.571429 -13.9195 sixteen 0.571429 -13.9195 eiichiro 0.75 -13.9196 (gravano 0.75 -13.9196 634 0.75 -13.9196 neighbors, 0.75 -13.9196 ato 0.75 -13.9196 appreciation 0.75 -13.9196 radicalbigg 0.75 -13.9196 26.83 0.6 -13.9198  the  0.6 -13.9198 ∞, 0.454545 -13.9201 are essential 0.179104 -13.9201 landmark 0.75 -13.9207 presupposing 0.75 -13.9207 ce-candidates 0.75 -13.9207 non-actionable 0.75 -13.9207 (94) 0.6 -13.9207 grammars: 0.5 -13.9207 xsi,j 0.5 -13.9207 0.733 0.318182 -13.9214 pushes 0.6 -13.9216 clasify 0.6 -13.9216 2.9% 0.6 -13.9216 columns: 0.0833333 -13.9217 6: 0.75 -13.9218 58.42 0.75 -13.9218 68.33 0.75 -13.9218 historybased 0.75 -13.9218 91.02 0.75 -13.9218 non-regularized 0.75 -13.9218 personally 0.75 -13.9218 underlining 0.75 -13.9218 mi2 0.75 -13.9218 surroundings 0.75 -13.9218 83.12 0.454545 -13.9221 stsg, 0.6 -13.9224 /t/ 0.6 -13.9224 wisconsin-madison 0.6 -13.9224 emile 0.454545 -13.9227 inherits 0.5 -13.9228 meth 0.75 -13.9229 sub-sampled 0.75 -13.9229 0.01), 0.75 -13.9229 w∈w 0.75 -13.9229 bilexicalized 0.75 -13.9229 ceased 0.75 -13.9229 transcriptional 0.75 -13.9229 endeavor 0.75 -13.9229 if≻x 0.75 -13.9229 act: 0.75 -13.9229 1161 0.416667 -13.9232 subdivided 0.6 -13.9233 (true 0.6 -13.9233 crfs, 0.6 -13.9233 combatant. 0.6 -13.9233 (est) 0.75 -13.924 uvt-wsd 0.75 -13.924 arcs), 0.75 -13.924 conference: 0.75 -13.924 10p 0.75 -13.924 42.23 0.75 -13.924 divisive 0.75 -13.924 bead 0.75 -13.924 edition, 0.75 -13.924 16.77 0.6 -13.9242 2.1: 0.6 -13.9242 1970), 0.6 -13.9242 s(k,lscript) 0.6 -13.9242 whole process 0.6 -13.9242 reviewlength 0.222222 -13.9246 supposed 0.6 -13.925 sport 0.6 -13.925 300000 0.6 -13.925 always) 0.6 -13.925 s-full 0.5 -13.925 fac 0.75 -13.9251 (t,c) 0.75 -13.9251 tsochantaridis 0.75 -13.9251 p(r) 0.75 -13.9251 (spot list 0.318182 -13.9251 l=1 0.571429 -13.9253 reg+ 0.384615 -13.9257 vino 0.5 -13.9258 feldman 0.6 -13.9259 insufficiently 0.6 -13.9259 templatic 0.6 -13.9259 common: 0.571429 -13.9261 charts, 0.571429 -13.9261 (hassan 0.571429 -13.9261 “both 0.571429 -13.9261 applied; 0.375 -13.9261 and daniel 0.75 -13.9262 p(z) 0.75 -13.9262 combinationof 0.75 -13.9262 2.5k 0.75 -13.9262 disseminated 0.75 -13.9262 88.76 0.75 -13.9262 logr 0.75 -13.9262 90.02 0.75 -13.9262 su4 0.75 -13.9262 37.91 0.318182 -13.9265 gold part-of-speech 0.6 -13.9268 room” 0.571429 -13.9269 de madrid 0.571429 -13.9269 crisis. 0.571429 -13.9269 wsem. 0.571429 -13.9269 she denies 0.571429 -13.9269 bonjour 0.304348 -13.927 appears. 0.145631 -13.9271 standard. 0.75 -13.9273 86400 0.75 -13.9273 statement: 0.75 -13.9273 icd9 0.75 -13.9273 non-definition 0.75 -13.9273 +hypo 0.75 -13.9273 feder 0.75 -13.9273 federalist 0.75 -13.9273 talkers 0.75 -13.9273 topologies 0.75 -13.9273 (become 0.333333 -13.9275 are judged 0.6 -13.9277 lproductdisplay 0.6 -13.9277 multicommodity 0.6 -13.9277 stm 0.5 -13.9279 ei1 0.155556 -13.9282 is built 0.75 -13.9284 top-down, 0.75 -13.9284 anchor-title 0.6 -13.9285 (‡) 0.6 -13.9285 (manually 0.6 -13.9285 tcs 0.6 -13.9285 6-3 0.6 -13.9285 reweighted 0.0722611 -13.9289 algorithm. 0.0725995 -13.9291 that were 0.6 -13.9294 r; 0.6 -13.9294 vectorλ 0.6 -13.9294 diacritized 0.6 -13.9294 t(69) 0.6 -13.9294 (calvo 0.236842 -13.9295 combined model 0.75 -13.9295 broder 0.75 -13.9295 1,082 0.75 -13.9295 stack: 0.75 -13.9295 llr, 0.6 -13.9303 (good) 0.75 -13.9306 p+t 0.75 -13.9306 champion 0.75 -13.9306 sition 0.75 -13.9306 works: 0.75 -13.9306 vegas, 0.75 -13.9306 ccgs 0.75 -13.9306 oportunity 0.75 -13.9306 (lts) 0.75 -13.9306 ao’s, 0.454545 -13.9307 competitively 0.168831 -13.9311 commentary 0.6 -13.9312 (25%) 0.6 -13.9312 θm 0.6 -13.9312 rose-regpos 0.6 -13.9312 verbal, 0.6 -13.9312 q1: 0.5 -13.9316 (fa) 0.75 -13.9317 (above) 0.75 -13.9317 exchanges, 0.75 -13.9317 (g 0.75 -13.9317 (strict 0.75 -13.9317 previous/next 0.75 -13.9317 scorer, 0.75 -13.9317 (ls) 0.75 -13.9317 googleweb, 0.75 -13.9317 1089 0.75 -13.9317 1275 0.75 -13.9317 probability-based 0.275862 -13.9321 zollmann 0.148515 -13.9324 point in 0.291667 -13.9325 was necessary 0.75 -13.9328 (social 0.75 -13.9328 nick 0.75 -13.9328 books: 0.75 -13.9328 (essentially 0.75 -13.9328 warmer 0.75 -13.9328 bril 0.75 -13.9328 multiplies 0.6 -13.9329 s-pos 0.6 -13.9329 xt, 0.5 -13.9331 dublin, 0.5 -13.9331 reap 0.6 -13.9338 “obama 0.6 -13.9338 contraction 0.6 -13.9338 2011); 0.6 -13.9338 chromosomes 0.75 -13.9339 (s2) 0.75 -13.9339 cabral, 0.75 -13.9339 (small) 0.75 -13.9339 ados 0.75 -13.9339 p@10 0.75 -13.9339 mean: 0.75 -13.9339 nodes), 0.75 -13.9339 θd, 0.75 -13.9339 694 0.75 -13.9339 iprime 0.416667 -13.9347 np@1 0.6 -13.9347 broadcasting 0.75 -13.935 (q,f,q 0.75 -13.935 ren, 0.75 -13.935 (in-domain) 0.75 -13.935 sandhi 0.75 -13.935 / 47 0.5 -13.9353 (jj) 0.5 -13.9353 (integer) 0.6 -13.9356 pir1 0.6 -13.9356 community: 0.75 -13.9361 karl 0.75 -13.9361 hypergraphs, 0.75 -13.9361 66.98 0.75 -13.9361 18.12 0.75 -13.9361 reformulations 0.75 -13.9361 nonrecursive 0.75 -13.9361 (krallinger 0.75 -13.9361 notc 0.75 -13.9361 96.19 0.6 -13.9365 85.45 0.6 -13.9365 opjk 0.6 -13.9365 sbar, 0.5 -13.9367 j∈i0 0.75 -13.9372 eoc 0.75 -13.9372 aristotle 0.75 -13.9372 trie-based 0.75 -13.9372 volcano 0.75 -13.9372 (form) 0.75 -13.9372 gut 0.6 -13.9374 “forget 0.6 -13.9374 tornar 0.6 -13.9374 l’ 0.6 -13.9374 painting 0.6 -13.9374 arise, 0.5 -13.9375 things: 0.230769 -13.9379 candidate. 0.258065 -13.9379 ganchev et 0.6 -13.9382 tas 0.6 -13.9382 nominal, 0.75 -13.9383 analogies, 0.75 -13.9383 [noun] 0.75 -13.9383 subject: 0.75 -13.9383 pinyin-correction 0.416667 -13.9384 yue 0.5 -13.939 error types. 0.6 -13.9391 bioinformatics, 0.6 -13.9391 system” 0.6 -13.9391 3.11 0.6 -13.9391 n+1, 0.6 -13.9391 quarterly 0.6 -13.9391 978 0.6 -13.9391 relatives 0.75 -13.9394 run) 0.75 -13.9394 556 0.75 -13.9394 eigenvalues 0.75 -13.9394 s3h, 0.75 -13.9394 75.50 0.75 -13.9394 a[pi]γl[pi] 0.75 -13.9394 bauer, 0.75 -13.9394 parents, 0.75 -13.9394 683 0.75 -13.9394 62.67 0.5 -13.9397 trust, 0.6 -13.94 963 0.6 -13.94 lexically-triggered 0.6 -13.94 “get 0.6 -13.94 and: 0.5 -13.9404 similar way, 0.75 -13.9405 fact: 0.75 -13.9405 kerry 0.75 -13.9405 metonymy, 0.75 -13.9405 87.85 0.75 -13.9405 amber, 0.164557 -13.9405 darpa 0.6 -13.9409 bin, 0.6 -13.9409 plants 0.5 -13.9412 second scenario, 0.5 -13.9412 (isozaki 0.75 -13.9416 eth 0.75 -13.9416 40.1% 0.75 -13.9416 s001 0.75 -13.9416 books’ 0.75 -13.9416 cbs, 0.75 -13.9416 1001 0.6 -13.9418 victory 0.6 -13.9418 sub-word tagging. 0.5 -13.9419 aline 0.0974576 -13.9424 to include 0.75 -13.9427 time-sensitive 0.75 -13.9427 (of, 0.75 -13.9427 573 0.75 -13.9427 itm, 0.75 -13.9427 25.77 0.75 -13.9427 primary subrequest 0.75 -13.9427 3.86 0.75 -13.9427 chih-chung chang and 0.75 -13.9427 0.78 0.78 0.75 -13.9427 co-citations 0.75 -13.9427 [b c] 0.75 -13.9427 billion) 0.75 -13.9427 no split cl. 0.0896057 -13.9429 replaced 0.304348 -13.9431 penn treebank, 0.6 -13.9436 spans; 0.6 -13.9436 arpa 0.6 -13.9436 statue 0.6 -13.9445 microaveraged 0.5 -13.9449 il-2 0.173913 -13.9451 are provided 0.258065 -13.9452 parse features 0.6 -13.9453 affixed 0.352941 -13.9456 stays 0.230769 -13.9457 and j. 0.6 -13.9462 all above 0.6 -13.9462 (“in 0.0514354 -13.9462 the current 0.5 -13.9464 xiong, 0.108696 -13.9466 anonymous reviewers 0.6 -13.9471 lesser” 0.416667 -13.9477 (1991). 0.236842 -13.948 zhu et 0.6 -13.948 resultsof 0.6 -13.948 6b 0.6 -13.948 −= 0.291667 -13.948 first four 0.416667 -13.9489 occur within 0.6 -13.9489 profiles 0.6 -13.9489 guys 0.384615 -13.949 selection step 0.0950413 -13.9492 (which 0.352941 -13.9493 ro 0.6 -13.9498 deu→en 0.6 -13.9498 spectrogram 0.6 -13.9498 zt+1s 0.6 -13.9498 esr1 0.196429 -13.9499 alignment models 0.5 -13.9501 (accuracy 0.5 -13.9501 box. 0.6 -13.9507 (pmi): 0.6 -13.9507 685 0.5 -13.9508 lawrence 0.6 -13.9516 no-pred 0.6 -13.9516 parameterize 0.6 -13.9516 non-text 0.6 -13.9516 convinced 0.047379 -13.9519 in this paper, 0.115854 -13.9519 argmax 0.0306644 -13.9524 in section 0.00409269 -13.9525 of 0.6 -13.9525 argon, 0.6 -13.9525 bottomup 0.6 -13.9525 potassium, 0.6 -13.9525 g(ν,µ) 0.236842 -13.9526 received a 0.171053 -13.9527 findings, 0.5 -13.9531 don 0.6 -13.9534 rerun 0.6 -13.9534 nulli, 0.6 -13.9534 lop 0.384615 -13.9536 can derive 0.6 -13.9543 growing, 0.6 -13.9543 0.785 0.155556 -13.9549 although it 0.6 -13.9552 m.a. 0.384615 -13.9553 class” 0.5 -13.9553 letter substitution 0.416667 -13.9558 eugenio 0.444444 -13.9558 a90a90a26a26 0.444444 -13.9558 fabrizio sebastiani. 0.6 -13.9561 mfp 0.5 -13.9561 93.7 0.352941 -13.9562 rada 0.28 -13.9564 volumes 0.135593 -13.9567 an algorithm 0.416667 -13.957 (2008b) 0.6 -13.957 model-1, 0.5 -13.9576 events involving 0.291667 -13.958 (mcclosky 0.087108 -13.9586 build a 0.6 -13.9588 treatments, 0.6 -13.9588 o(ns) 0.142857 -13.9591 approaches are 0.291667 -13.9593 co-segmentation 0.117647 -13.9597 average number 0.6 -13.9597 shakespeare 0.444444 -13.9598 sir . 0.444444 -13.9598 vincent 0.6 -13.9606 argue, 0.6 -13.9606 dogs, 0.6 -13.9606 (multiple 0.5 -13.9606 clothes 0.444444 -13.9611 sfb 0.416667 -13.9614 deletion rules 0.416667 -13.9614 implicativity 0.6 -13.9615 (dm) 0.6 -13.9615 scholar 0.5 -13.9621 /ada/ 0.6 -13.9624 all-total 0.164557 -13.9629 base. 0.5 -13.9629 0.860 0.6 -13.9633 slovak, 0.6 -13.9633 postulated 0.6 -13.9642 2.49 0.2 -13.9645 creates a 0.352941 -13.9647 (doddington, 0.6 -13.9651 201) 0.6 -13.9651 volatile 0.5 -13.9651 current system, 0.105263 -13.9655 consist 0.315789 -13.9657 restarts 0.6 -13.9661 (hale, 0.6 -13.9661 lexicons: 0.6 -13.9661 in-domain, 0.0669456 -13.9666 the above 0.0672269 -13.9666 we found 0.6 -13.967 counterm 0.6 -13.967 {w 0.6 -13.967 ∆(y, 0.6 -13.967 0.748 0.6 -13.967 54.35 0.6 -13.967 initialize: 0.6 -13.967 ronald 0.169014 -13.9672 context features 0.6 -13.9688 re-evaluated 0.6 -13.9697 haha 0.6 -13.9697 583 0.5 -13.9697 narrative container. 0.5 -13.9697 visually-distinguishable 0.28 -13.9699 feature type 0.5 -13.9704 will probably 0.6 -13.9706 caveats 0.6 -13.9706 (precision 0.6 -13.9706 1836 0.416667 -13.9715 jin-dong 0.6 -13.9724 word-image 0.6 -13.9724 user-specific 0.132231 -13.9733 model, which 0.6 -13.9734 (comparing 0.444444 -13.9739 traveling 0.6 -13.9743 ecocs 0.6 -13.9743 0.039 0.6 -13.9743 (reflecting 0.6 -13.9743 adam’s 0.0539811 -13.9745 are used 0.6 -13.9752 83.53 0.6 -13.9752 465 0.6 -13.9752 indicator’s 0.444444 -13.9753 classifier: 0.129032 -13.9754 each pair 0.5 -13.9758 releasing 0.444444 -13.976 z1 0.6 -13.9761 800, 0.6 -13.9761 no-weight 0.6 -13.9761 (a) baseball-hockey 0.103448 -13.9762 in part 0.291667 -13.9766 alexandre 0.5 -13.9766 fail. 0.444444 -13.9767 1.21 0.444444 -13.9767 conv 0.6 -13.977 enrolled 0.6 -13.977 503 0.0726817 -13.9776 national 0.6 -13.978 roger 0.6 -13.978 60.5% 0.6 -13.978 weight tuning 0.5 -13.9781 gu 0.169014 -13.9783 done on 0.291667 -13.9784 are put 0.258065 -13.9787 features like 0.183333 -13.9788 name. 0.6 -13.9789 10.9% 0.6 -13.9789 andrews 0.6 -13.9789 (bilmes 0.266667 -13.9795 an smt 0.5 -13.9796 τj 0.6 -13.9798 39, 0.6 -13.9798 twa 0.6 -13.9798 (θ) 0.25 -13.9801 corpus). 0.384615 -13.9816 rank candidate 0.6 -13.9816 “prevent 0.6 -13.9816 “topic 0.416667 -13.9818 commentary, 0.333333 -13.9825 denero et 0.6 -13.9826 i;hb;j 0.5 -13.9827 eighth 0.6 -13.9835 “first 0.6 -13.9835 {y 0.444444 -13.9835 eprime∈e 0.5 -13.9835 iy 0.352941 -13.9836 be representative 0.304348 -13.9836 ‘ 0.444444 -13.9842 eng1 0.230769 -13.9842 accomplished 0.6 -13.9844 candidates: 0.6 -13.9844 882 0.6 -13.9844 “people 0.158537 -13.9849 groups. 0.140187 -13.9851 (lin 0.333333 -13.9851 will now 0.14433 -13.9853 w2 0.6 -13.9853 (related 0.6 -13.9853 l ). 0.6 -13.9853 0.439 0.6 -13.9853 sm,n 0.6 -13.9853 (melamed 0.6 -13.9863 principles, 0.6 -13.9863 iker 0.6 -13.9863 += 0.6 -13.9863 glycosylation 0.0923695 -13.9866 are then 0.0513141 -13.9866 results for 0.6 -13.9872 nations, 0.6 -13.9872 josef och 0.160494 -13.9879 evaluation, we 0.6 -13.9881 “v 0.6 -13.9881 abraham 0.6 -13.9891 multigrams 0.6 -13.9891 sortal anaphora 0.242424 -13.9896 unlike previous 0.208333 -13.9898 deviations 0.291667 -13.991 than 100 0.315789 -13.9914 decreases as 0.384615 -13.9923 p(i 0.0815047 -13.9924 to do 0.28 -13.9928 o-en 0.6 -13.9937 (a2) 0.6 -13.9937 identification: 0.5 -13.9943 outperforms state-of-the-art 0.5 -13.9943  0.5 -13.9951 p(post 0.444444 -13.9952 (somasundaran et 0.6 -13.9956 wikipedia1 0.6 -13.9956 aoa 0.6 -13.9956 (imp) 0.5 -13.9967 find optimal 0.5 -13.9967 radicalbigsummationtext 0.357143 -13.997 bm 0.384615 -13.9971 line indicates 0.304348 -13.9973 keystroke 0.6 -13.9975 recalled 0.6 -13.9975 (v,w) 0.126984 -13.9977 have presented 0.269231 -13.9981 blitzer et 0.5 -13.9982 파이 0.6 -13.9984 wherev 0.6 -13.9994 #3 0.230769 -14 this time 0.444444 -14.0001 jagarlamudi 0.28 -14.0002 fund 0.219512 -14.0002 be built 0.6 -14.0003 offices 0.6 -14.0012 setting: 0.6 -14.0012 inhibited 0.6 -14.0012 89.88 0.6 -14.0022 solidarity 0.6 -14.0022 pardo, 0.416667 -14.0025 false. 0.6 -14.0031 steroid 0.6 -14.0031 x0,1 0.6 -14.0031 formulas, 0.352941 -14.0034 [arg1 0.444444 -14.0036 basing 0.6 -14.0041 o(n2 0.444444 -14.0043 q-level 0.123188 -14.0045 al., 2011). 0.5 -14.0045 another aspect 0.357143 -14.0049 ˆl 0.6 -14.005 b&c 0.6 -14.005 prob. 0.6 -14.005 bethard 0.444444 -14.005 string-to-dependency 0.6 -14.006 duchamp 0.6 -14.006 (i,j)∈a 0.6 -14.006 71.26 0.5 -14.0061 xuv 0.315789 -14.0062 (sag 0.315789 -14.0062 aligned words 0.384615 -14.0062 sydney, 0.333333 -14.0064 .5 0.444444 -14.0071 doc-comb 0.5 -14.0077 noprop-dep 0.444444 -14.0078 relationally-clustered 0.6 -14.0079 φc 0.6 -14.0079 92.34 0.6 -14.0079 represent, 0.357143 -14.0083 dda 0.6 -14.0088 imbalance, 0.6 -14.0088 system act, 0.28 -14.0095 philipp 0.196078 -14.0095 an active 0.2 -14.0097 representations for 0.5 -14.0101 jab 0.5 -14.0101 lik 0.416667 -14.0111 svr-based 0.6 -14.0117 defined: 0.6 -14.0126 (pmi), 0.5 -14.0132 two dimensions 0.6 -14.0136 consistency: 0.6 -14.0136 1these 0.1125 -14.0137 is applied 0.25 -14.0141 this table, 0.6 -14.0145 special thanks 0.6 -14.0145 891 0.5 -14.0148 terme 0.384615 -14.0153 una 0.6 -14.0155 macherey 0.6 -14.0155 83.18 0.444444 -14.0155 wsj-only 0.357143 -14.0162 official results 0.6 -14.0164 17.57 0.6 -14.0164 republica 0.6 -14.0174 geo:state 0.6 -14.0174 notr(t) 0.315789 -14.0177 this operation 0.6 -14.0184 settlers 0.6 -14.0184 pseudo-document 0.6 -14.0184 fr–en 0.6 -14.0184 ‡school 0.444444 -14.0184 vb, 0.6 -14.0193 halted 0.6 -14.0193 aditionaly, 0.416667 -14.0197 schuler 0.6 -14.0203 0.0047 0.6 -14.0203 neighbor, 0.444444 -14.0205 403 0.357143 -14.0208 mqa 0.444444 -14.0212 2c 0.5 -14.022 (random 0.5 -14.022 0.856 0.444444 -14.0227 possessives 0.5 -14.0228 juri 0.444444 -14.0234 31.4 0.149425 -14.0235 following: 0.188679 -14.0237 english language 0.6 -14.0241 systemis 0.444444 -14.0248 inducer 0.113208 -14.025 refers to the 0.6 -14.0251 cognition, 0.0524194 -14.026 that can 0.6 -14.0261 off-diagonal 0.5 -14.0268 apps 0.6 -14.027 coined 0.444444 -14.0277  0.444444 -14.0277 class induction 0.6 -14.028 voc 0.6 -14.028 z-score 0.6 -14.028 presenceof 0.333333 -14.0282 informality 0.444444 -14.0284 .001). 0.5 -14.0284 (i,t) 0.333333 -14.0288 restrict our 0.6 -14.029 p(ℓs 0.6 -14.029 (5): 0.6 -14.029 wants, 0.6 -14.029 u(i)(y(i)−1) 0.123077 -14.0293 otherwise. 0.6 -14.0299 (wn,c(wn−11 0.6 -14.0299 richard c. 0.6 -14.0309 mechanisms, 0.6 -14.0309 voiceless 0.093617 -14.0309 i=1 0.5 -14.0309 antonio 0.5 -14.0309 dela 0.154762 -14.031 (callison-burch 0.204082 -14.031 method described 0.3 -14.0314 berlin 0.6 -14.0319 inflection: 0.149425 -14.0326 purpose, 0.357143 -14.0329 [φ 0.291667 -14.0332 she has 0.3 -14.0334 were constructed 0.444444 -14.0334 possessive, 0.6 -14.0338 5for 0.291667 -14.0341 what was 0.5 -14.0341 each group. 0.6 -14.0348 troops 0.5 -14.0349 al-ghad 0.384615 -14.0352 we impose 0.6 -14.0358 dong 0.6 -14.0358 suchas 0.6 -14.0358 shay 0.444444 -14.0363 restr 0.357143 -14.0369 utah 0.6 -14.0377 quiet 0.5 -14.0382 bbc 0.6 -14.0387 mcguffey 0.6 -14.0387 frees 0.6 -14.0387 wdhmm 0.5 -14.039 guy 0.6 -14.0397 metrics, including 0.6 -14.0397 21.90 0.291667 -14.0398 white, 0.444444 -14.0399 larv 0.164384 -14.0402 they used 0.157895 -14.0404 be performed 0.5 -14.0406 justin 0.6 -14.0407 term: 0.128205 -14.0412 is presented 0.5 -14.0414 (vilar 0.6 -14.0416 625 0.28 -14.042 china. 0.209302 -14.0421 evident 0.5 -14.0423 (attali 0.071066 -14.0425 also be 0.6 -14.0426 origtweets 0.6 -14.0426 468 0.6 -14.0426 cder 0.6 -14.0426 “topics” 0.6 -14.0426 656 0.3 -14.0427 draft 0.219512 -14.0435 is commonly 0.444444 -14.0436 islamic 0.3 -14.0451 attribute. 0.384615 -14.0452 =⇒ 0.384615 -14.0452 and joshi, 0.6 -14.0456 animacy, 0.6 -14.0456 n×k 0.174603 -14.0458 the next section 0.6 -14.0466 conceiver 0.6 -14.0466 syntactically, 0.6 -14.0466 joshua’s 0.357143 -14.0475 treating each 0.357143 -14.0475 2006a). 0.6 -14.0475 opus 0.6 -14.0475 ords 0.6 -14.0475 98195 0.315789 -14.0479 al., 2000), 0.235294 -14.048 acld 0.5 -14.048 multiplication. 0.5 -14.048 share information 0.6 -14.0485 s-features 0.6 -14.0495 pain, 0.6 -14.0495 alternately, 0.144444 -14.0504 bleu (papineni et 0.6 -14.0505 english-to-french 0.6 -14.0505 1262 0.6 -14.0515 623 0.6 -14.0515 (highest 0.357143 -14.0516 catvar 0.3 -14.052 articles about 0.384615 -14.0522 default. 0.6 -14.0525 data): 0.5 -14.053 matthias 0.180328 -14.0531 one possible 0.6 -14.0535 §3, 0.0472441 -14.0539 assigned 0.6 -14.0545 ncf,e 0.444444 -14.0545 erjavec, 0.5 -14.0554 (zelenko 0.6 -14.0555 alty 0.5 -14.0563 和 0.6 -14.0565 bitte 0.6 -14.0565 answering system. 0.5 -14.0571 (frequency 0.6 -14.0575 disc 0.269231 -14.0576 data contains 0.6 -14.0585 unique), 0.6 -14.0585 87.17 0.214286 -14.0587 sind 0.5 -14.0588 ˆu 0.333333 -14.0594 manning. 0.6 -14.0595 interfacing 0.6 -14.0595 94.33 0.6 -14.0595 teachers’ 0.357143 -14.0599 80k 0.5 -14.0604 baccianella 0.5 -14.0604 noisy-seed) 0.5 -14.0604 arcleft 0.5 -14.0604 ia/x 0.5 -14.0604 xb 0.5 -14.0604 yang liu, 0.5 -14.0604 euro, 0.5 -14.0604 e∈dn(f) 0.5 -14.0604 hurting 1 -14.0605 (jarvelin 1 -14.0605 markups. 1 -14.0605 ta,i,v 1 -14.0605 adjusted. 1 -14.0605 augmentations 1 -14.0605 senses? 1 -14.0605 masks 1 -14.0605 cycl 1 -14.0605 context-related 1 -14.0605 (n−1) 1 -14.0605 攸 /lkn/but 1 -14.0605 news, wikipedia), 1 -14.0605 ogren 1 -14.0605 51.77 1 -14.0605 korean; 1 -14.0605 gen(x; λ) 1 -14.0605 ,cause 1 -14.0605 batteries. 1 -14.0605 locked-in syndrome 1 -14.0605 “vista” 1 -14.0605 hz. 1 -14.0605 (linguistically) 1 -14.0605 hsmq-learning. 1 -14.0605 longforms 1 -14.0605 186 discourses, 1 -14.0605 postags thepartofspeechtagscorrespondingtothewords 1 -14.0605 lexeme (t,lscript), 1 -14.0605 edition (mendonc¸a 1 -14.0605 (arranz 1 -14.0605 weekly. 1 -14.0605 summ 1 -14.0605 (67.8%), 1 -14.0605 twitchell 1 -14.0605 “bootstrapped” 1 -14.0605 distractors). 1 -14.0605 preservers. 1 -14.0605 adjunct). 1 -14.0605 data(%bleu-cased). 1 -14.0605 honeypots 1 -14.0605 1467 1 -14.0605 crqq tmnz 1 -14.0605 46–55, 1 -14.0605  without 1 -14.0605 99.5 1.005 1 -14.0605 (±1.84) 1 -14.0605 “warplanes” 1 -14.0605 η ·∥w∥2 1 -14.0605 #splits(e) 1 -14.0605 (ζ1,...,ζn) 1 -14.0605 cu. 1 -14.0605 theorem2. 1 -14.0605 feature-type-based 1 -14.0605 doe, j. 1 -14.0605 (hampstead, 1 -14.0605 berkum 1 -14.0605 cludes 1 -14.0605 j∈ji pijcijgn 1 -14.0605 mturkers 1 -14.0605 (bitexts) 1 -14.0605 excitability 1 -14.0605 (summing 1 -14.0605 truncated, 1 -14.0605 (false positives) 1 -14.0605 york? 1 -14.0605 67%, 1 -14.0605 vbz be::v 1 -14.0605 nuit rh´enane 1 -14.0605 devaluing 1 -14.0605 mrf (pl-mrf) 1 -14.0605 peter. 1 -14.0605 a2:donate 1 -14.0605 orthant-wise limited-memory 1 -14.0605 nerima 1 -14.0605 internet-enabled 1 -14.0605 pnull 1 -14.0605 smoking. 1 -14.0605 keh-yih 1 -14.0605 everybody, 1 -14.0605 overly-specific 1 -14.0605 append tnew 1 -14.0605 noticed, 1 -14.0605 side; 1 -14.0605 hypothesized, 1 -14.0605 been, 1 -14.0605 recursing. 1 -14.0605 stts-emg 1 -14.0605 non-coreferent 1 -14.0605 (compliment, friends) 1 -14.0605 .702 1 -14.0605 27(4):521–544. 1 -14.0605 disputed authorship: 1 -14.0605 alsoincludingthewordsinthe 1 -14.0605 spinner.” 1 -14.0605 fst-based 1 -14.0605 qc. 1 -14.0605 m´ario 1 -14.0605 clang, 1 -14.0605 reannotate 1 -14.0605 subtitling 1 -14.0605 {prices, markets, 1 -14.0605 nonobvious 1 -14.0605 (w1,w2,n,l,m,r). 1 -14.0605 maxpaths. 1 -14.0605 datatext 1 -14.0605 demonstrativenpfeaturestrueifthegivennoun 1 -14.0605 bos >(n\n) 1 -14.0605 kanji. 1 -14.0605 6a/b 1 -14.0605 1/k. 1 -14.0605 watson: 1 -14.0605 bucket) ‘bucket 1 -14.0605 hassan sajjad 1 -14.0605 development/testing 1 -14.0605 englishtablecaption 1 -14.0605 herein. 1 -14.0605 $1275 1 -14.0605 (unigram) 1 -14.0605 cars. 1 -14.0605 “sxsw”. 1 -14.0605 discourse-new. 1 -14.0605 5s. 1 -14.0605 “nice 1 -14.0605 olympus 1 -14.0605 83.61 1 -14.0605 (spe) 1 -14.0605 friend. 1 -14.0605 categories2. 1 -14.0605 solicitation 1 -14.0605 twitter; 1 -14.0605 object-pairs 1 -14.0605 .05,†at 1 -14.0605 (6.05b 1 -14.0605 directionality: 1 -14.0605 model2. 1 -14.0605 lefthand-side 1 -14.0605 (13.2b). 1 -14.0605 marianna 1 -14.0605 standard’s 1 -14.0605 90.3% 1 -14.0605 scripting 1 -14.0605 generalizability. 1 -14.0605 date; 1 -14.0605 postprocessing, 1 -14.0605 project). 1 -14.0605 22.67 25.32 1 -14.0605 sdbw 1 -14.0605 icomputation 1 -14.0605 (lch): 1 -14.0605 mark_() 1 -14.0605 (=hiero) 1 -14.0605 radiographic 1 -14.0605 finden 1 -14.0605 79.67 1 -14.0605 p(f,e) 1 -14.0605 http://aclweb.org/aclwiki/ 1 -14.0605 52.85% 1 -14.0605 baseline), 1 -14.0605 conversing 1 -14.0605 travel. 1 -14.0605 [−1,1]. 1 -14.0605 hybridization 1 -14.0605 idf) 1 -14.0605 [4..6] 1 -14.0605 ptsg/pcfg 1 -14.0605 4.6%. 1 -14.0605 puppy 1 -14.0605 52.9% 1 -14.0605 knowledge). 1 -14.0605 (s-pos), 1 -14.0605 web1t, 1 -14.0605 𝑚 1 -14.0605 noun/adjective 1 -14.0605 84–90, 1 -14.0605 collaboration, 1 -14.0605 unmarked. 1 -14.0605 interruptions. 1 -14.0605 agencies, 1 -14.0605 31.16 1 -14.0605 fr syntacticanotations fr semanticanotations 1 -14.0605 (fa)+gw). 1 -14.0605 reference-aware 1 -14.0605 scl. 1 -14.0605 .005 1 -14.0605 giza-weak). 1 -14.0605 〈1,v=>+ept〉(22)) 1 -14.0605 cube-pruning, 1 -14.0605 stolcke a., 1 -14.0605 cancombine(null, ˜null): 1 -14.0605 links), 1 -14.0605 koehn. 1 -14.0605 9464 55.98 1 -14.0605 mastered 1 -14.0605 (32%). 1 -14.0605 juice. 1 -14.0605 2details 1 -14.0605 metavariables, 1 -14.0605 parts; 1 -14.0605 cross-validation: 1 -14.0605 0.98, 1 -14.0605 viscus, 1 -14.0605 pp*vp 1 -14.0605 ’and’ 1 -14.0605 repackaged 1 -14.0605 (seretan, 1 -14.0605 3,00 1 -14.0605 x(su) 1 -14.0605 (nlu). 1 -14.0605 (tm), 1 -14.0605 bag-of-concepts. 1 -14.0605 #bpoil #bp 1 -14.0605 shrestha 1 -14.0605 exercises: 1 -14.0605 tially 1 -14.0605 avg.f-score 1 -14.0605 -0.411 1 -14.0605 aamir sohail 1 -14.0605 emotionally. 1 -14.0605 weuse 1 -14.0605 ss) 1 -14.0605 .us 1 -14.0605 2011 conllsharedtask 1 -14.0605 (pilz 1 -14.0605 (ave.) 1 -14.0605 64.91 1 -14.0605 10-20 bow 1 -14.0605 f1,f8 1 -14.0605 schema_cluster_match 1 -14.0605 available2 1 -14.0605 elkiss 1 -14.0605 enhancements. 1 -14.0605 integration: 1 -14.0605 stage) 1 -14.0605 (cap), 1 -14.0605 dificulty. 1 -14.0605 corpus.3 1 -14.0605 manually-compiled 1 -14.0605 days). 1 -14.0605 gender-location 1 -14.0605 experimentand 1 -14.0605 mmpresent.:7 1 -14.0605 hysom. 1 -14.0605 fern’andez, 1 -14.0605 1122 1 -14.0605 yoav artzi 1 -14.0605 numberof systemsin 1 -14.0605 reltype 1 -14.0605 minimum-cut. 1 -14.0605 congressional floor-debate 1 -14.0605 rebholzschuhmann 1 -14.0605 wood. 1 -14.0605 eun young 1 -14.0605 microphone, 1 -14.0605 thata 1,...,an 1 -14.0605 (rouge-n) 1 -14.0605 btec, 1 -14.0605 dmean 1 -14.0605 conf set(wi) 1 -14.0605 desc ?n 1 -14.0605 show.v 1 -14.0605 brichst brechst 1 -14.0605 substructures, 1 -14.0605 9,263 1 -14.0605 a9 1 -14.0605 artist? 1 -14.0605 nonspecialized 1 -14.0605 nugues. 1 -14.0605 3rd acm/iee 1 -14.0605 (1.0), 1 -14.0605 position, responsibility, 1 -14.0605 jp, fr, 1 -14.0605 0.967. 1 -14.0605 (isa) 1 -14.0605 r5 r6 1 -14.0605 /b/-/d/ 1 -14.0605 ∑h∈h1 pih 1 -14.0605 eric nyberg 1 -14.0605 hs03’s 1 -14.0605 kop, 1 -14.0605 august ’99 1 -14.0605 o’donovan 1 -14.0605 deceasing 1 -14.0605 noise; 1 -14.0605 nextpp= builtpps.nextpp(bestarc) 1 -14.0605 skipped, 1 -14.0605 (parametric) 1 -14.0605 •preprocessing. 1 -14.0605 destiny. 1 -14.0605 morristown, 1 -14.0605 fl øk 1 -14.0605 nadir durrani 1 -14.0605 −−→tverb 1 -14.0605 thementions.andthen 1 -14.0605 eyssel, 1 -14.0605 vs. etwa). 1 -14.0605 resource-intensive. 1 -14.0605 assistant. 1 -14.0605 amusing 1 -14.0605 “/ f(jiqiao)/skill” 1 -14.0605 car.’ 1 -14.0605 {(xi,zi) 1 -14.0605 (lrec'10). 1 -14.0605 nucleic acids, 1 -14.0605 woz07 1 -14.0605 analogous. 1 -14.0605 determinant, 1 -14.0605 unbroken 1 -14.0605 enlightening 1 -14.0605 (gonz´alezbrenes 1 -14.0605 0.76 0.82 1 -14.0605 2/2=1 2/3=0.66 1 -14.0605 -ire. 1 -14.0605 pls. 1 -14.0605 discriminating? 1 -14.0605 viscus”. 1 -14.0605 (easley 1 -14.0605 0.70. 1 -14.0605 j.lang-3@sms.ed.ac.uk, 1 -14.0605 metaphorically. 1 -14.0605 threading. 1 -14.0605 simplistic. 1 -14.0605 subjects participated 1 -14.0605 poibeau lattice, 1 -14.0605 9since 1 -14.0605 hague, 1 -14.0605 0.83. 1 -14.0605 (chunk) 1 -14.0605 functor. 1 -14.0605 07-09, 1 -14.0605 76.71 52.83 1 -14.0605 alive. 1 -14.0605 woepos. 1 -14.0605 12), 1 -14.0605 spearman's 1 -14.0605 binarized, 1 -14.0605 (reduced). 1 -14.0605 relativebeamwidth 1 -14.0605 -.89 1 -14.0605 kl-divergence. 1 -14.0605 summed. 1 -14.0605 (+1.12) 1 -14.0605 martin. 1 -14.0605 acid esters 1 -14.0605 child-like 1 -14.0605 ttable 1 -14.0605 glosser, 1 -14.0605 34.01 32.23 1 -14.0605 96.73 1 -14.0605 sweep p 1 -14.0605 nuggets) 1 -14.0605 propositions,partial 1 -14.0605 (patient) 1 -14.0605 䉂 奒 1 -14.0605 “grammatical relation⇒ne 1 -14.0605 (puerta melguizo 1 -14.0605 (cui 1 -14.0605 sub-corpora. 1 -14.0605 cedure 1 -14.0605 69.84% 1 -14.0605 co-segmented 1 -14.0605 symbol’s 1 -14.0605 shell. 1 -14.0605 18.7% 1 -14.0605 complexes. 1 -14.0605 vestigate 1 -14.0605 graph-clustering 1 -14.0605 µi. 1 -14.0605 understandings 1 -14.0605 avignon, france 1 -14.0605 msds 1 -14.0605 budanitsky, 1 -14.0605 n-foil. 1 -14.0605 advertising. 1 -14.0605 we) 1 -14.0605 raghavu@microsoft.com 1 -14.0605 strategically 1 -14.0605 sapporo, 1 -14.0605 shey@cs.biu.ac.il 1 -14.0605 +0.05 1 -14.0605 simulated annealing, 1 -14.0605 nets. 1 -14.0605 melody. 1 -14.0605 (bad, 1 -14.0605 ilps. 1 -14.0605 professionally-produced 1 -14.0605 yielding none, 1 -14.0605 reconstituted 1 -14.0605 lexicalised, 1 -14.0605 negotiated 1 -14.0605 (baseline): 1 -14.0605 2010.) 1 -14.0605 outperform. 1 -14.0605 do? 1 -14.0605 b.plank@rug.nl 1 -14.0605 infancy. 1 -14.0605 f¨ur maschinelle 1 -14.0605 e3, 1 -14.0605 25000 1 -14.0605 ays 1 -14.0605 frameworks: 1 -14.0605 142–151, 1 -14.0605 czech-english. 1 -14.0605 textbook, 1 -14.0605 childes: 1 -14.0605 lifting 1 -14.0605 (senses 1 -14.0605 gobi 1 -14.0605 jegelka 1 -14.0605 wuvyuv 1 -14.0605 (+1.4) 1 -14.0605 0,m 1 -14.0605 engendered 1 -14.0605 treebank-derived 1 -14.0605 subordination, 1 -14.0605 (backward 1 -14.0605 palestine liberation 1 -14.0605 lemma) 1 -14.0605 breastcancer.org 1 -14.0605 individual) 1 -14.0605 class-seeking 1 -14.0605 sailing vessel 1 -14.0605 millisecond 1 -14.0605 p5’s 1 -14.0605 11–21, 1 -14.0605 “munged” 1 -14.0605 kozareva, 1 -14.0605 coreferencedatasetace 1 -14.0605 positionagreement 1 -14.0605 presentable 1 -14.0605 mini-tacitus. 1 -14.0605 guy lapalme, 1 -14.0605 ,.all, 1 -14.0605 creole. 1 -14.0605 2.07% 1 -14.0605 (per, 1 -14.0605 ireland (sfi), 1 -14.0605 charge. 1 -14.0605 71.69 1 -14.0605 (seq/desired symbol)mean 1 -14.0605 38.78 1 -14.0605 references; 1 -14.0605 spammer 1 -14.0605 yuw ∈{0,1}. 1 -14.0605 dt: 1 -14.0605 jargons, 1 -14.0605 dt), 1 -14.0605 paltoglou 1 -14.0605 intestine 1 -14.0605 manipulated. 1 -14.0605 36.74 1 -14.0605 ‘o’. 1 -14.0605 5.3.1). 1 -14.0605 {cq1,...,cqk}. 1 -14.0605 classroom. 1 -14.0605 [p spicy 1 -14.0605 1992; jurafsky 1 -14.0605 (hri). 1 -14.0605 figures-of-speech 1 -14.0605 (fig.1). 1 -14.0605 58.93 1 -14.0605 mentiondetectionmethod. 1 -14.0605 unoa ffi 1 -14.0605 (futagi 1 -14.0605 difficultand challenging 1 -14.0605 dictionary-like 1 -14.0605 laughter 1 -14.0605 1650-1800. 1 -14.0605 hwc 1 -14.0605 69.97 1 -14.0605 pr´evot 1 -14.0605 heuristically. 1 -14.0605 !w 1 -14.0605 “x(topic marker)” 1 -14.0605 starred. 1 -14.0605 data-sets, 1 -14.0605 (.96) 1 -14.0605 conference). 1 -14.0605 semi-lexical 1 -14.0605 peer, 1 -14.0605 kprime 1 -14.0605 thurlow, crispin 1 -14.0605 fr–en. 1 -14.0605 word-final 1 -14.0605 generators) 1 -14.0605 tony (wenda) 1 -14.0605 preprocess, 1 -14.0605 “mental age” 1 -14.0605 §2. 1 -14.0605 tures. 1 -14.0605 idt 1 -14.0605 -1.64. 1 -14.0605 sentence? 1 -14.0605 a80a80a80data 1 -14.0605 colaboration. 1 -14.0605 termhood-based 1 -14.0605 xii. 1 -14.0605 p(c) ={i∈u(c) 1 -14.0605 coindexations. 1 -14.0605 g2p. 1 -14.0605 bracketing. 1 -14.0605 z∗(qry 1 -14.0605 moses-gc. 1 -14.0605 kambhatla, 1 -14.0605 friends). 1 -14.0605 (carenini 1 -14.0605 label.7 1 -14.0605 anping-fort. 1 -14.0605 ag+ dp+ 1 -14.0605 wordnet’s, 1 -14.0605 summationtext p,r 1 -14.0605 “binary+patterns”, “coarse+patterns”, 1 -14.0605 complq; expl 1 -14.0605 (alegria 1 -14.0605 sedivy 1 -14.0605 there? 1 -14.0605 over-representation 1 -14.0605 similarity/distance 1 -14.0605 preterminalized litgs 1 -14.0605 0134, 1 -14.0605 aux, did] 1 -14.0605 domains (headphones, 1 -14.0605 mckp, 1 -14.0605 divisions. 1 -14.0605 nonrebuttals 1 -14.0605 language-processing 1 -14.0605 reported; 1 -14.0605 29(1). 1 -14.0605 (franz 1 -14.0605 token#base form#tag#> 1 -14.0605 scope). 1 -14.0605 multi-clause 1 -14.0605 morocco. 1 -14.0605 evidence) 1 -14.0605 1,511 1 -14.0605 -, –} 1 -14.0605 correction: 1 -14.0605 (hsieh 1 -14.0605 verses. 1 -14.0605 need: 1 -14.0605 {hartung,frank}@cl.uni-heidelberg.de 1 -14.0605 reduction), 1 -14.0605 erasing 1 -14.0605 decoding). 1 -14.0605 d-66123 saarbr¨ucken 1 -14.0605 (wh)iµsummationtext vhav 1 -14.0605 norwegian metacenter 1 -14.0605 {kgimpel,nasmith}@cs.cmu.edu 1 -14.0605 2009-01-27: 1 -14.0605 (lmde) 1 -14.0605 ieer. 1 -14.0605 grounded, 1 -14.0605 mada. 1 -14.0605 word.3 1 -14.0605 orientations) 1 -14.0605 nicely. 1 -14.0605 wible. 1 -14.0605 thefirst 1 -14.0605 atool 1 -14.0605 bit-strings 1 -14.0605 iwashita, 1 -14.0605 c(xi, ˆy,yi) 1 -14.0605 macr: 1 -14.0605 declaratively-specified 1 -14.0605 translationese, 1 -14.0605 (iv)). 1 -14.0605 violation. 1 -14.0605 (1650-1700, 1 -14.0605 (1t 1 -14.0605 reversible. 1 -14.0605 (tam, 1 -14.0605 6639 1 -14.0605 (1% 1 -14.0605 “both fine” 1 -14.0605 bless. 1 -14.0605 dale2 1 -14.0605 (74.8% 1 -14.0605 '01 1 -14.0605 j∈in(i) s(j)× 1 -14.0605 nonrelational 1 -14.0605 developing, 1 -14.0605 cooperating 1 -14.0605 documenting 1 -14.0605 (gim´enez, 1 -14.0605 wikis, 1 -14.0605 that-less ccs 1 -14.0605 document/data 1 -14.0605 potential) 1 -14.0605 łyk. 1 -14.0605 decorating 1 -14.0605 vectorθ, 1 -14.0605 (kilany 1 -14.0605 bean 1 -14.0605 jason. 1 -14.0605 vilain 1 -14.0605 210. 1 -14.0605 differences) 1 -14.0605 larry.hunter@ucdenver.edu 1 -14.0605 (liang, 1 -14.0605 0.77, 1 -14.0605 stringmatchingfeaturestrue 1 -14.0605 christiane. 1 -14.0605 met] [am-prp 1 -14.0605 itspoke: 1 -14.0605 tfln, fml, 1 -14.0605 (jiao 1 -14.0605 8way 1 -14.0605 sarah alkuhlani 1 -14.0605 (f0) 1 -14.0605 overtaken 1 -14.0605 0.1603 0.6095 1 -14.0605 (false). 1 -14.0605 preserv. 1 -14.0605 budget, 1 -14.0605 describe: 1 -14.0605 dubious 1 -14.0605 pre-tutoring 1 -14.0605 de∇3,∇3 1 -14.0605 borow 1 -14.0605 adults, 1 -14.0605 banea, carmen, 1 -14.0605 cu-marecek. 1 -14.0605 searchbased 1 -14.0605 personnes, 1 -14.0605 windowsizeof2. 1 -14.0605 bavelas 1 -14.0605 reconsidering 1 -14.0605 unsatisfying 1 -14.0605 maoxi li 1 -14.0605 “tomorrow 1 -14.0605 (gfs). 1 -14.0605 94.59 88.80 1 -14.0605 document%1:06:00 1 -14.0605 “b”, “c”, 1 -14.0605 (e. 1 -14.0605 replace. 1 -14.0605 treebank) 1 -14.0605 pers.coll 1 -14.0605 2.3.4 1 -14.0605 institute. 1 -14.0605 (kiyoki 1 -14.0605 objets trouvés. 1 -14.0605 c-lda-a, 1 -14.0605 trade-off. 1 -14.0605 rochester. 1 -14.0605 apostrophes. 1 -14.0605 non-intrusive 1 -14.0605 authorship, 1 -14.0605 files). 1 -14.0605 ,...,λ(d)md} 1 -14.0605 pedersen. 1 -14.0605 clustering). 1 -14.0605 grammy awards, 1 -14.0605 vdss. 1 -14.0605 promoters. 1 -14.0605 classification/regression 1 -14.0605 〈e1e2,f1f2〉. 1 -14.0605 s(ǫ)) 1 -14.0605 sentenceswithin 1 -14.0605 “door”. 1 -14.0605 domainspecific, 1 -14.0605 track). 1 -14.0605 honeymooners. 1 -14.0605 lete lesco 1 -14.0605 feature-hmm: 1 -14.0605 italicized. 1 -14.0605 engagement. 1 -14.0605 fold: 1 -14.0605 (italian 1 -14.0605 conflict. 1 -14.0605 locally t-threshold 1 -14.0605 (shaikh 1 -14.0605 aa∗(b+ 1 -14.0605 #ofclasses 1 -14.0605 bharati 1 -14.0605 coling 2008. 1 -14.0605 summaries). 1 -14.0605 j. (196) 1 -14.0605 analyzer/generator 1 -14.0605 mst, 1 -14.0605 words.) 1 -14.0605 sproat. 1 -14.0605 footing. 1 -14.0605 ∗corresponding 1 -14.0605 1411 1 -14.0605 englishgermanevaluation. 1 -14.0605 spoustova 1 -14.0605 captions. 1 -14.0605 vital. 1 -14.0605 ‘eric clapton’ 1 -14.0605 inbuilt inscript 1 -14.0605 (schegloff 1 -14.0605 turntaking. 1 -14.0605 x,y ϵ 1 -14.0605 λi,t ←λi,t−αt 1 -14.0605 staggering 1 -14.0605 ai-tr-483, 1 -14.0605 features+statistics-based 1 -14.0605 “joy”. 1 -14.0605 segment’s 1 -14.0605 learned extraction-only 1 -14.0605 propbank) 1 -14.0605 attachable 1 -14.0605 (koolen 1 -14.0605 e→f 1 -14.0605 shares. 1 -14.0605 sequencec 1 -14.0605 br :={at(r,n1),ik(o,at(r,n1)), 1 -14.0605 (dr), 1 -14.0605 stacked, 1 -14.0605 (sammons 1 -14.0605 steps) 1 -14.0605 circumvents 1 -14.0605 (fsa) 1 -14.0605 contributors, 1 -14.0605 (setting 1 -14.0605 (-1,1) 1 -14.0605 seddah, 1 -14.0605 ai’s 1 -14.0605 n-fold 1 -14.0605 wished. 1 -14.0605 sampling) 1 -14.0605 clauses; 1 -14.0605 3200 1 -14.0605 projectable, 1 -14.0605 semi-automate 1 -14.0605 0.51.52.53.54.55.56.57.58.59.5 1 -14.0605 sprimev∈nmf 1 -14.0605 representations; 1 -14.0605 discouraged. 1 -14.0605 40.99 1 -14.0605 banter. 1 -14.0605 85.78 1 -14.0605 (prendinger 1 -14.0605 equivalent). 1 -14.0605 papers: 1 -14.0605 historically, submodular 1 -14.0605 say “congratulations 1 -14.0605 category/domain 1 -14.0605 nativespeaker 1 -14.0605 0.1481 1 -14.0605 rte-3 1 -14.0605 padt. 1 -14.0605 insertions. 1 -14.0605 punjabi, 1 -14.0605 depletes 1 -14.0605 args, 1 -14.0605 precisionand 1 -14.0605 morpheus (crane, 1 -14.0605 flanking 1 -14.0605 y2. 1 -14.0605 forgot farrah 1 -14.0605 ’? 1 -14.0605 knowledgeable 1 -14.0605 figures: 1 -14.0605 kilgarriff’s 1 -14.0605 “collaborative” 1 -14.0605 238m 1 -14.0605 i=ld(1..8) 1 -14.0605 package? 1 -14.0605 1964) 1 -14.0605 bel(b, wildo(a, 1 -14.0605 tiedemann, 1 -14.0605 booth 1 -14.0605 ϵk. 1 -14.0605 multiplication). 1 -14.0605 “ad”. 1 -14.0605 ‘informative’ 1 -14.0605 algorithm? 1 -14.0605 treebank 1x 1 -14.0605 tool), 1 -14.0605 en-dir. 1 -14.0605 geo, 1 -14.0605 unexploited. 1 -14.0605 schuler@ling.ohio-state.edu 1 -14.0605 sensor kinase 1 -14.0605 96.7%. 1 -14.0605 meat). 1 -14.0605 mek-ess-ta. 1 -14.0605 lemas 1 -14.0605 biology) 1 -14.0605 ation. 1 -14.0605 una iniciativa 1 -14.0605 1892. 1 -14.0605 tempeval-2010. 1 -14.0605 “size.” 1 -14.0605 (itai 1 -14.0605 (ob 1 -14.0605 arg1-arg1, arg0-arg1, 1 -14.0605 parser sec/sent 1 -14.0605 “must-link” 1 -14.0605 180.67 1 -14.0605 overs 1 -14.0605 ia/x (etprime..t/fu..uprime 1 -14.0605 2.32% 1 -14.0605 annual amia 1 -14.0605 commissiona 1 -14.0605 on-demand/on-the-fly 1 -14.0605 sadness—go 1 -14.0605 massachusetts: 1 -14.0605 delineations 1 -14.0605 bmt, 1 -14.0605 54.93 1 -14.0605 phrasesand 1 -14.0605 phillips. 1 -14.0605 disorders. 1 -14.0605 val`encia 46022 1 -14.0605 marie anne 1 -14.0605 crsi22 127510 1 -14.0605 voc(ee). 1 -14.0605 bothside 1 -14.0605 low? 1 -14.0605 (-1.3, 1 -14.0605 space.1 1 -14.0605 λiφi(t,j,s), 1 -14.0605 ˜dlscript. 1 -14.0605 books). 1 -14.0605 back-off mixing, 1 -14.0605 weps 1 -14.0605 rate” 1 -14.0605 isabel trancoso 1 -14.0605 bren”) 1 -14.0605 branching. 1 -14.0605 liblinear l2-l1 1 -14.0605 schedule (ados) 1 -14.0605 “will medtronic’s 1 -14.0605 .049 1 -14.0605 priming. 1 -14.0605 retraining. 1 -14.0605 nonconvex, 1 -14.0605 again). 1 -14.0605 convergence). 1 -14.0605 hyphenation. 1 -14.0605 directives. 1 -14.0605 comparatives. 1 -14.0605 partitioned. 1 -14.0605 cooperation. 1 -14.0605 non-o 1 -14.0605 experiments’ 1 -14.0605 regulation) 1 -14.0605 pubmed/medline 1 -14.0605 /ws. 1 -14.0605 scription 1 -14.0605 limsi’s 1 -14.0605 oes. 1 -14.0605 demos. 1 -14.0605 taggings 1 -14.0605 wearable 1 -14.0605 scattering 1 -14.0605 broader/narrower 1 -14.0605 ‘fish’ 1 -14.0605 v(s). 1 -14.0605 avail 1 -14.0605 [v1w1,v2w2], 1 -14.0605 wikulu. 1 -14.0605 45.00 1 -14.0605 actuallya 1 -14.0605 #3. 1 -14.0605 traian basescu 1 -14.0605 −→n 1 -14.0605 character-level. 1 -14.0605 0.46. 1 -14.0605 ferreira leite 1 -14.0605 s0 w4”,x)] 1 -14.0605 discount. 1 -14.0605 κ), 1 -14.0605 question? 1 -14.0605 (kintsch, 1 -14.0605 #3, 1 -14.0605 (muc-4 1 -14.0605 summationtextj mjpj 1 -14.0605 answer; 1 -14.0605 as). 1 -14.0605 dso 1 -14.0605 octuplets 1 -14.0605 chicago. 1 -14.0605 powerbook 1 -14.0605 (possibly null) 1 -14.0605 hyperedges) 1 -14.0605 ‘normalised’ 1 -14.0605 unit). 1 -14.0605 2005, ai 1 -14.0605 verhagen. 1 -14.0605 efflux.n, 1 -14.0605 parity 1 -14.0605 face”. 1 -14.0605 a1a2 1 -14.0605 unforeseen 1 -14.0605 bruce springsteen 1 -14.0605 insegnando [s/he 1 -14.0605 scores.) 1 -14.0605 ˆpintp. 1 -14.0605 preformed 1 -14.0605 t0 dct 1 -14.0605 “long tail” 1 -14.0605 1ticc 1 -14.0605 cases.) 1 -14.0605 a1s 1 -14.0605 (downstream) 1 -14.0605 parser1, 1 -14.0605 mcollins@cs.columbia.edu 1 -14.0605 ˆx2. 1 -14.0605 (mutton 1 -14.0605 (modulo 1 -14.0605 ratings (r(54003) 1 -14.0605 %bleu. 1 -14.0605 ∈x.x(s) 1 -14.0605 -iren 1 -14.0605 hal.’ 1 -14.0605 low-memory 1 -14.0605 sketch: 1 -14.0605 thann 1 -14.0605 91.9% 1 -14.0605 ageno. 1 -14.0605 hispanics 1 -14.0605 baayen 1 -14.0605 2nd amendment, 1 -14.0605 phrasesor 1 -14.0605 disease). 1 -14.0605 analyzer, adam, 1 -14.0605 council. 1 -14.0605 curators. 1 -14.0605 research; 1 -14.0605 “hub-removed” 1 -14.0605 monosemous. 1 -14.0605 elman 1 -14.0605 contexts). 1 -14.0605 featuretypes 1 -14.0605 string-to-string, 1 -14.0605 t/v. 1 -14.0605 {lzy.thu, 1 -14.0605 combiner, 1 -14.0605 24.47 63.26 1 -14.0605 +0.73 1 -14.0605 articulating 1 -14.0605 dld: 1 -14.0605 centroid, 1 -14.0605 negative/positive 1 -14.0605 d’avignon (uapv), 1 -14.0605 28. 1 -14.0605 c=0 1 -14.0605 cc, but] 1 -14.0605 attributes). 1 -14.0605 proposition-level 1 -14.0605 sub-parts. 1 -14.0605 unlimited). 1 -14.0605 41.62% 1 -14.0605 apw 1 -14.0605 91.77 85.74 1 -14.0605 .359 1 -14.0605 (mrl). 1 -14.0605 price. 1 -14.0605 defended 1 -14.0605 taxon. 1 -14.0605 mitton, 1 -14.0605 mult(µ). 1 -14.0605 centroids. 1 -14.0605 1957), 1 -14.0605 matrixes 1 -14.0605 exp(β1x1 +···+βkxk) 1 -14.0605 co-occurs. 1 -14.0605 consulted. 1 -14.0605 non-concave 1 -14.0605 trigger/role 1 -14.0605 detail). 1 -14.0605 unused 1 -14.0605 purpose). 1 -14.0605 #2, 1 -14.0605 conveniently. 1 -14.0605 nyt95/nyt95prime 1 -14.0605 (rst). 1 -14.0605 worked. 1 -14.0605 algorithms’ 1 -14.0605 (part-of-speech, 1 -14.0605 simplify. 1 -14.0605 non-capitalized, 1 -14.0605 75013 1 -14.0605 lab ltrc 1 -14.0605 2(2). 1 -14.0605 ”disc” 1 -14.0605 vetting 1 -14.0605 (asymmetric) 1 -14.0605 blunt 1 -14.0605 parsers? 1 -14.0605 edge-flip. 1 -14.0605 cij, inegationslash=j 1 -14.0605 (13.5%) 1 -14.0605 germane. 1 -14.0605 vicinity 1 -14.0605 interface; 1 -14.0605 burkholderia xenovorans 1 -14.0605 wordinternal 1 -14.0605 reran 1 -14.0605 md: 1 -14.0605 implicatures. 1 -14.0605 andrei.popescu-belis@idiap.ch 1 -14.0605 character-, 1 -14.0605 flight crew 1 -14.0605 500+ 1 -14.0605 non-explicit 1 -14.0605 test-sets. 1 -14.0605 30.86 1 -14.0605 hypergraph), 1 -14.0605 feats. 1 -14.0605 malaria mosquito 1 -14.0605 interpret: 1 -14.0605 ignored). 1 -14.0605 machine translation). 1 -14.0605 11b, 1 -14.0605 consortium (ldc): 1 -14.0605 head-to-head comparisons, 1 -14.0605 automatically). 1 -14.0605 σ2. 1 -14.0605 molecules. 1 -14.0605 constraints” 1 -14.0605 mending, 1 -14.0605 subproblems. 1 -14.0605 epstein. 1 -14.0605 bittar, 1 -14.0605 confidenceweighted 1 -14.0605 neal, 1 -14.0605 craven, 1 -14.0605 janyce. 1 -14.0605 schumaker 1 -14.0605 䠮 ‘iraq’ 1 -14.0605 λpz. 1 -14.0605 conjunctions). 1 -14.0605 linkages (“who 1 -14.0605 α3. 1 -14.0605 multi-aspect 1 -14.0605 s1i (s1i 1 -14.0605 (bleu). 1 -14.0605 ]); 1 -14.0605 tainted 1 -14.0605 compatibilities 1 -14.0605 unpluged. 1 -14.0605 epsilon1]. 1 -14.0605 scc, 1 -14.0605 ;a/, 1 -14.0605 al-masdar. 1 -14.0605 é) 1 -14.0605 np/pp 1 -14.0605 viethen1,2 1 -14.0605 (tomai 1 -14.0605 inu. 1 -14.0605 1c=0.0,epsilon1=0.0010 1 -14.0605 “prenormalization” 1 -14.0605 systematic. 1 -14.0605 dtd creation 1 -14.0605 30.84 1 -14.0605 modal/logical 1 -14.0605 (prof. 1 -14.0605 ∃j. 1 -14.0605 chai. 1 -14.0605 expresive 1 -14.0605 renter 1 -14.0605 learned) 1 -14.0605 55.9% 1 -14.0605 historians. 1 -14.0605 “npov dispute”. 1 -14.0605 foresaw 1 -14.0605 otherwise) 1 -14.0605 (p.c.) 1 -14.0605 t”. 1 -14.0605 (0.9%) 1 -14.0605 timexes 1 -14.0605 ptm. 1 -14.0605 fλx.cost(x)∧f(x) 1 -14.0605 3,021 1 -14.0605 telic role), 1 -14.0605 non-dependencies 1 -14.0605 libsvm: 1 -14.0605 ‘modifier’ 1 -14.0605 10-3 1 -14.0605 scalar. 1 -14.0605 p(t), 1 -14.0605 vb. 1 -14.0605 .txt 1 -14.0605 chisquare 1 -14.0605 onlyidentitycoreferenceisconsideredwhileattributive 1 -14.0605 sprouted 1 -14.0605 (corrects 1 -14.0605 ace2003. 1 -14.0605 semanticaly 1 -14.0605 71.96% 1 -14.0605 tutor: [does 1 -14.0605 too! 1 -14.0605 speech-like 1 -14.0605 61.96 1 -14.0605 fine-grain 1 -14.0605 overrides 1 -14.0605 ≻z,(φ′∧x)= z 1 -14.0605 astronomically 1 -14.0605 (constructing 1 -14.0605 ref automotive, 1 -14.0605 ratnaparkhi’s 1 -14.0605 =mt(rl1..nl) 1 -14.0605 metaphorical. 1 -14.0605 essence/stufflike 1 -14.0605 (experiment-iv) 1 -14.0605 4k. 1 -14.0605 “cohesion” 1 -14.0605 julia@cs.columbia.edu 1 -14.0605 .020 1 -14.0605 “gelesen” 1 -14.0605 chapters, 1 -14.0605 57.6k 1 -14.0605 fp(rf(s1),q1) 1 -14.0605 (iarpa). 1 -14.0605 disambiguators 1 -14.0605 phraseis 1 -14.0605 hedged 1 -14.0605 1.1). 1 -14.0605 “equal 1 -14.0605 pluralizing 1 -14.0605 capillary 1 -14.0605 delete, replace, 1 -14.0605 contrasts, 1 -14.0605 discussion). 1 -14.0605 82.2% 65.0% 1 -14.0605 ngs-u’s 1 -14.0605 springen sprengen 1 -14.0605 available); 1 -14.0605 i=2 i=10 1 -14.0605 previouswork 1 -14.0605 mean-field factorization: 1 -14.0605 (dreyer 1 -14.0605 –coref’s 1 -14.0605 1 -14.0605 inversion. 1 -14.0605 crossvalidation) 1 -14.0605 streamlining 1 -14.0605 (45%). 1 -14.0605 (thater10), 1 -14.0605 causativity 1 -14.0605 exemplification 1 -14.0605 idafa (possessive) 1 -14.0605 target→source 1 -14.0605 ketrs, 1 -14.0605 consideredsystems). 1 -14.0605 “meteor”. 1 -14.0605 asterisk. 1 -14.0605 talkin bout 1 -14.0605 flattens 1 -14.0605 hrl. 1 -14.0605 和(h 1 -14.0605 ‘off-the-shelf’ 1 -14.0605 77.14 1 -14.0605 res virt 1 -14.0605 april. 1 -14.0605 years.’ 1 -14.0605 rcv1 (static) 1 -14.0605 fr1 1 -14.0605 ”israr $ytany” 1 -14.0605 1094 1 -14.0605 (hachey, 1 -14.0605 88.84 1 -14.0605 51.91 1 -14.0605 winter olympics, 1 -14.0605 λx.λy.λdi.λz.u10(x,y,di,p) 1 -14.0605 dtus. 1 -14.0605 implementation1. 1 -14.0605 adj nn: 1 -14.0605 strong) 1 -14.0605 legs. 1 -14.0605 countt(fprime,e) 1 -14.0605 k;j i;j 1 -14.0605 (owczarzak 1 -14.0605 medians 1 -14.0605 fet 1 -14.0605 o(n2)). 1 -14.0605 w+. 1 -14.0605 anotationmatcher 1 -14.0605 re-order 1 -14.0605 four-word phrase 1 -14.0605 phenomenon-specific 1 -14.0605 tiberi, 1 -14.0605 51.96 1 -14.0605 hayes, 1 -14.0605 !"#$ 1 -14.0605 12–21, 1 -14.0605 rna, 1 -14.0605 65.91 1 -14.0605 wherek 1 -14.0605 opt. 1 -14.0605 re-align 1 -14.0605 ia/x (etprime..t/fvprime..v 1 -14.0605 (auto/auto), 1 -14.0605 savg, 1 -14.0605 0null1. 1 -14.0605 (ireland 1 -14.0605 proto-agent 1 -14.0605 result(%) 1 -14.0605 mozambique, cyprus, 1 -14.0605 malouf’s 1 -14.0605 pinhan ‘latent’ 1 -14.0605 cu−2bexact 1 -14.0605 recasting 1 -14.0605 decompounding, 1 -14.0605 syntbip, 1 -14.0605 shift-reduce 87.43% 1 -14.0605 red). 1 -14.0605 maxi≤r 1 -14.0605 eventualities. 1 -14.0605 monuments 1 -14.0605 =1.04; 1 -14.0605 y∈g(x) 2∑ 1 -14.0605 hdlr, 1 -14.0605 19(2):313-330. 1 -14.0605 association strength. 1 -14.0605 (sk) 1 -14.0605 hu‘become’. 1 -14.0605 vbn dt+np 1 -14.0605 subtracted. 1 -14.0605 countermeasures, 1 -14.0605 $a 1 -14.0605 single-stream 1 -14.0605 cause> cause> 0 1 -14.0605 -他人 aka-no-tanin 1 -14.0605 “candlestick park” 1 -14.0605 terminoontology. 1 -14.0605 modelingunrestrictedcoreference 1 -14.0605 äga ’to 1 -14.0605 (langkilde, 1 -14.0605 aggressive” 1 -14.0605 underrates 1 -14.0605 98.17 1 -14.0605 il-10 production 1 -14.0605 feature.) 1 -14.0605 vehicles’ 1 -14.0605 linton 1 -14.0605 clarification, rewording, 1 -14.0605 incoherence. 1 -14.0605 phraseappearsin 1 -14.0605 rede galega 1 -14.0605 k¨uhe kuh 1 -14.0605 metavariable, 1 -14.0605 inverse-permuted 1 -14.0605 enormous. 1 -14.0605 experiments.10 1 -14.0605 standardmentionboundaries 1 -14.0605 sc). 1 -14.0605 frontier(a). 1 -14.0605 pursued. 1 -14.0605 romanized 1 -14.0605 john. 1 -14.0605 ldc2007t40) 1 -14.0605 blake, 1 -14.0605 pa, 2002. 1 -14.0605 manifold. 1 -14.0605 hcrfs 1 -14.0605 operation). 1 -14.0605 104). 1 -14.0605 tweet’s 1 -14.0605 smart-casing 1 -14.0605 2.5m 1 -14.0605 85.9% 1 -14.0605 n.o. 1 -14.0605 gmm. 1 -14.0605 bullish. 1 -14.0605 complexity; 1 -14.0605 79.76% 1 -14.0605 skeleton, 1 -14.0605 he-nom computer-acc 1 -14.0605 initiation–response 1 -14.0605 (or constrained) 1 -14.0605 annotation? 1 -14.0605 sorensen 1 -14.0605 retrained, 1 -14.0605 sri-lm toolkit’s 1 -14.0605 http://www-nlpir.nist.gov/related_projects/muc/ 1 -14.0605 reordering-rule 1 -14.0605 certificates. 1 -14.0605 malouf (2000) 1 -14.0605 86.73 1 -14.0605 (39603 1 -14.0605 attested. 1 -14.0605 belself. 1 -14.0605 wmt10. 1 -14.0605 departure/arrival 1 -14.0605 band). 1 -14.0605 agglomeration 1 -14.0605 up)? 1 -14.0605 newstest2010. 1 -14.0605 getoor 1 -14.0605 out-of-sample 1 -14.0605 (hrl) 1 -14.0605 prlm). 1 -14.0605 decompounding. 1 -14.0605 ∼15% 1 -14.0605 erors. 1 -14.0605 (festgenommen) 1 -14.0605 stewart. 1 -14.0605 corefer, 1 -14.0605 nullifying 1 -14.0605 sub-decisions 1 -14.0605 1..nv: 1 -14.0605 (14). 1 -14.0605 “member” 1 -14.0605 hansen, 1 -14.0605 significances 1 -14.0605 con. 1 -14.0605 501-1000 1-1000 1 -14.0605 suck 1 -14.0605 velopment 1 -14.0605 p5. 1 -14.0605 nineties 1 -14.0605 writing: 1 -14.0605 subdomains 1 -14.0605 n.quant.def 1 -14.0605 [0.0,1.0]. 1 -14.0605 ihmm 1 -14.0605 bleached 1 -14.0605 behavior!” 1 -14.0605 subtype="pattern" pattern="[@v←@v_[" 1 -14.0605 h=1 1 -14.0605 rd b1·θ 1 -14.0605 svenska akademiens 1 -14.0605 recently mentioned 1 -14.0605 watson’s 1 -14.0605 (capitalized) 1 -14.0605 day, date, 1 -14.0605 37–41, 1 -14.0605 1only 1 -14.0605 eisner. 1 -14.0605   1 -14.0605 combos 1 -14.0605 ‘zhou shuren’ 1 -14.0605 lexicalisation, 1 -14.0605 expparenleftbigθ1 ·f⟨1,y1,x⟩parenrightbig 1 -14.0605 vapnik. 1 -14.0605 mance. 1 -14.0605 94k 1 -14.0605 verbally, 1 -14.0605 whichwefirstuseexactstringmatchingto 1 -14.0605 sall. 1 -14.0605 sub-trees, 1 -14.0605 (nwo). 1 -14.0605 (mgi) 1 -14.0605 mirrors? 1 -14.0605 scree 1 -14.0605 f1-score. 1 -14.0605 timeliness 1 -14.0605 (saers 1 -14.0605 multi-functionality, 1 -14.0605 (no.60973053, 1 -14.0605 max r,b 1 -14.0605 incapability 1 -14.0605 removing duplicaterules, 1 -14.0605 sfb/tr8 1 -14.0605 asked: 1 -14.0605 tutalk 1 -14.0605 web-search 1 -14.0605 0.980 1 -14.0605 crabbé 1 -14.0605 top-md 1 -14.0605 memberships. 1 -14.0605 candy cigarette, 1 -14.0605 (lado, 1 -14.0605 (joachims 1 -14.0605 noun–preposition 1 -14.0605 3,800 1 -14.0605 functionl(e,d(f)). 1 -14.0605 slackness 1 -14.0605 matically 1 -14.0605 γ(w). 1 -14.0605 mwetoolkit, 1 -14.0605 sfst. 1 -14.0605 zajic 1 -14.0605 high-dimension 1 -14.0605 em-hmm: 1 -14.0605 project1. 1 -14.0605 raymond, 1 -14.0605 playful. 1 -14.0605 sgall 1 -14.0605 lexicalised. 1 -14.0605 (*: p<0.05; 1 -14.0605 33.96 1 -14.0605 lenient) 1 -14.0605 corectly. 1 -14.0605 act-level 1 -14.0605 paolillo 1 -14.0605 dethlefs@uni-bremen.de 1 -14.0605 (counted 1 -14.0605 hlen. 1 -14.0605 miwa10 1 -14.0605 “가장 ([ga-jang]: 1 -14.0605 (cp) 1 -14.0605 klapaftis, 1 -14.0605 +person 1 -14.0605 opinable 1 -14.0605 (bandura, 1 -14.0605 conditions? 1 -14.0605 6other 1 -14.0605 demanding. 1 -14.0605 x2 (film) 1 -14.0605 bop (3813) 1 -14.0605 beams. 1 -14.0605 kpbl2. 1 -14.0605 (nucle). 1 -14.0605 vlachos, 1 -14.0605 basque). 1 -14.0605 crabby:1, grumpy:1, 1 -14.0605 r∈r∪{none}, 1 -14.0605 em-estimated 1 -14.0605 abbrev.100, 1 -14.0605 fabrizio morbini, 1 -14.0605 non-ambiguous, 1 -14.0605 stsg, stssg, 1 -14.0605 (clwsd) 1 -14.0605 p<.001. 1 -14.0605 macleod 1 -14.0605 overcoverage, 1 -14.0605 healthcare. 1 -14.0605 dis-preferred 1 -14.0605 analyzers, 1 -14.0605 time zone, 1 -14.0605 t-start 1 -14.0605 (iteration 1 -14.0605 exp parenleftbiggsummationdisplay 1 -14.0605 “possible” 1 -14.0605 e-mail. 1 -14.0605 hunpos. 1 -14.0605 arase 1 -14.0605 international. 1 -14.0605 .104 1 -14.0605 1(d). 1 -14.0605 tier; 1 -14.0605 .1) 1 -14.0605 working) 1 -14.0605 600. 1 -14.0605 constraint-level 1 -14.0605 researcher. 1 -14.0605 probability/back-off 1 -14.0605 parser3 1 -14.0605 hsw 1 -14.0605 way” 1 -14.0605 section2. 1 -14.0605 covington 1 -14.0605 79.68 (80.13/79.23) 1 -14.0605 1both 1 -14.0605 62.22% 1 -14.0605 p(x1,...,xk 1 -14.0605 indeterminacy 1 -14.0605 65–70, 1 -14.0605 message-pairs 1 -14.0605 41.48 -0.18 1 -14.0605 ic1. 1 -14.0605 ˜τth. 1 -14.0605 ,∗)ing 1 -14.0605 pl.01 1 -14.0605 (lignos 1 -14.0605 ligature 1 -14.0605 0.56. 1 -14.0605 (tuning, 1 -14.0605 ptbstyle 1 -14.0605 abundance, 1 -14.0605 h0(a0,r0) 1 -14.0605 i.alegria@ehu.es 1 -14.0605 ar, 1 -14.0605 hybridizations. 1 -14.0605 postverbal 1 -14.0605 method1. 1 -14.0605 “run”. 1 -14.0605 query-ad 1 -14.0605 periments 1 -14.0605 compressive, 1 -14.0605 +ts). 1 -14.0605 emmanuel prochasson 1 -14.0605 totprd 1 -14.0605 allergies. 1 -14.0605 fluctuating 1 -14.0605 accomplished. 1 -14.0605 k.u. 1 -14.0605 isinstance=“no”> 1 -14.0605 departed 1 -14.0605 nbc 1 -14.0605 predefine 1 -14.0605 headword percolation 1 -14.0605 astubbs@cs.brandeis.edu 1 -14.0605 market investor 1 -14.0605 (§3.3). 1 -14.0605 concrete). 1 -14.0605 pi(n) 1 -14.0605 markables; 1 -14.0605 (carvalho 1 -14.0605 reducir, bajar, 1 -14.0605 matters: 1 -14.0605 requests: 1 -14.0605 miles south 1 -14.0605 ?affected 1 -14.0605 unpairedt-test. 1 -14.0605 groundwork 1 -14.0605 mesage understanding 1 -14.0605 list-wise 1 -14.0605 permissive, 1 -14.0605 mentions(also 1 -14.0605 instructor’s 1 -14.0605 ...’, 1 -14.0605 pron. 1 -14.0605 47.86 1 -14.0605 constraint) 1 -14.0605 (brinton 1 -14.0605 (jolliffe, 1 -14.0605 0-100. 1 -14.0605 (compose, sonata) 1 -14.0605 two-module 1 -14.0605 two-participant 1 -14.0605 73592 5503 1 -14.0605 53.74 1 -14.0605 bird’s 1 -14.0605 is-a, has-part, 1 -14.0605 interference. 1 -14.0605 parse; 1 -14.0605 ranker-learning 1 -14.0605 (dalrymple, 1 -14.0605 ??) 1 -14.0605 97%). 1 -14.0605 word-toword 1 -14.0605 mj. 1 -14.0605 musician). 1 -14.0605 step-wise 1 -14.0605 sieves. 1 -14.0605 nation). 1 -14.0605 he [went]x 1 -14.0605 inputs i1...i−1 1 -14.0605  log  probability 1 -14.0605 2.1.2, 1 -14.0605 credits, 1 -14.0605 mon≻mon. 1 -14.0605 processes; 1 -14.0605 addressing. 1 -14.0605 emissions, 1 -14.0605 bullinaria 1 -14.0605 viceversa. 1 -14.0605 (negatively 1 -14.0605 95.7% 1 -14.0605 cierva 1 -14.0605 §3.2, 1 -14.0605 implementationeffort. 1 -14.0605 (garner 1 -14.0605 +2.5 1 -14.0605 (0.622p/0.612r) 1 -14.0605 (mert; 1 -14.0605 p(wm1 tm1 1 -14.0605 blanket 1 -14.0605 0.583 0.309 1 -14.0605 non-baseline 1 -14.0605 c/i 1 -14.0605 (mitchel 1 -14.0605 ‘genre’ 1 -14.0605 white-space. 1 -14.0605 7639 1 -14.0605 (learning -testing) 1 -14.0605 s32s55 1 -14.0605 85.78% 1 -14.0605 in§2.1, 1 -14.0605 (qg, 1 -14.0605 plain text) 1 -14.0605 optimization-based 1 -14.0605 .678 1 -14.0605 ashenden, 1 -14.0605 lexrank, 1 -14.0605 segment-level. 1 -14.0605 metrics? 1 -14.0605 shift– 1 -14.0605 kraljic 1 -14.0605 n00141010117. 1 -14.0605 “c:±1” 1 -14.0605 dsummationdisplay d=1 1 -14.0605 [h3,k,h4h5,j], 1 -14.0605 mask 1 -14.0605 retained). 1 -14.0605 human-assigned 1 -14.0605 pen trebank 1 -14.0605 piece. 1 -14.0605 xuv +xvw−xuw≤1 1 -14.0605 {fb, ddi} 1 -14.0605 proficiency: 1 -14.0605 havethreetypes:names,nominalmentionsandpronouns. 1 -14.0605 tomanek 1 -14.0605 49–57, 1 -14.0605 b∈inter(e) (1−sb), 1 -14.0605 (es..sprime/fvprime..v 1 -14.0605 online-b, 1 -14.0605 anonymous, 1 -14.0605 (4b), 1 -14.0605 writers. 1 -14.0605 quantifiers), 1 -14.0605 effect), 1 -14.0605 (begum 1 -14.0605 clustering’s 1 -14.0605 chiang. 1 -14.0605 context-identifier. 1 -14.0605 29.08 1 -14.0605 tae yano 1 -14.0605 pre-boundary 1 -14.0605 (la repubblica 1 -14.0605 congruence. 1 -14.0605 rphilli, lester}@ncsu.edu 1 -14.0605 numerical. 1 -14.0605 exhaustive: 1 -14.0605 taboada. 1 -14.0605 mr hicks 1 -14.0605 (74) 1 -14.0605 1-15. 1 -14.0605 rule-based. 1 -14.0605 2654(3.65%) 1 -14.0605 summarization) 1 -14.0605 lcs’s 1 -14.0605 “lock” icon). 1 -14.0605 zc≤k 1 -14.0605 ltrc, 1 -14.0605 en–de, 1 -14.0605 (64) 1 -14.0605 confidence-driven 1 -14.0605 ’07. 1 -14.0605 page). 1 -14.0605 modalities. 1 -14.0605 zipper. 1 -14.0605 lacalle, 1 -14.0605 definitenounphrase. 1 -14.0605 pobject 1 -14.0605 .003 1 -14.0605 ont attaqu´e 1 -14.0605 computer-nom 1 -14.0605 near-optimal solution. 1 -14.0605 1 -14.0605 celex (baayen 1 -14.0605 astheevaluationcorpus.the 1 -14.0605 aunual 1 -14.0605 monolinguals. 1 -14.0605 ‘addendum’ 1 -14.0605 commonplace. 1 -14.0605 septembrie 1 -14.0605 giesbrecht, 2010), 1 -14.0605 ineffective, 1 -14.0605 http://ngrams.googlelabs.com/datasets. 1 -14.0605 glacier’s 1 -14.0605 alignments). 1 -14.0605 meurers (2003b) 1 -14.0605 bow: 1 -14.0605 left-attach, 1 -14.0605 outside. 1 -14.0605 están brillantes 1 -14.0605 (flickinger 1 -14.0605 edgefactored 1 -14.0605 appendix: 1 -14.0605 (pi) 1 -14.0605 hit; 1 -14.0605 braves macintosh 1 -14.0605 musicians. 1 -14.0605 graphs). 1 -14.0605 damages, 1 -14.0605 typef1 1 -14.0605 (portnuf, 1 -14.0605 neologisms, 1 -14.0605 multi-avatar 1 -14.0605 φn(wprime) 1 -14.0605 itspoke. 1 -14.0605 avgf0.5 1 -14.0605 0.70.8 1 -14.0605 s=0.4 1 -14.0605 schatzmann 1 -14.0605 35%, 1 -14.0605 zhao-eng. 1 -14.0605 pseudo-lesk 1 -14.0605 moore. 1 -14.0605 0.0001. 1 -14.0605 (aissen, 1 -14.0605 utrecht leave? 1 -14.0605 single-cluster 1 -14.0605 (forman, 1 -14.0605 us? 1 -14.0605 cul-desac 1 -14.0605 longest-match 1 -14.0605 lrec’10. 1 -14.0605 biocuration. 1 -14.0605 rejects, 1 -14.0605 tri-grams 1 -14.0605 varma. 1 -14.0605 ceoof(x1,x2) 1 -14.0605 subresult 1 -14.0605 tratz, 1 -14.0605 .69, 1 -14.0605 ‘x’-chain. 1 -14.0605 jette.viethen@mq.edu.au 1 -14.0605 large-coverage 1 -14.0605 etime. 1 -14.0605 “iz(only).” 1 -14.0605 orthographically. 1 -14.0605 attenuate 1 -14.0605 (wellington, 1 -14.0605 subsequence. 1 -14.0605 term.) 1 -14.0605 infarction, 1 -14.0605 role’s 1 -14.0605 s2j (s2j 1 -14.0605 2-4) 1 -14.0605 cathedrals. 1 -14.0605 realisations, 1 -14.0605 99.76 71.6 1 -14.0605 please? 1 -14.0605 relationspecific 1 -14.0605 java. 1 -14.0605 (bottom), 1 -14.0605 genes/proteins 1 -14.0605 stringintotheidenticalpre-clusters.withthesepre- 1 -14.0605 opaque, 1 -14.0605 topic? 1 -14.0605 incorect 1 -14.0605 mother. 1 -14.0605 realizer, 1 -14.0605 accentuated 1 -14.0605 iso 639-3 1 -14.0605 end-goal 1 -14.0605 ves 1 -14.0605 (+1.5) 1 -14.0605 mqa. 1 -14.0605 objective-spam. 1 -14.0605 giver” 1 -14.0605 do?). 1 -14.0605 67%. 1 -14.0605 (blackwood 1 -14.0605 emission. 1 -14.0605 referral 1 -14.0605 tapers 1 -14.0605 ££entity-sequence. 1 -14.0605 (σ) 1 -14.0605 summationtextss=1fs(zs) 1 -14.0605 80.09 1 -14.0605 thaiherbminer, 1 -14.0605 1969; chen, 1 -14.0605 ill-suited 1 -14.0605 whitespace, 1 -14.0605 (huet, 1 -14.0605 lexeme (w,vectorc) 1 -14.0605 (scriptwriters) nonconsciously 1 -14.0605 dhi, 1 -14.0605 voca. 1 -14.0605 pinker 1 -14.0605 sells, 1 -14.0605 (we). 1 -14.0605 bg882>d 1 -14.0605 lei, 1 -14.0605 helpfulness; 1 -14.0605 xiaoyi ma 1 -14.0605 m∗n 1 -14.0605 7,400 1 -14.0605 precision@ 1 -14.0605 poesio’s 1 -14.0605 browses. 1 -14.0605 “cleaning 1 -14.0605 testing: 1 -14.0605 stressing 1 -14.0605 recherche (anr) 1 -14.0605 78.95% 1 -14.0605 barr, 1 -14.0605 uploader 1 -14.0605 nadya.f.sg=erg 1 -14.0605 42.75 49.30 1 -14.0605 xtract. 1 -14.0605 (relatively concrete), 1 -14.0605 openly 1 -14.0605 subspace. 1 -14.0605 (candidates) 1 -14.0605 schultz, tazo, 1 -14.0605 met1.2 met1.3 1 -14.0605 levine (2002), 1 -14.0605 fprime countt(f 1 -14.0605 lo1. 1 -14.0605 pt(x)ps(x). 1 -14.0605 advertisements. 1 -14.0605 spmt: 1 -14.0605 (schr¨oder 1 -14.0605 emotion-associated 1 -14.0605 degrees? 1 -14.0605 θ). 1 -14.0605 conventionalized 1 -14.0605 117–120, 1 -14.0605 universidad complutense 1 -14.0605 semantic). 1 -14.0605 treasure trove 1 -14.0605 yymarton@us.ibm.com 1 -14.0605 d.f., 1 -14.0605 (atb), 1 -14.0605 associated. 1 -14.0605 “state” 1 -14.0605 (auto/gold) 1 -14.0605 resumptions 1 -14.0605 importantthroughout 1 -14.0605 reuse, 1 -14.0605 ¯w. 1 -14.0605 clda-t 1 -14.0605 cited-or-not 1 -14.0605 vectorr. 1 -14.0605 bılal=ko ısara 1 -14.0605 def= summationtextt 1 -14.0605 alessandro lenci 1 -14.0605 (sinha 2009a): 1 -14.0605 mation 1 -14.0605 ‘path’, ‘straight’). 1 -14.0605 eventsent 1 -14.0605 22,805,699 1 -14.0605 (posted 1 -14.0605 first; 1 -14.0605 kedis, 1 -14.0605 corect. 1 -14.0605 histograms (lh) 1 -14.0605 0.72. 1 -14.0605 216594 (classic 1 -14.0605 dysarthria, 1 -14.0605 no-con no-emb 1 -14.0605 nonfiltered 1 -14.0605 x-axis: 1 -14.0605 five-character 1 -14.0605 (orig. 1 -14.0605 triphones. 1 -14.0605 multifunctional, 1 -14.0605 coreferencing 1 -14.0605 mercury lounge 1 -14.0605 (4.4) 1 -14.0605 eopt 1 -14.0605 6.3). 1 -14.0605 inclusion. 1 -14.0605 prune, 1 -14.0605 actors. 1 -14.0605 56.62 1 -14.0605 o.w. 1 -14.0605 10−7 1 -14.0605 parent). 1 -14.0605 efficiency/accuracy 1 -14.0605 blick 1 -14.0605 lex. 1 -14.0605 decouples 1 -14.0605 with-def-vapor. 1 -14.0605 using/not 1 -14.0605 discarded). 1 -14.0605 corpus’s 1 -14.0605 temporally, 1 -14.0605 thai. 1 -14.0605 bethesda, 20894, 1 -14.0605 matsubayashi 1 -14.0605 11-point 1 -14.0605 javier artiles, 1 -14.0605 workspace. 1 -14.0605 user-model 1 -14.0605 b+l, 1 -14.0605 pay. 1 -14.0605 2.21 45.10 1 -14.0605 p-values. 1 -14.0605 metric; 1 -14.0605 hardisty, 1 -14.0605 e-rater, 1 -14.0605 (clte) 1 -14.0605 illogical 1 -14.0605 .001) 1 -14.0605 npov. 1 -14.0605 maffi 1 -14.0605 (az) 1 -14.0605 msh 1 -14.0605 raskin 1 -14.0605 4.8, 1 -14.0605 (chosen 1 -14.0605 high-value 1 -14.0605 non-reduced 1 -14.0605 decipherments. 1 -14.0605 anne vilnat 1 -14.0605 giuliano, 1 -14.0605 san diego, 1 -14.0605 sketch, 1 -14.0605 builtpps.previouspp(bestarc) 1 -14.0605 mj). 1 -14.0605 twenty-two 1 -14.0605 58.1%1606 1 -14.0605 h5, 1 -14.0605 words”. 1 -14.0605 (lappin 1 -14.0605 fruit). 1 -14.0605 “world heritage 1 -14.0605 english–dutch. 1 -14.0605 (gud) 1 -14.0605 1(a)), 1 -14.0605 75083-0688 1 -14.0605 ours) 1 -14.0605 §4.1) 1 -14.0605   1 -14.0605 lesk. 1 -14.0605 every/d line/h] 1 -14.0605 watermarker. 1 -14.0605 >mt2. 1 -14.0605 82%. 1 -14.0605 en–de. 1 -14.0605 crossclassified 1 -14.0605 (→) 1 -14.0605 next: 1 -14.0605 (2007)); 1 -14.0605 2000)”. 1 -14.0605 autoslog-ts 1 -14.0605 markus saers 1 -14.0605 available.1 1 -14.0605 g1850 g1839g3013 1 -14.0605 (luscher, 1 -14.0605 fp7), 1 -14.0605 astray. 1 -14.0605 krs reddy@students.iiit.ac.in 1 -14.0605 question]. 1 -14.0605 n1g 2w1 1 -14.0605 p<.05 1 -14.0605 posa nega 1 -14.0605 “is teaching” 1 -14.0605 (particle presence) 1 -14.0605 18.27 30.66 1 -14.0605 incoherences. 1 -14.0605 (-2.9, 1 -14.0605 1215 1 -14.0605 (616 1 -14.0605 (85.25/81.30)** 1 -14.0605 f­measure all phrases 1 -14.0605 d.f. 1 -14.0605 b, auto-feedback, 1 -14.0605 jzhao@nlpr.ia.ac.cn 1 -14.0605 agreement”. 1 -14.0605 p<0.05: 1 -14.0605 pairwise-independent 1 -14.0605 kahani ‘story’ 1 -14.0605 diseases). 1 -14.0605 unfairly 1 -14.0605 76.00 1 -14.0605 lpsolve: http://lpsolve.sourceforge.net/ 1 -14.0605 caitra 1 -14.0605 (pbsmt), 1 -14.0605 well-preserved 1 -14.0605 “emotion”. 1 -14.0605 re-predict 1 -14.0605 .cm 1 -14.0605 osaka) 1 -14.0605 eat-pst-dec 1 -14.0605 alfonso perez 1 -14.0605 meta-relations 1 -14.0605 abbreviates bels,p, 1 -14.0605 issue). 1 -14.0605 perpetrators 1 -14.0605 kübler 1 -14.0605 coper-pye, 1 -14.0605 wi…wn 1 -14.0605 92.3∗ 1 -14.0605 fujiki 1 -14.0605 expressivity. 1 -14.0605 (4a). 1 -14.0605 #relevant intersectiontext#retrieved 1 -14.0605 f-measures. 1 -14.0605 label-collapsed 1 -14.0605 word-gram 1 -14.0605 00.0 78.2 1 -14.0605 mbrsc-e/conjoin 1 -14.0605 0.89% 1 -14.0605 myle ott, 1 -14.0605 a a∼ 1 -14.0605 class-by-class 1 -14.0605 83.21 1 -14.0605 α}. 1 -14.0605 d. ye.” 1 -14.0605 (2.1m 1 -14.0605 reaffirmed 1 -14.0605 σ∗1·“σβk...σγk”·σ∗2,pi〉 1 -14.0605 (0.17) 1 -14.0605 italics). 1 -14.0605 qg, 1 -14.0605 toy-ess-ta. 1 -14.0605 factor; 1 -14.0605 traversed. 1 -14.0605 .95 .77 1 -14.0605 thrax. 1 -14.0605 34–42, 1 -14.0605 0.617f1 1 -14.0605 communicate, 1 -14.0605 contextless 1 -14.0605 (tong, 1 -14.0605 rank≤2 rank≤3 1 -14.0605 tlebyusinggoldstandardmentionboundaries.also 1 -14.0605 section,we 1 -14.0605 few) 1 -14.0605 log-likelihood) 1 -14.0605 cost/benefit trade-off 1 -14.0605 sabatini, 1 -14.0605 99.2 1.008 1 -14.0605 time-outs. 1 -14.0605 (seidl 1 -14.0605 robert.dale@mq.edu.au 1 -14.0605 pow. 1 -14.0605 beacon 1 -14.0605 ( 21.74%) 1 -14.0605 unanswerable 1 -14.0605 belong. 1 -14.0605 meansnoun 1 -14.0605 jmwe’s 1 -14.0605 16%. 1 -14.0605 “play role”. 1 -14.0605 (target)? 1 -14.0605 52.67% 1 -14.0605 computermediated 1 -14.0605 s(i) =α 1 -14.0605 2,303 1 -14.0605 8 bitsa 1 -14.0605 1900’10 ’20 1 -14.0605 vowel). 1 -14.0605 selflabeled 1 -14.0605 stabilization 1 -14.0605 corpusextracted 1 -14.0605 see. 1 -14.0605 rprime =〈k,s∗1·“∇”·s∗2,σ∗1·“∇”·σ∗2,piprime〉 1 -14.0605 game-like 1 -14.0605 saar. 1 -14.0605 62,061 1 -14.0605 incompatibility 1 -14.0605 discriminatively, 1 -14.0605 q-queries. 1 -14.0605 dekang lin. 1 -14.0605 n3(d). 1 -14.0605 (seroussi 1 -14.0605 57.87% 1 -14.0605 pasive 1 -14.0605 79.82++ 1 -14.0605 tents, 1 -14.0605 “u. 1 -14.0605 cha, hca, 1 -14.0605 0.929 1 -14.0605 chunk– pair–distance 1 -14.0605 [0.1, 1 -14.0605 m=20). 1 -14.0605 cognate, 1 -14.0605 short),which 1 -14.0605 tobach mit 1 -14.0605 unmistakable. 1 -14.0605 (1/4). 1 -14.0605 4x 1 -14.0605 compare. 1 -14.0605 minimality 1 -14.0605 “-1” 1 -14.0605 past; 1 -14.0605 (pbt:normal 1 -14.0605 st. helens 1 -14.0605 345 929 1 -14.0605 wsd/wsi 1 -14.0605 catlett, 1 -14.0605 stringas 1 -14.0605 (e, agent) 1 -14.0605 underestimated 1 -14.0605 one-node 1 -14.0605 ‘be’. 1 -14.0605 auto-f. 1 -14.0605 985 1 -14.0605 r(x’, 1 -14.0605 stamp. 1 -14.0605 s32s57 s32s49s48 1 -14.0605 family; 1 -14.0605 marched, 1 -14.0605 c1···cm. 1 -14.0605 0.00001. 1 -14.0605 (ogren 1 -14.0605 teknik link¨opings 1 -14.0605 (cart) 1 -14.0605 headings. 1 -14.0605 citationbased 1 -14.0605 deng, 1 -14.0605 fine+pattern 1 -14.0605 oracle skylines 1 -14.0605 multi-step 1 -14.0605 gapping. 1 -14.0605 entirety, 1 -14.0605 huffy, 1 -14.0605 flexibilities. 1 -14.0605 bayesiansolution:d 1 -14.0605 (“high”) 1 -14.0605 phrases”. 1 -14.0605 thirst.n, 1 -14.0605 pak, a., 1 -14.0605 probabilities), 1 -14.0605 free” 1 -14.0605 (82% 1 -14.0605 alan finlayson 1 -14.0605 nonparallel. 1 -14.0605   1 -14.0605 terminate, 1 -14.0605 15f 1 -14.0605 scorable 1 -14.0605 akcay, 1 -14.0605 tropical storms 1 -14.0605 quick, 1 -14.0605 mdl. 1 -14.0605 super-entities. 1 -14.0605 genomes. 1 -14.0605 michel gagnon 1 -14.0605 mwd 1 -14.0605 syartyn 1 -14.0605 mbytes. 1 -14.0605 28.29 1 -14.0605 ˆei←ecur 1 -14.0605 types.) 1 -14.0605 occurring, 1 -14.0605 (1.3 1 -14.0605 64.9%. 1 -14.0605 table’s 1 -14.0605 reodering. 1 -14.0605 unnecessary. 1 -14.0605 shift (sh) 1 -14.0605 +0.78 1 -14.0605 train (english): 1 -14.0605 genova, 1 -14.0605 1 -14.0605 sections 2.2 1 -14.0605 micromacro 1 -14.0605 ldc2004t07, 1 -14.0605 zeroed 1 -14.0605 cation, 1 -14.0605 (tildewidef,tildewidee) 1 -14.0605 context-vectors, 1 -14.0605 replicability 1 -14.0605 x∈a,y∈b d(x,y) 1 -14.0605 forgo 1 -14.0605 byblos hr 1 -14.0605 s1’s 1 -14.0605 54.84 1 -14.0605 improbable. 1 -14.0605 ©2011associationforcomputationallinguistics computationallinguistics 1 -14.0605 isolated. 1 -14.0605 {act art} 1 -14.0605 paragraph-sized 1 -14.0605 “`a deux 1 -14.0605 grouping. 1 -14.0605 thesaurus) 1 -14.0605 x3 x2 1 -14.0605 fourth. 1 -14.0605 placed. 1 -14.0605 wmt2010, 1 -14.0605 scientifique, bp 1 -14.0605 uniform(1,m). 1 -14.0605 0.970 1 -14.0605 usual. 1 -14.0605 screens, 1 -14.0605 andxj. 1 -14.0605 null-grams 1 -14.0605 baskaran sankaran 1 -14.0605 (526) 1 -14.0605 eisele 1 -14.0605 spink. 1 -14.0605 designers. 1 -14.0605 mysql database, 1 -14.0605 thek-nn 1 -14.0605 mission. 1 -14.0605 oovcorp. 1 -14.0605 pbsmt. 1 -14.0605 illinois-coref, 1 -14.0605 (mulkar-mehta 1 -14.0605 expend 1 -14.0605 24.38 64.26 1 -14.0605 (mead). 1 -14.0605 (prep to, 1 -14.0605 philadelphia: 1 -14.0605 (candidate) 1 -14.0605 srlfeatures,andnamedentityfeatures 1 -14.0605 tj1. 1 -14.0605 0≤u≤uprime 1 -14.0605 corpora." 1 -14.0605 (62.5 1 -14.0605 count(?r) 1 -14.0605 h(c) 1 -14.0605 election) 1 -14.0605 a→ epsilon1/epsilon1 1 -14.0605 villages 1 -14.0605 untargeted 1 -14.0605 dealers. 1 -14.0605 drs. 1 -14.0605 4.2: 1 -14.0605 px(e,o) 1 -14.0605 colectivity 1 -14.0605 cachepipe. 1 -14.0605 use-cases 1 -14.0605 experi 1 -14.0605 ∀λr 1 -14.0605 re-scored, 1 -14.0605 harris (1955) 1 -14.0605 -6.28514 1 -14.0605 invocation 1 -14.0605 synonym), 1 -14.0605 56.0% 1 -14.0605 non-gappy 1 -14.0605 halfway 1 -14.0605 lengthnormalized 1 -14.0605 (eprime ∈dn(f) 1 -14.0605 neue 1 -14.0605 l)ising 1 -14.0605 central. 1 -14.0605 trials 4.825 1 -14.0605 terminology). 1 -14.0605 1969; powell, 1 -14.0605 pool-based 1 -14.0605 smdps 1 -14.0605 loss-driven 1 -14.0605 ses, 1 -14.0605 romanian). 1 -14.0605 concurrently 1 -14.0605 union) 1 -14.0605 w1,w2∈σ∗, 1 -14.0605 a¯ar. 1 -14.0605 64.99 1 -14.0605 constellations 1 -14.0605 ”golden age” 1 -14.0605 discuses 1 -14.0605 vg. 1 -14.0605 prone. 1 -14.0605 okayandwhatdidthey tellyou? 1 -14.0605 161–169, 1 -14.0605 successfully, 1 -14.0605 800. 1 -14.0605 “al-df-p 1 -14.0605 d.j.a.de.kok@rug.nl 1 -14.0605 laws. 1 -14.0605 (memms) 1 -14.0605 isolation; 1 -14.0605 (sec.) 1 -14.0605 (korkontzelos 1 -14.0605 subj, obj, 1 -14.0605 (lf/uf) 1 -14.0605 lowercase. 1 -14.0605 5197 1 -14.0605 j−1 1 -14.0605 cii, 1 -14.0605 multiset. 1 -14.0605 artist. 1 -14.0605 p(a), 1 -14.0605 j” mktb 1 -14.0605 frequently-occurring 1 -14.0605 [aε1,aε2]. 1 -14.0605 sawaf, 1 -14.0605 lexicalsemantic 1 -14.0605 gscs 1 -14.0605 tandis 1 -14.0605 lim. 1 -14.0605 (2011a), 1 -14.0605 rmunro@stanford.edu 1 -14.0605 0.332±0.002 1 -14.0605 builtpps. 1 -14.0605 8when 1 -14.0605 34.78 1 -14.0605 40.62 1 -14.0605 (ships, vessels) 1 -14.0605 delimiter, 1 -14.0605 lower). 1 -14.0605 27.1%. 1 -14.0605 n-gram; 1 -14.0605 trale (meurers, 1 -14.0605 headline) 1 -14.0605 33.74 1 -14.0605 damerau 1 -14.0605 tences. 1 -14.0605 dale-chall 1 -14.0605 box). 1 -14.0605 .676 1 -14.0605 da-to-msa 1 -14.0605 interoperability. 1 -14.0605 cmu-heaf. 1 -14.0605 swedish 82.55 1 -14.0605 68.86 1 -14.0605 (bergkirkpatrick 1 -14.0605 jmwe. 1 -14.0605 elimination, 1 -14.0605 73.0% 1 -14.0605 dost´al, 1 -14.0605 78.21% 1 -14.0605 (sch¨onhofen, 1 -14.0605 -0.02 82.39 1 -14.0605 non-verbally, 1 -14.0605 (dfg). 1 -14.0605 exacerbate 1 -14.0605 uk’s 1 -14.0605 ⊗sl ⊗βv) 1 -14.0605 (nanba 1 -14.0605 n+ap. 1 -14.0605 {amit,hal}@umiacs.umd.edu 1 -14.0605 soundex. 1 -14.0605 cross-linked 1 -14.0605 ontology’s 1 -14.0605 svm-n-gram 1 -14.0605 described) 1 -14.0605 concept– 1 -14.0605 300%. 1 -14.0605 relief. 1 -14.0605 pinpointing 1 -14.0605 person/years. 1 -14.0605 fires, 1 -14.0605 supervector 1 -14.0605 order; 1 -14.0605 tla 1 -14.0605 [wubben 1 -14.0605 (tikk 1 -14.0605 compositionality? 1 -14.0605 fl* .465 1 -14.0605 caucci 1 -14.0605 cook. 1 -14.0605 +0.96 1 -14.0605 initialised 1 -14.0605 [the park]ground. 1 -14.0605 higher), 1 -14.0605 es→en en→es 1 -14.0605 (bermingham 1 -14.0605 .9873p,lp 1 -14.0605 adding 3.21% 1 -14.0605 (very good), 1 -14.0605 1994)) 1 -14.0605 240. 1 -14.0605 adapt) 1 -14.0605 00115193-a 1 -14.0605 φn(w) 1 -14.0605 reassignments 1 -14.0605 pass-pst-dec 1 -14.0605 inning. 1 -14.0605 identity: 1 -14.0605 major. 1 -14.0605 rastrow 1 -14.0605 multi-review 1 -14.0605 30–37, 1 -14.0605 300% 1 -14.0605 concept→word 1 -14.0605 explicated 1 -14.0605 greta. 1 -14.0605 write. 1 -14.0605 normalizedcut 1 -14.0605 xxx.txt 1 -14.0605 legal structuresym 1 -14.0605 constraint solving. 1 -14.0605 en-en 1 -14.0605 entraining 1 -14.0605 (14/100) 1 -14.0605 74.56 45.04 1 -14.0605 mlp, 1 -14.0605 fw. 1 -14.0605 3c. 1 -14.0605 amazons 1 -14.0605 inst(x, 1)) 1 -14.0605 icwr+cofer: (δ 1 -14.0605 cssdb. 1 -14.0605 cj? 1 -14.0605 maintained, 1 -14.0605 forcoreferenceresolutioninsection3,evaluateitin section4,andconcludeinsection5. 1 -14.0605 pennacchiotti. 1 -14.0605 encode. 1 -14.0605 (woodsend 1 -14.0605 vijay-shanker, 1 -14.0605 idiomatic. 1 -14.0605 }”) 1 -14.0605 identically, 1 -14.0605 discrepancy.n. 1 -14.0605 z-score. 1 -14.0605 )· ·p(cnknk−1+1,nk−1,gk)} 1 -14.0605 covariance) 1 -14.0605 .c2m2/d 1 -14.0605 “still”, “already”, 1 -14.0605 saliency. 1 -14.0605 rtl, 1 -14.0605 (n´ev´eol 1 -14.0605 elaborately 1 -14.0605 (elliptical 1 -14.0605 monolingual. 1 -14.0605 (re 1 -14.0605 rescoring: 1 -14.0605 greedy-based 1 -14.0605 ero leof 1 -14.0605 definiteness. 1 -14.0605 » (ke) 1 -14.0605 package3 1 -14.0605 r-su4, 1 -14.0605 exploit. 1 -14.0605 (fr-en) 1 -14.0605 misc, 1 -14.0605 max: y(q)j 1 -14.0605 tutoring r(97) 1 -14.0605 ‘hundred’. 1 -14.0605 (rutter 1 -14.0605 (listnet, 1 -14.0605 s(y), 1 -14.0605 language-model-based 1 -14.0605 transformations) 1 -14.0605 firstextractall 1 -14.0605 n(0,0.01). 1 -14.0605 neats 1 -14.0605 kosovo was 1 -14.0605 antwerp, belgium. 1 -14.0605 cups 1 -14.0605 replicability. 1 -14.0605 (triggers) 1 -14.0605 46.9% 1 -14.0605 geo:area ?a 1 -14.0605 43–53, 1 -14.0605 innovations. 1 -14.0605 (e,f). 1 -14.0605 unequivocally 1 -14.0605 f’s 1 -14.0605 (2006)’s 1 -14.0605 trust. 1 -14.0605 clustered, 1 -14.0605 2,290 1 -14.0605 mprimek, 1 -14.0605 (caraballo 1 -14.0605 18.16% 1 -14.0605 average-rating 1 -14.0605 applications; 1 -14.0605 news-making 1 -14.0605 re-tweeting 1 -14.0605 280m 1 -14.0605 tomorrow-morning 1 -14.0605 live, 1 -14.0605 masked 1 -14.0605 urge 1 -14.0605 subproblem. 1 -14.0605 bite 1 -14.0605 pln 1 -14.0605 big bang 1 -14.0605 opportunities. 1 -14.0605 sixteenth 1 -14.0605 (messages 1 -14.0605 asigning 1 -14.0605 transductive, 1 -14.0605 palladians 1 -14.0605 92.64 92.04 1 -14.0605 sql 1 -14.0605 group%1:03:00. 1 -14.0605 (dummy), 1 -14.0605 calibration. 1 -14.0605 pyp-1hmm-lm. 1 -14.0605 hello. 1 -14.0605 ntc-8 1 -14.0605 50). 1 -14.0605 sub-thread 1 -14.0605 78.49 1 -14.0605 infinity, 1 -14.0605 subsumer (lowest 1 -14.0605 performed; 1 -14.0605 i0. 1 -14.0605 0.66±0.08 1 -14.0605 .724 1 -14.0605 na´s´safa saliymun 1 -14.0605 doi-nbc n10ap20031, 1 -14.0605 ?� 1 -14.0605 (8.5 1 -14.0605 collins, tommi 1 -14.0605 mcc course 1 -14.0605 agreements, 1 -14.0605 λis 1 -14.0605 70.86 1 -14.0605 describesour 1 -14.0605 (bloodgood 1 -14.0605 (mccowan 1 -14.0605 (gw). 1 -14.0605 33.8k 1 -14.0605 hirasawa 1 -14.0605 svos 1 -14.0605 sensibility. 1 -14.0605 lopes (1999) 1 -14.0605 – 1086 1 -14.0605 endeavor. 1 -14.0605 non-synonymous 1 -14.0605 themaindifferencebetweenontonotesand 1 -14.0605 fetches 1 -14.0605 observation-based 1 -14.0605 volumeinfo 1 -14.0605 inspects 1 -14.0605 pre-clusterpairs. 1 -14.0605 batch-mode 1 -14.0605 cd(wi) 1 -14.0605 n-foil’s 1 -14.0605 directly; 1 -14.0605 binary-dirt 1 -14.0605 2400 1 -14.0605 (zeman, 1 -14.0605 implementa 1 -14.0605 abnormality 1 -14.0605 inlinks. 1 -14.0605 drop-down 1 -14.0605 crowdflower. 1 -14.0605 duplication, 1 -14.0605 cebuano 1 -14.0605 unfair, 1 -14.0605 cloa. 1 -14.0605 mead. 1 -14.0605 sub-models 1 -14.0605 pienaar, 1 -14.0605 same). 1 -14.0605 (rumshisky 1 -14.0605 13.5, 1 -14.0605 correct? 1 -14.0605 (map). 1 -14.0605 (clda). 1 -14.0605 pre. 1 -14.0605 24.09 63.82 1 -14.0605 icall 1 -14.0605 sucks 1 -14.0605 (p/a). 1 -14.0605 accession numbers. 1 -14.0605 model.) 1 -14.0605 kasten 1 -14.0605 microplanning 1 -14.0605 no-weight, doc-pair, 1 -14.0605 near-native 1 -14.0605 rerankers) 1 -14.0605 nphard. 1 -14.0605 hyperonyms 1 -14.0605 timothy baldwin♠♥ 1 -14.0605 antonio fernández, 1 -14.0605 cognitive-affective 1 -14.0605 range). 1 -14.0605 fooled 1 -14.0605 co-decoding, 1 -14.0605 (mcnemar, 1947). 1 -14.0605 (wer), 1 -14.0605 pr ´ecis 1 -14.0605 ���q 1 -14.0605 mature. 1 -14.0605 pi1. 1 -14.0605 misuse, 1 -14.0605 forecasting, 1 -14.0605 ilp, sub-modular 1 -14.0605 alsumait 1 -14.0605 whole-utterance 1 -14.0605 1322 1 -14.0605 conciseness. 1 -14.0605 ep/g020981/1. 1 -14.0605 universally. 1 -14.0605 k.f. 1 -14.0605 hindering 1 -14.0605 two? 1 -14.0605 (klsh) 1 -14.0605 word-link 1 -14.0605 pivot. 1 -14.0605 hardware requirements 1 -14.0605 (meyers, 1 -14.0605 noun-baseline. 1 -14.0605 bin−2 bin−2−lr 1 -14.0605 s=0.1 s=0.8 1 -14.0605 case.3 1 -14.0605 probabilistically-motivated 1 -14.0605 fazer exame 1 -14.0605 0.2901 0.2805 1 -14.0605 pnct 1 -14.0605 tesla-b. 1 -14.0605 zipfian, 1 -14.0605 objection 1 -14.0605 presignature 1 -14.0605 attached, 1 -14.0605 urban legends 1 -14.0605 unifiedclassificationmodelfor 1 -14.0605 fold’s 1 -14.0605 pages 388–395. 1 -14.0605 learning-based. 1 -14.0605 partialparses. 1 -14.0605 80.11++ 1 -14.0605 mica, 1 -14.0605 +%'$$ ,%$$$ 1 -14.0605 codewords, 1 -14.0605 system,we 1 -14.0605 advance; 1 -14.0605 gr/t04540/01). 1 -14.0605 ural 1 -14.0605 site 10185 1 -14.0605 preslav 1 -14.0605 steven bethard, 1 -14.0605 multisystem 1 -14.0605 screening 1 -14.0605 piprime =〈pi1,...,pij,k,pij+1,...,pik−1〉 1 -14.0605 24–29, 1 -14.0605 bhunter 1 -14.0605 51.8 .398 1 -14.0605 exp2: 1 -14.0605 (sch¨onhofen 1 -14.0605 “barack hussein 1 -14.0605 cwcd 1 -14.0605 pos(o) 1 -14.0605 keyboard, 1 -14.0605 23%. 1 -14.0605 seventy. 1 -14.0605 θpz,y 1 -14.0605 borders. 1 -14.0605 ocf 1 -14.0605 prefer. 1 -14.0605 do.pass-comp 1 -14.0605 sourcing. 1 -14.0605 reassembled 1 -14.0605 yuki 1 -14.0605 10th. 1 -14.0605 syntaxsemantics 1 -14.0605 felicitous. 1 -14.0605 box 2704, 1 -14.0605 (inside) 1 -14.0605 tuple-bonus 1 -14.0605 sums. 1 -14.0605 wordspaces 1 -14.0605 (characterised 1 -14.0605 equating 1 -14.0605 hoeks, 1 -14.0605 77.67 1 -14.0605 herb 1 -14.0605 million), 1 -14.0605 ciency, 1 -14.0605 frequency-weighted 1 -14.0605 denotesthe 1 -14.0605 1640 1 -14.0605 reminders 1 -14.0605 2011c). 1 -14.0605 (q&a) 1 -14.0605 cognate. 1 -14.0605 in-vivo 1 -14.0605 nullnullnullnullnullnullnullr. 1 -14.0605 data3. 1 -14.0605 rotated 1 -14.0605 sharif. 1 -14.0605 g75g393g286g374g855 g898g271g396g381g410g346g286g396g856g910g899 1 -14.0605 codewords. 1 -14.0605 gelman 1 -14.0605 (nn handel)) 1 -14.0605 (agrawal 1 -14.0605 princess diana’s 1 -14.0605 s.t.∀u,v,w∈v 1 -14.0605 switchboard. 1 -14.0605 textpro 1 -14.0605 samples; 1 -14.0605 separatelyanduse 1 -14.0605 treatment: 1 -14.0605 genetics. 1 -14.0605 ,...,o(q)n(q) 1 -14.0605 dialectal affixational 1 -14.0605 1964). 1 -14.0605 i1,2, girl3,5 1 -14.0605 weblogs. 1 -14.0605 acceptance. 1 -14.0605 uima: 1 -14.0605 prometeo/2009/014. 1 -14.0605 -2.0460036 1 -14.0605 prima facie 1 -14.0605 (o) os/o 1 -14.0605 ,x)= 1 -14.0605 gutenberg, 1 -14.0605 frontiers 1 -14.0605 cos(w0, ˆw) 1 -14.0605 54.40% 70.19% 1 -14.0605 (glass 1 -14.0605 vladimir i. 1 -14.0605 telephony-based 1 -14.0605 1’s. 1 -14.0605 (generation). 1 -14.0605 200gd avan. 1 -14.0605 khz. 1 -14.0605 structures) 1 -14.0605 everyh∈h 1 -14.0605 water). 1 -14.0605 and... 1 -14.0605 tks 1 -14.0605 smoking habits 1 -14.0605 γin γout 1 -14.0605 braceleftbigsummationtext m,n∈e,nnegationslash=mψa(m,n) 1 -14.0605 1120 1 -14.0605 constituents’ 1 -14.0605 (ldc, 1 -14.0605 embracing 1 -14.0605 randomized, 1 -14.0605 gtb. 1 -14.0605 how. 1 -14.0605 ¬causal 1 -14.0605 salt. 1 -14.0605 dfu 1 -14.0605 billot 1 -14.0605 goodresultsis thatwe 1 -14.0605 row: 1 -14.0605 subtopics, 1 -14.0605 (outside) 1 -14.0605 320,000 1 -14.0605 (ppc) 1 -14.0605 overfits 1 -14.0605 passing. 1 -14.0605 syntax; 1 -14.0605 (language-specific) 1 -14.0605 directionality. 1 -14.0605 domains (news, 1 -14.0605 norming 1 -14.0605 agg, 1 -14.0605 mb. 1 -14.0605 harsh, 1 -14.0605 ispossibletreatment(penicillin, strep 1 -14.0605 diaries. 1 -14.0605 cross-lingually, 1 -14.0605 itive 1 -14.0605 r5. 1 -14.0605 hengji@cs.qc.cuny.edu 1 -14.0605 bottleneck, 1 -14.0605 (0.14) 1 -14.0605 llcc 1 -14.0605 (asvm) 1 -14.0605 hyper-links 1 -14.0605 me-rec, 1 -14.0605 ”. 1 -14.0605 λy 1 -14.0605 pseudosentiment 1 -14.0605 (kamp, 1 -14.0605 disputant. 1 -14.0605 cyk-like 1 -14.0605 2–3 1 -14.0605 forests) 1 -14.0605 fitzgerald. 1 -14.0605 0.088 1 -14.0605 (eisner, 1996b; 1 -14.0605 red–yellow. 1 -14.0605 ]a 1 -14.0605 reject. 1 -14.0605 similarity), 1 -14.0605 m0rnin 1 -14.0605 w(s). 1 -14.0605 (f-score 1 -14.0605 sport. 1 -14.0605 giuliano 1 -14.0605 zero-order 1 -14.0605 inevitable. 1 -14.0605 ,smile’. 1 -14.0605 3.9 ipsini 1 -14.0605 (mdp). 1 -14.0605 continent, 1 -14.0605 margins. 1 -14.0605 bestsub(c) 1 -14.0605 pages 1083 1 -14.0605 editsga 1 -14.0605 finalfscores 1 -14.0605 cyk-3 1 -14.0605 elements). 1 -14.0605 gold3. 1 -14.0605 programmed 1 -14.0605 east. 1 -14.0605 ini. 1 -14.0605 claypool 1 -14.0605 §2.1, 1 -14.0605 semiautomatically 1 -14.0605 senseannotated 1 -14.0605 “the gunman 1 -14.0605 0.51 0.53 1 -14.0605 trainees 1 -14.0605 further). 1 -14.0605 log-conca 1 -14.0605 10k) 1 -14.0605 non-textual 1 -14.0605 bed. 1 -14.0605 posteriors, 1 -14.0605 available.6 1 -14.0605 part). 1 -14.0605 redistribution, 1 -14.0605 pcfg-style 1 -14.0605 dissociation, 1 -14.0605 (ips) 1 -14.0605 visualize) 1 -14.0605 afective companion 1 -14.0605 231720 (euromatrix 1 -14.0605 15,720 1 -14.0605 comlex. 1 -14.0605 †instituto 1 -14.0605 learners' 1 -14.0605 (buckley 1 -14.0605 non-singleton 1 -14.0605 aggregator 1 -14.0605 wx 1 -14.0605 78.75 0.07 1 -14.0605 boldface). 1 -14.0605 disinformers 1 -14.0605 1http://www.statmt.org/wmt10/ 1 -14.0605 m4). 1 -14.0605 segmental), 1 -14.0605 port. 1 -14.0605 slipnet. 1 -14.0605  1 -14.0605 latter). 1 -14.0605 unreachable 1 -14.0605 61.86 1 -14.0605 s32s52 s32s53 1 -14.0605 n.pron.def.pers 1 -14.0605 sense-inventories 1 -14.0605 clte, 1 -14.0605 fixes. 1 -14.0605 →p(v 1 -14.0605 ic. 1 -14.0605 per-gender 1 -14.0605 (x): 1 -14.0605 winnow’s 1 -14.0605 clusterings (views) 1 -14.0605 enumerable, 1 -14.0605 lookahead. 1 -14.0605 (oard, 1 -14.0605 .5) 1 -14.0605 lively 1 -14.0605 7-factor 1 -14.0605 martin chodorow 1 -14.0605 outdegree. 1 -14.0605 p" type="phosphorylation" 1 -14.0605 choose, 1 -14.0605 yo shidahara 1 -14.0605 (reader 1 -14.0605 precursor 1 -14.0605 cmu-portugal program, 1 -14.0605 712 1 -14.0605 θ: 1 -14.0605 belly 1 -14.0605 pos(t) ∈{nn} 1 -14.0605 (edr) 1 -14.0605 g87g396g286 g272g349g400 1 -14.0605 task2(logs2) 1 -14.0605 biography. 1 -14.0605 4885 1 -14.0605 3.5), 1 -14.0605 intuition: 1 -14.0605 iaa: 1 -14.0605 tfk: 1 -14.0605 system(s) 1 -14.0605 28.22 57.06 1 -14.0605 (rather) 1 -14.0605 ‘kerry taylor’ 1 -14.0605 1 -14.0605 post-soviet 1 -14.0605 §4). 1 -14.0605 blood pressure, 1 -14.0605 anytime. 1 -14.0605 0.44. 1 -14.0605 harmonizes 1 -14.0605 v3(p) 1 -14.0605 (1968). 1 -14.0605 what? 1 -14.0605 u(s). 1 -14.0605 copper, 1 -14.0605 lt. 1 -14.0605 trade-offs. 1 -14.0605 assessors, 1 -14.0605 summation. 1 -14.0605 tf-idf. 1 -14.0605 introduce. 1 -14.0605 tagger, constructor, 1 -14.0605 judgements) 1 -14.0605 essay, 1 -14.0605 36–40, 1 -14.0605 non-lossy 1 -14.0605 preller 1 -14.0605 13note 1 -14.0605 u =q= 1 -14.0605 gdlm. 1 -14.0605 word-dependency 1 -14.0605 hanks. 1 -14.0605 “if” 1 -14.0605 tm, 1 -14.0605 63.61 1 -14.0605 difficult”. 1 -14.0605 aspect-based 1 -14.0605 upspeak/downspeak 1 -14.0605 morphadorner. 1 -14.0605 diacritics. 1 -14.0605 analogy. 1 -14.0605 “is”. 1 -14.0605 moran, d. 1 -14.0605 descriptiveness 1 -14.0605 a1a3 1 -14.0605 (grefenstette, 1994a, 1 -14.0605 (ipeirotis 1 -14.0605 user). 1 -14.0605 hitoshi isahara. 1 -14.0605 agendas 1 -14.0605 tj 1tjwj 1 -14.0605 sanchis-trilles 1 -14.0605 esign 1 -14.0605 engineering” 1 -14.0605 (passive ats) 1 -14.0605 politician’s 1 -14.0605 ‘two 1 -14.0605 saw. 1 -14.0605 strengthening. 1 -14.0605 injured. 1 -14.0605 best-fitting 1 -14.0605 2010)]. 1 -14.0605 rearranged 1 -14.0605 federico@fbk.eu 1 -14.0605 tied. 1 -14.0605 24.33 1 -14.0605 argument-atomicity 1 -14.0605 png 1 -14.0605 berlin. 1 -14.0605 with whole-segment 1 -14.0605 prp). 1 -14.0605 bart cramer, 1 -14.0605 adequado 1 -14.0605 features.7 1 -14.0605 wsj5k 1 -14.0605 hierarchy). 1 -14.0605 v1.3 1 -14.0605 consolidation. 1 -14.0605 .1). 1 -14.0605 that-less cc) 1 -14.0605 rethink 1 -14.0605 (gray 1 -14.0605 reordered). 1 -14.0605 style: 1 -14.0605 synsetbased 1 -14.0605 word-net 1 -14.0605 400–931, 1 -14.0605 5.3.3 1 -14.0605 colin portnuf 1 -14.0605 complications. 1 -14.0605 given) 1 -14.0605 once), 1 -14.0605 1960’s. 1 -14.0605 asymmetry, 1 -14.0605 verbalisation. 1 -14.0605 pre.(%) 1 -14.0605 gests 1 -14.0605 (fleischman, 1 -14.0605 reliability) 1 -14.0605 “grew 1 -14.0605 “early 1 -14.0605 os. 1 -14.0605 mismatching 1 -14.0605 multiwords 1 -14.0605 neuromotor 1 -14.0605 syndicate 1 -14.0605 (upper part) 1 -14.0605 (/,/ 1 -14.0605 extensive, 1 -14.0605 instance-label 1 -14.0605 dublin. 1 -14.0605 nonfat 1 -14.0605 s0h2, s0l2, 1 -14.0605 twins ada 1 -14.0605 emily thomforde, 1 -14.0605 motorbike. 1 -14.0605 (goldwater, 1 -14.0605 (cathcart 1 -14.0605 game-play 1 -14.0605 [8]. 1 -14.0605 tiab 1 -14.0605 (r,c) 1 -14.0605 rationales. 1 -14.0605 indexers 1 -14.0605 summarizations 1 -14.0605 “far” 1 -14.0605 28.66 1 -14.0605 exclamations 1 -14.0605 scoping. 1 -14.0605 instantiate. 1 -14.0605 0.16). 1 -14.0605 ?y 1 -14.0605 punishment. 1 -14.0605 culturespecific (gage, 1 -14.0605 woz-08. 1 -14.0605 tasa. 1 -14.0605 communitywide 1 -14.0605 pribadi 1 -14.0605 errado 1 -14.0605 features(log1,log2) 1 -14.0605 (purity) 1 -14.0605 efficiently; 1 -14.0605 challenges? 1 -14.0605 35.3% 1 -14.0605 towers 1 -14.0605 lawsuits 1 -14.0605 seattle’sin 1 -14.0605 (levenberg 1 -14.0605 ed pronou 1 -14.0605 forestbased 1 -14.0605 relies, 1 -14.0605 (chall 1 -14.0605 proven. 1 -14.0605 s54 s56 1 -14.0605 recognition; 1 -14.0605 evaluativity. 1 -14.0605 maxent lexical(n=798) 1 -14.0605 precomputed, 1 -14.0605 52.1m 1 -14.0605 qapb(pi1,pi2) 1 -14.0605 supertags) 1 -14.0605 elec. 1 -14.0605 padova 1 -14.0605 “toner cartridges 1 -14.0605 rcv1+ep 1 -14.0605 (dramas, 1 -14.0605 burden. 1 -14.0605 awards #0917170 1 -14.0605 datapoint 1 -14.0605 unrolled 1 -14.0605 soil). 1 -14.0605 hk. 1 -14.0605 illusory 1 -14.0605 englishspanish 1 -14.0605 (i;x )2y 1 -14.0605 iit kharagpur, 1 -14.0605 ilustrate 1 -14.0605 slav@google.com 1 -14.0605 10:457–485. 1 -14.0605 highly. 1 -14.0605 usb 1 -14.0605 +det2+lmm +person+fn*ngr 1 -14.0605 et.al 1 -14.0605 8it 1 -14.0605 misidentified. 1 -14.0605 generalisability 1 -14.0605 infinite) 1 -14.0605 reli 1 -14.0605 86.1% 1 -14.0605 syntacto-semantic 1 -14.0605 (http://www.edikt.org.uk/). 1 -14.0605 dynamic-programmingbased 1 -14.0605 wordnet,propbank, 1 -14.0605 medtronic. 1 -14.0605 2815 29.9% 1 -14.0605 (fujii 1 -14.0605 stresses. 1 -14.0605 mono-words. 1 -14.0605 phoneme [msec] 1 -14.0605 halving 1 -14.0605 officers. 1 -14.0605 constraints: connectedness, 1 -14.0605 cardinals. 1 -14.0605 wmt-11. 1 -14.0605 hmeant, 1 -14.0605 rennr¨ader. 1 -14.0605 ˚j. 1 -14.0605 logistic regresion 1 -14.0605 (ljubešić 1 -14.0605 bilngual 1 -14.0605 c0. 1 -14.0605 kg882>e 1 -14.0605 presented). 1 -14.0605 ([v1,v2]). 1 -14.0605 #question 1 -14.0605 [h1,j,h1j,j 1 -14.0605 ipeirotis 1 -14.0605 0.3%. 1 -14.0605 isn’t. 1 -14.0605 (inverse 1 -14.0605 system.3 1 -14.0605 all-good 1 -14.0605 {ξ ·q(cnk−1 1 -14.0605 (blair-goldensohn 1 -14.0605 subtypologies 1 -14.0605 74.61 1 -14.0605 room’ 1 -14.0605 (4b). 1 -14.0605 (havasi 1 -14.0605 pijwprimeij. 1 -14.0605 pronounfeaturestrueif thegivennounphraseis 1 -14.0605 model.computeposterior(l); 1 -14.0605 knows. 1 -14.0605 cs.cmu.edu 1 -14.0605 sentence-final 1 -14.0605 [1..2] 1 -14.0605 woszczyna 1 -14.0605 bag. 1 -14.0605 set3. 1 -14.0605 5:4 1 -14.0605 comma+sbar+. 1 -14.0605 lm-based 1 -14.0605 vasileios hatzivassiloglou 1 -14.0605 (etts), 1 -14.0605 adjective-noun, 1 -14.0605 a vulgar 1 -14.0605 vanderheyden 1 -14.0605 (kauchak 1 -14.0605 arabic-english. 1 -14.0605 withc 1 -14.0605 concrete-abstract, 1 -14.0605 extrapolates 1 -14.0605 ‘_a_’ 1 -14.0605 360,985,492) 1 -14.0605 shelling 1 -14.0605 (-3.6, 1 -14.0605 guidelines) 1 -14.0605 copies, 1 -14.0605 s-neg. 1 -14.0605 self-referential 1 -14.0605 operationalize. 1 -14.0605 etwan 1 -14.0605 contributor, 1 -14.0605 sub-steps 1 -14.0605 lexicalized) 1 -14.0605 73–83, 1 -14.0605 automata (fsa), 1 -14.0605 48.5 unsup, 1 -14.0605 co-citations. 1 -14.0605 dinarelli 1 -14.0605 w.v. 1 -14.0605 5411 1 -14.0605 equivalent) 1 -14.0605 1443-1452. 1 -14.0605 anagrams 1 -14.0605 shaj 1 -14.0605 4.1.3. 1 -14.0605 (bionlp-st, 1 -14.0605 (pointwise 1 -14.0605 noise”. 1 -14.0605 boosting) 1 -14.0605 granted. 1 -14.0605 +hier reo 1 -14.0605 pre-social 1 -14.0605 perturbation quotient. 1 -14.0605 (faq) 1 -14.0605 ldahm 1 -14.0605 92.9%, 1 -14.0605 loop-free 1 -14.0605 wmt08n 1 -14.0605 down”. 1 -14.0605 45.31 57.22 1 -14.0605 learnability. 1 -14.0605 t-chunk 1 -14.0605 np-nplink, 1 -14.0605 joe biden, 1 -14.0605 (otc), 1 -14.0605 obj -= 1 -14.0605 .275 1 -14.0605 (1–2). 1 -14.0605 gak( g(i−1)a1 1 -14.0605 unshaded 1 -14.0605  1 -14.0605 emulating 1 -14.0605 politically 1 -14.0605 items). 1 -14.0605 dahlmeier, 1 -14.0605 somepre-clusters.it 1 -14.0605 progression, 1 -14.0605 3-fold cross-validation. 1 -14.0605 (w1,w2,n,r). 1 -14.0605 jamesp@cs.brandeis.edu 1 -14.0605 protein). 1 -14.0605 something? 1 -14.0605 foundationof 1 -14.0605 §2.2. 1 -14.0605 begging 1 -14.0605 t25 1 -14.0605 7,423 1 -14.0605 naturally-occurring 1 -14.0605 11.0%. 1 -14.0605 scisumm. 1 -14.0605 termination) 1 -14.0605 mentionranking 1 -14.0605 pitch-accent, 1 -14.0605 sum). 1 -14.0605 feasibly 1 -14.0605 parameters1) 1 -14.0605 newsire 1 -14.0605 coupled. 1 -14.0605 deveria[should] ter[have] 1 -14.0605 downloads) 1 -14.0605 (emadzadeh 1 -14.0605 important? 1 -14.0605 (qian 1 -14.0605 furthering 1 -14.0605 dbgap 1 -14.0605 o(mn). 1 -14.0605 questionnaires, 1 -14.0605 40,000, 1 -14.0605 uted 1 -14.0605 utsumi 1 -14.0605 204–215, 1 -14.0605 lange winkelstraat 1 -14.0605 writingexpert contentexpert 1 -14.0605 “fully agree” 1 -14.0605 pairs4. 1 -14.0605 score-based 1 -14.0605 sorted. 1 -14.0605 faisal mahbub 1 -14.0605 sequence kernels. 1 -14.0605 stadiumoreventvenue 1 -14.0605 usedifferentfeaturestodeterminewhetherthiskindoflink is 1 -14.0605 (frames) 1 -14.0605 nbtree, 1 -14.0605 henceforth). 1 -14.0605 betterthanusinganunifiedclassificationmodel.for 1 -14.0605 ignore. 1 -14.0605 sonner. 1 -14.0605 (4.c) 1 -14.0605 predicate2) 1 -14.0605 sentiment; 1 -14.0605 (nr) 1 -14.0605 36.95 51.79 1 -14.0605 85.22% 1 -14.0605 obviously surpasses 1 -14.0605 70.15 83.31 1 -14.0605 “michael jackson” 1 -14.0605 “resource”, 1 -14.0605 swier 1 -14.0605 (links). 1 -14.0605 distinction; 1 -14.0605 (iyer 1 -14.0605 88.9(94.4) 1 -14.0605 bias) 1 -14.0605 commonplace, 1 -14.0605 n×t 1 -14.0605 analogously.) 1 -14.0605 fashion; 1 -14.0605 progeny 1 -14.0605 houses. 1 -14.0605 (discussion 1 -14.0605 pra’s 1 -14.0605 vp2 1 -14.0605 surface-string 1 -14.0605 sha 1 -14.0605 formalisms) 1 -14.0605 shallower. 1 -14.0605 [x,y], 1 -14.0605 harper† †department 1 -14.0605 notices anorexia. 1 -14.0605 post/response 1 -14.0605 amazon. 1 -14.0605 verbalising 1 -14.0605 sci. 1 -14.0605 matched-pairs 1 -14.0605 author? 1 -14.0605 1754 1 -14.0605 45.61 1 -14.0605 pbase. 1 -14.0605 boundary). 1 -14.0605 > 10000000) 1 -14.0605 blue–green 1 -14.0605 lmax. 1 -14.0605 (0.8, 1 -14.0605 0.04). 1 -14.0605 systems” 1 -14.0605 96.70% 1 -14.0605 clusterresults. 1 -14.0605 t-test: 1 -14.0605 model.5 1 -14.0605 downhill-simplex 1 -14.0605 thisdatasetis alsotaggedwithsyntactictrees,high 1 -14.0605 91–97, 1 -14.0605 byrd 1 -14.0605 17). 1 -14.0605 parsim’s 1 -14.0605 angbracketrightbig rrbracket 1 -14.0605 presidential debate 1 -14.0605 wni∈s1,wnj∈s2 jcn(wni,wnj), 1 -14.0605 wel. 1 -14.0605 non-scopal 1 -14.0605 inferior) 1 -14.0605 convexity. 1 -14.0605 microblog. 1 -14.0605 threshold). 1 -14.0605 fricatives, affricates, 1 -14.0605 ehud reiter 1 -14.0605 indicate. 1 -14.0605 daoud@metrica.net 1 -14.0605 central torpedo 1 -14.0605 821 1 -14.0605 fired.” 1 -14.0605 turn-final 1 -14.0605 (192). 1 -14.0605 expresion. 1 -14.0605 ˇ zabokrtsk`y 1 -14.0605 [v1w1,v2w2] 1 -14.0605 moat (multilingual 1 -14.0605 (0.9) 1 -14.0605 village. 1 -14.0605 & loa-istc-cnr, 1 -14.0605 (top1, 1 -14.0605 hebrew-english 1 -14.0605 stars, 1 -14.0605 (sp 1 -14.0605 controlled...levels 1 -14.0605 unstable, 1 -14.0605 14), 1 -14.0605 ies. 1 -14.0605 !!, 1 -14.0605 (diwa) 1 -14.0605 2bed.2bath 1 -14.0605 friendlier 1 -14.0605 reranking (>10counts) 1 -14.0605 2–4 1 -14.0605 49-52. 1 -14.0605 (roi) 1 -14.0605 (dice, 1945) 1 -14.0605 priors) 1 -14.0605 detections. 1 -14.0605 visit sequence: 1 -14.0605 cabré 1 -14.0605 designs. 1 -14.0605 mistype “shenme” 1 -14.0605 debts 1 -14.0605 disclosure. 1 -14.0605 stabilize. 1 -14.0605 misspellings. 1 -14.0605 (shan 1 -14.0605 headword. 1 -14.0605 sollten sie 1 -14.0605 bngs, 1 -14.0605 1/3, 1 -14.0605 thementionshave thesameposition(beginning,middleorend)insentences 1 -14.0605 33.72 1 -14.0605 townshend. 1 -14.0605 adjacency. 1 -14.0605 substrings. 1 -14.0605 government employee’s 1 -14.0605 targeted. 1 -14.0605 (p-trans) 1 -14.0605 student-manual1 1 -14.0605 coordinator 1 -14.0605 svmrank 1 -14.0605 312.4m 1 -14.0605 whole-number 1 -14.0605 icons 1 -14.0605 blink. 1 -14.0605 3.84 1 -14.0605 nomenclatures. 1 -14.0605 “pain” 1 -14.0605 cb3 0fd, 1 -14.0605 nist08 1 -14.0605 (huffman, 1 -14.0605 adrian nastase 1 -14.0605 “mere rayee 1 -14.0605 nulll, nullp-vit(g), 1 -14.0605 individualization. 1 -14.0605 싸고 1 -14.0605 (piao 1 -14.0605 2,4 0,4 1 -14.0605 lavrenko 1 -14.0605 postnull1 1 -14.0605 diminishers 1 -14.0605 (non-core) 1 -14.0605 positively. 1 -14.0605 wlscript ∼ 1 -14.0605 hypothesesfor 1 -14.0605 aguiar‡ 1 -14.0605 10-75% 1 -14.0605 intutively, 1 -14.0605 think. 1 -14.0605 (res). 1 -14.0605 (iaa). 1 -14.0605 1/12th 1 -14.0605 random.) 1 -14.0605 {x(l),y(l)}l∈sl 1 -14.0605 71.5% 1 -14.0605 pronoun 88.4 1 -14.0605 1996)). 1 -14.0605 inefficiency 1 -14.0605 ited 1 -14.0605 θt. 1 -14.0605 intercorrelations 1 -14.0605 1.1.1 1 -14.0605 sharedtasknextbest 1 -14.0605 adafre 1 -14.0605 meet. 1 -14.0605 sys π11 1 -14.0605 thy 1 -14.0605 24,200 1 -14.0605 +simm&l 1 -14.0605 1,772 1 -14.0605 model.updateparameters(counts); 1 -14.0605 〈s,t〉, 1 -14.0605 backoffs. 1 -14.0605 list.”] 1 -14.0605 lukewarm 1 -14.0605 field-s. 1 -14.0605 tos. 1 -14.0605 immediate. 1 -14.0605 42620 1 -14.0605 stemming). 1 -14.0605 expressionevent 1 -14.0605 leaders. 1 -14.0605 senses; 1 -14.0605 generalize, 1 -14.0605 http://svmlight. 1 -14.0605 urdu-english, 1 -14.0605 http://www.gutenberg.org 1 -14.0605 missed, 1 -14.0605 dd-admm, 1 -14.0605 chinese). 1 -14.0605 coventry 1 -14.0605 88.69 1 -14.0605 facial taging 1 -14.0605 z∗(prf)q 1 -14.0605 fracture. 1 -14.0605 eleftherios avramidis, 1 -14.0605 qsbp. 1 -14.0605 (bornstein, 1 -14.0605 springboard 1 -14.0605 px, 1 -14.0605 aristotle, 1 -14.0605 (hungarian). 1 -14.0605 sd. 1 -14.0605 1965. 1 -14.0605 erases 1 -14.0605 houfengwang† 1 -14.0605 di matematica 1 -14.0605 -10 -8-6 1 -14.0605 wish. 1 -14.0605 cross-validations 1 -14.0605 {1,...,m} 1 -14.0605 bpcs. 1 -14.0605 knowledge-, corpus-, 1 -14.0605 harmonized 1 -14.0605 −.746 1 -14.0605 it”. 1 -14.0605 (giving 1 -14.0605 geo). 1 -14.0605 shown; 1 -14.0605 design: 1 -14.0605 “rank” 1 -14.0605 second). 1 -14.0605 577-585. 1 -14.0605 multi-directional 1 -14.0605 pre-clusterpaircanbeseenasa 1 -14.0605 1,030 1 -14.0605 crpt niqwla 1 -14.0605 theme, sink) 1 -14.0605 a3, 1 -14.0605 firmness 1 -14.0605 tf.idf, 1 -14.0605 veryc 1 -14.0605 misinterpret 1 -14.0605 obtaining, 1 -14.0605 d′t 1 -14.0605 b terminoontology 1 -14.0605 76.2% 1 -14.0605 w∈ri min(cnt(w,t,ri),cnt(w,t,ci)) 1 -14.0605 dispreference 1 -14.0605 additions. 1 -14.0605 jožef stefan 1 -14.0605 conciliation. 1 -14.0605 inconsistencies. 1 -14.0605 collection-member 1 -14.0605 madnani’s 1 -14.0605 durational 1 -14.0605 baselines; 1 -14.0605 min-cost 1 -14.0605 ja antaa 1 -14.0605 resource; 1 -14.0605 σnullxvotingamericannull(x) ∧∃y[y 1 -14.0605 69.06 1 -14.0605 dmax 1 -14.0605 similarity” 1 -14.0605 convention: 1 -14.0605 backchannel. 1 -14.0605 money. 1 -14.0605 lvnounobj 1 -14.0605 vsms, 1 -14.0605 unconn. 1 -14.0605 tutor). 1 -14.0605 testbed. 1 -14.0605 on/p. 1 -14.0605 2002).1 1 -14.0605 m `arquez 1 -14.0605 strapparava, giovanni 1 -14.0605 bhattacharyya2 1 -14.0605 please, 1 -14.0605 ˆpji). 1 -14.0605 (news). 1 -14.0605 track: 1 -14.0605 step-bystep 1 -14.0605 50.08 1 -14.0605 (bins 1 -14.0605 the-west”) 1 -14.0605 sie verhindern 1 -14.0605 lakoff (1973)) 1 -14.0605 verbalizer, 1 -14.0605 minimum-effort 1 -14.0605 dj pauly 1 -14.0605 ”chunking”). 1 -14.0605 riety 1 -14.0605 verify, 1 -14.0605 in-link 1 -14.0605 (2.2) 1 -14.0605 selects: 1 -14.0605 treetaggerchunker 1 -14.0605 93.8%. 1 -14.0605 (gratch 1 -14.0605 ryant, 1 -14.0605 experiencer obj 1 -14.0605 md-i. 1 -14.0605 [np:a0 i] 1 -14.0605 cumby 1 -14.0605  none: 1 -14.0605 antonyms. 1 -14.0605 bio-entity 1 -14.0605 expert-generated 1 -14.0605 habitat. 1 -14.0605 initially. 1 -14.0605 geographical. 1 -14.0605 nishimura 1 -14.0605 adjnn v- 1 -14.0605 subtree (ew 1 -14.0605 poster. 1 -14.0605 gnd. 1 -14.0605 wizards, 1 -14.0605 persists, 1 -14.0605 pmi(i, “excellent”) 1 -14.0605 631 3,725 1 -14.0605 attardi. 1 -14.0605 (1981). 1 -14.0605 experiment1. 1 -14.0605 tetreaultchunk 1 -14.0605 (+1.47 1 -14.0605 structures), 1 -14.0605 (das) 1 -14.0605 safe. 1 -14.0605 verb/argument 1 -14.0605 budget-driven shrinkage 1 -14.0605 70.65 70.88 1 -14.0605 75.54 1 -14.0605 enhanced, 1 -14.0605 mapping) 1 -14.0605 randomisation 1 -14.0605 modular. 1 -14.0605 ibm-1, 1 -14.0605 cramer 1 -14.0605 scfghmd 1 -14.0605 rm. 1 -14.0605 argument-finding 1 -14.0605 gaing(eprime): 1 -14.0605 unmatch_count 1 -14.0605 w∈rny×nx bardblmy 1 -14.0605 mg2, 1 -14.0605 (flight,[f light]) 1 -14.0605 (rlm1, rlm2, 1 -14.0605 kalita, 1 -14.0605 (sender, recipient) 1 -14.0605 co-occuring 1 -14.0605 ouro preto. 1 -14.0605 18%. 1 -14.0605 84.5%. 1 -14.0605 77.91 1 -14.0605 spelling-based 1 -14.0605 (1969), 1 -14.0605 skewed: 1 -14.0605 boy.” 1 -14.0605 propagated. 1 -14.0605 gero uste 1 -14.0605 u(r), 1 -14.0605 nulex 1 -14.0605 (gsc) 1 -14.0605 “junk” 1 -14.0605 effect’ 1 -14.0605 5-8). 1 -14.0605 unobservable 1 -14.0605 maamouri 1 -14.0605 leaked 1 -14.0605 dodi 5230.29, 1 -14.0605 anexample 1 -14.0605 train/evaluation 1 -14.0605 wl∈v #(wl,dk) 1 -14.0605 pos-tags; 1 -14.0605 “listens” missing 1 -14.0605 (‘colorings’) 1 -14.0605 parliamentary tomato, 1 -14.0605 beam-width. 1 -14.0605 19940512.0003 1 -14.0605 presupposed 1 -14.0605 (wehrli 1 -14.0605 canon g3 1 -14.0605 background). 1 -14.0605 iteration’s 1 -14.0605 saudi arabia 1 -14.0605 star. 1 -14.0605 election] 1 -14.0605 (cqa) 1 -14.0605 relief-centric 1 -14.0605 fusion: 1 -14.0605 ranking; 1 -14.0605 1st sf+ss_v+wsf_n 1 -14.0605 (mbsp) 1 -14.0605 brinton 1 -14.0605 similarities). 1 -14.0605 others”. 1 -14.0605 uzcrs 1 -14.0605 korea (nrf) 1 -14.0605 counting. 1 -14.0605 239 vandoeuvre-l`es-nancy, 1 -14.0605 web) 1 -14.0605 npcomplete. 1 -14.0605 to-string 1 -14.0605 synsets). 1 -14.0605 visited, 1 -14.0605 65.72 1 -14.0605 filtered). 1 -14.0605 non-dialogue 1 -14.0605 enron.com 1 -14.0605 stanford parser6 1 -14.0605 / 1 -14.0605 restrain 1 -14.0605 thyme. 1 -14.0605 travel-planning 1 -14.0605 & manning., 1 -14.0605 scope3 1 -14.0605 onr muri 1 -14.0605 1948). 1 -14.0605 78.79% 1 -14.0605 2,847 1 -14.0605 (far 1 -14.0605 prerequisites. 1 -14.0605 ulrich reffle, 1 -14.0605 lch. 1 -14.0605 (2006a). 1 -14.0605 aljoscha 1 -14.0605 2.2.2, 1 -14.0605 been. 1 -14.0605 (fr. 1 -14.0605 balanced topic] 1 -14.0605 forbidden, 1 -14.0605 wessner, 1 -14.0605 jansche 1 -14.0605 formaliser (tem-full+f) 1 -14.0605 keypresses 1 -14.0605 “transition”, 1 -14.0605 appearing, 1 -14.0605 213250). 1 -14.0605 .08) 1 -14.0605 (awn) (elkateb, 1 -14.0605 mixtures. 1 -14.0605 bursty 1 -14.0605 subject/object, 1 -14.0605 silke.scheible, 1 -14.0605 sio, 1 -14.0605 s3 britain 1 -14.0605 twintig jar. 1 -14.0605 signicantly 1 -14.0605 +epsilon1. 1 -14.0605 fǎlǜ 1 -14.0605 gold.2 1 -14.0605 picchi, 1 -14.0605 1997)) 1 -14.0605 45.62 1 -14.0605 f(ynull). 1 -14.0605 ids. 1 -14.0605 1...n} 1 -14.0605 “x2:在” 1 -14.0605 (rebuttals). 1 -14.0605 hypotheses) 1 -14.0605 faculty, 1 -14.0605 sie. 1 -14.0605 lowercase/capitalized 1 -14.0605 citation, 1 -14.0605 antisymmetric. 1 -14.0605 maximum-a-posterior 1 -14.0605 (discretized) 1 -14.0605 synonyms). 1 -14.0605 αi +βj 1 -14.0605 best-ranked 1 -14.0605 1/9 1 -14.0605 non-relevant. 1 -14.0605 carport, 1 -14.0605 (too  general.) 1 -14.0605 expensive) 1 -14.0605 gold-pos pred-pos 1 -14.0605 “no”). 1 -14.0605 inter-lingual 1 -14.0605 9.70 1 -14.0605 “words” 1 -14.0605 -.91 1 -14.0605 objectively, 1 -14.0605 affectid 1 -14.0605 anaphorically. 1 -14.0605 clauses (szpakowicz 1 -14.0605 ideological. 1 -14.0605 {danielhe,nght}@comp.nus.edu.sg 1 -14.0605 maximalnp: 1 -14.0605 92.13% 1 -14.0605 p)). 1 -14.0605 32gb 1 -14.0605 ho ‘be’ 1 -14.0605 dificult. 1 -14.0605 65%, 1 -14.0605 reordering-as-preprocessing 1 -14.0605 {act com} 1 -14.0605 modifications (negation 1 -14.0605 okurowski. 1 -14.0605 (mahmoud, 1 -14.0605 nonpolar 1 -14.0605 100-word-limited 1 -14.0605 '10). 1 -14.0605 scus. 1 -14.0605 “dependency 1 -14.0605 reproducibility. 1 -14.0605 non-overlapped occurrences;lscriptdocument 1 -14.0605 (1 mm) 1 -14.0605 o(mn2), 1 -14.0605 ablation, 1 -14.0605 dates) 1 -14.0605 expressiveness. 1 -14.0605 moses+del. 1 -14.0605 lagrangian. 1 -14.0605 demonstrated, 1 -14.0605 nakatani, 1 -14.0605 (dos santos 1 -14.0605 =0, 1 -14.0605 statisticsbased 1 -14.0605 inter-language 1 -14.0605 usp-bayes 1 -14.0605 kilometers. 1 -14.0605 efficiency: 1 -14.0605 (gold/auto), 1 -14.0605 theme, cause, 1 -14.0605 lattice-like 1 -14.0605 pathway-to-event 1 -14.0605 [predicate cooked] 1 -14.0605 disposes 1 -14.0605 kdd. 1 -14.0605 ·minparenleftbige1−rprime 1 -14.0605 life), 1 -14.0605 (leopold 1 -14.0605 .517 1 -14.0605 pre-translation 〈t,u〉 1 -14.0605 triage. 1 -14.0605 cu,θ. 1 -14.0605 discovered) 1 -14.0605 thesauri, 1 -14.0605 {rise:1, lift:4, 1 -14.0605 a29a36a41 a29a36a42 1 -14.0605 traders. 1 -14.0605 hvsbioevent 1 -14.0605 antal.vdnbosch@uvt.nl 1 -14.0605 helpful; 1 -14.0605 invertebrate. 1 -14.0605 item-based 1 -14.0605 filimonov†‡ 1 -14.0605 innate. 1 -14.0605 novak 1 -14.0605 task1(logs2) 1 -14.0605 dicto: 1 -14.0605 k: 1 -14.0605 zq. 1 -14.0605 mostfalse, 1 -14.0605 super-entity 1 -14.0605 paiwan /seti/, 1 -14.0605 c-parameter 1 -14.0605 !"# 1 -14.0605 subactions. 1 -14.0605 league. 1 -14.0605 marginal, 1 -14.0605 genia-medco 1 -14.0605 correlation). 1 -14.0605 additive. 1 -14.0605 spells 1 -14.0605 ˥ 1 -14.0605 backtracking, 1 -14.0605 pscbased 1 -14.0605 (briscoe, 1 -14.0605 green df(6): 1 -14.0605 downsample 1 -14.0605 i(npj, 1 -14.0605 ’cow’ >>> 1 -14.0605 character). 1 -14.0605 careless 1 -14.0605 everybody loved 1 -14.0605 tandem. 1 -14.0605 ashburner 1 -14.0605 disgust. 1 -14.0605 never, n’t, 1 -14.0605 liegl, 1 -14.0605 boring exciting 1 -14.0605 bbn’s 1 -14.0605 gfs, 1 -14.0605 θne+de, 1 -14.0605 word-graph 1 -14.0605 make2. 1 -14.0605 cuisine, 1 -14.0605 w 1 -14.0605 today...?” 1 -14.0605 steadily. 1 -14.0605 (s-neut). 1 -14.0605 user-generated contents: 1 -14.0605 non-entailment 1 -14.0605 scfg-hmd 41.31 1 -14.0605 relatedness, commoness 1 -14.0605 gratitude. 1 -14.0605 9,11 10,12 1 -14.0605 1972. 1 -14.0605 pos). 1 -14.0605 reactants, 1 -14.0605 masi-ess-ta. 1 -14.0605 iwsd; pprecision,r- 1 -14.0605 1986; brennan 1 -14.0605 upperbound 4.80 1 -14.0605 frequenc 1 -14.0605 singletonpre-cluster, 1 -14.0605 problem,and 1 -14.0605 (step (3)) 1 -14.0605 ∆2 1 -14.0605 nmod pmod 1 -14.0605 zeuch (in 1 -14.0605 bandwidth 1 -14.0605 chrisq@microsoft.com 1 -14.0605 construction-focused 1 -14.0605 1338 1 -14.0605 premise-hypothesis 1 -14.0605 node@7277 1 -14.0605 (maslennikov 1 -14.0605 35.83 1 -14.0605 (deruyter 1 -14.0605 75080 1 -14.0605 dtd. 1 -14.0605 emotions? 1 -14.0605 (199). 1 -14.0605 (coling/acl-06). 1 -14.0605 (gap+) [third-pron] 1 -14.0605 “true siding” 1 -14.0605 conceal 1 -14.0605 gram. 1 -14.0605 election is-anullnullnull 1 -14.0605 domain-independent. 1 -14.0605 `a l’expiration 1 -14.0605 conference) 1 -14.0605 adverbs). 1 -14.0605 preposition” 1 -14.0605 browses newspaper”, 1 -14.0605 giza-strong). 1 -14.0605 §3.2. 1 -14.0605 miner 1 -14.0605 atis. 1 -14.0605 0.55. 1 -14.0605 a69a254a92a115a3a115 a88a65a121a110a65 1 -14.0605 š. 1 -14.0605 platt, 1 -14.0605 1862 1 -14.0605 www.gutenberg.org/ebooks/10681 1 -14.0605 idiomatic/natural, 1 -14.0605 contentwindow2, 1 -14.0605 favoured 1 -14.0605 dot. 1 -14.0605 814–824, 1 -14.0605 v, n.denot, 1 -14.0605 englishand 1 -14.0605 word-pair. 1 -14.0605 re-group 1 -14.0605 tonality 1 -14.0605 proto-patient 1 -14.0605 dissociates 1 -14.0605 expectation semiring, 1 -14.0605 (na 1 -14.0605 4one 1 -14.0605 oct. 1 -14.0605 crane 1 -14.0605 bolh. 1 -14.0605 (a–c) 1 -14.0605 52.74 1 -14.0605 givennounphrase. 1 -14.0605 haussler 1 -14.0605 (cieri 1 -14.0605 (rbmt) 1 -14.0605 rules1 1 -14.0605 dep(?,tj). 1 -14.0605 (ide, 1 -14.0605 86.9%. 1 -14.0605 ballpark” 1 -14.0605 rp. 1 -14.0605 austria). 1 -14.0605 tarau. 1 -14.0605 elizabeth boschee, 1 -14.0605 links.) 1 -14.0605 (+gw) 1 -14.0605 hyper-parameters, 1 -14.0605 bee 1 -14.0605 1ifana 1 -14.0605 dirichlet-multinomial 1 -14.0605 condensates 1 -14.0605 debug 1 -14.0605 59.86 1 -14.0605 nations (un) 1 -14.0605 allocation, overseas 1 -14.0605 40m 1 -14.0605 hatwidenullnull 1 -14.0605 rule). 1 -14.0605 pre–verbal 1 -14.0605 (ns) 1 -14.0605 (kn) 1 -14.0605 andcj. 1 -14.0605 non-sentential 1 -14.0605 stone cutter 1 -14.0605 trim” 1 -14.0605 task2 1 -14.0605 pre-test −.46 1 -14.0605 module’s 1 -14.0605 responses; 1 -14.0605 sub-graphs. 1 -14.0605 medications 1 -14.0605 79–86, 1 -14.0605 0.6435). 1 -14.0605 4th. 1 -14.0605 nesson 1 -14.0605 maximal. 1 -14.0605 europarl: 1 -14.0605 snapshot. 1 -14.0605 (extraction :type 1 -14.0605 useci 1 -14.0605 sprimen 1 -14.0605 wordain 1 -14.0605 1977. 1 -14.0605 43%. 1 -14.0605 eminem 1 -14.0605 bionlp-st. 1 -14.0605 (46%). 1 -14.0605 p(sprimei) 1 -14.0605 zzz 1 -14.0605 wu. 1 -14.0605 (sdp), 1 -14.0605 null(wnullnullt) 1 -14.0605 set1. 1 -14.0605 ldc2006e34, ldc2006e85, 1 -14.0605 (rambow, 1 -14.0605 ⊂yprime. 1 -14.0605 notexistential 1 -14.0605 lemmatization. 1 -14.0605 prime ministers 1 -14.0605 pb(w). 1 -14.0605 sampo pyysalo, 1 -14.0605 0.00.10.20.30.40.5 1 -14.0605 exercised (e14) 1 -14.0605 ter-plus. 1 -14.0605  c1s1o1 1 -14.0605 chalenge. 1 -14.0605 host, hostpart, 1 -14.0605 n[2]. 1 -14.0605 804–813, 1 -14.0605 yan qu, 1 -14.0605 ‘redundant’ 1 -14.0605 runs throughe1 1 -14.0605 41–45, 1 -14.0605 bestfirst 1 -14.0605 51.78 1 -14.0605 73,000 1 -14.0605 diagrams. 1 -14.0605 yves saint 1 -14.0605 homogeneous; completeness 1 -14.0605 (bounded 1 -14.0605 diag 1 -14.0605 §3.3 1 -14.0605 primaryargumentsof 1 -14.0605 (good). 1 -14.0605 don tuggener 1 -14.0605 fuzzification. 1 -14.0605 +det+png 1 -14.0605 perdocument 1 -14.0605 (5)). 1 -14.0605 unknowns. 1 -14.0605 inadequately 1 -14.0605 5–9), 1 -14.0605 toolkit1 1 -14.0605 attop 1 -14.0605 chua. 1 -14.0605 “sowohl... 1 -14.0605 metaphorically, 1 -14.0605 usa) 1 -14.0605 destinations. 1 -14.0605 (makhoul 1 -14.0605 t7: 1 -14.0605 30.89 1 -14.0605 justin bieber 1 -14.0605 twitterbased 1 -14.0605 “(definer” 1 -14.0605 yakovleva 1 -14.0605 albercht 1 -14.0605 152–160, 1 -14.0605 e-j. 1 -14.0605 [placing]. 1 -14.0605 andbaldwin,1998)andceaf(luo,2005). 1 -14.0605 µu−νqi 1 -14.0605 (mturk). 1 -14.0605 lscript1-regularization 1 -14.0605 circle. 1 -14.0605 intension 1 -14.0605 82.84%. 1 -14.0605 gw3 1 -14.0605 1617 1 -14.0605 geo:population. 1 -14.0605 substates 1 -14.0605 (brow 1 -14.0605 39.67 1 -14.0605 nomen 1 -14.0605 39. 1 -14.0605 chung yong 1 -14.0605 horn-clause 1 -14.0605 (burgoon 1 -14.0605 p(r3 1 -14.0605 whichis biasedtoward 1 -14.0605 αh, 1 -14.0605 r∈r(wm,wc) λr 1 -14.0605 distilling 1 -14.0605 wiley-blackwel. 1 -14.0605 answered; 1 -14.0605 bullets. 1 -14.0605 (score; 1 -14.0605 suominen 1 -14.0605 b0(s). 1 -14.0605 (c1)ii 1 -14.0605 tbl. 1 -14.0605 r=0.2. 1 -14.0605 iis-0812261. 1 -14.0605 43.84 1 -14.0605 connotations 1 -14.0605 sussna, 1 -14.0605 nnlms 1 -14.0605 tables,wecanseethatthescorescanbeimprovedlitt- 1 -14.0605 (ontology 1 -14.0605 null” 1 -14.0605 lightning rod 1 -14.0605 (ppis) 1 -14.0605 vb, sing] 1 -14.0605 delicious)” 1 -14.0605 w1...wn, 1 -14.0605 insight. 1 -14.0605 userproposedusages 1 -14.0605 42.61 1 -14.0605 [@ 1 -14.0605 stephen.clark@cl.cam.ac.uk 1 -14.0605 tal-2, lyl-1 1 -14.0605 speling correction 1 -14.0605 language-pair 1 -14.0605 (che 1 -14.0605 enigmatic 1 -14.0605 saralegi 1 -14.0605 taihen 1 -14.0605 prevalent, 1 -14.0605 atach 1 -14.0605 protein-related 1 -14.0605 97.78 1 -14.0605 relationbased 1 -14.0605 top5000 1 -14.0605 (100k 1 -14.0605 vex 1 -14.0605 sections.) 1 -14.0605 (wiebe, 1 -14.0605 ⁄ 1 -14.0605 barier” 1 -14.0605 /∈ {0,n} 1 -14.0605 p1s5: "medtronic" 1 -14.0605 one-sense-per-category 1 -14.0605 greedy. 1 -14.0605 9as 1 -14.0605 trade. 1 -14.0605 sentiments: 1 -14.0605 auch” 1 -14.0605 unmanageable. 1 -14.0605 skuebler@indiana.edu 1 -14.0605 order-of-magnitude 1 -14.0605 r=recall, 1 -14.0605 superior) 1 -14.0605 totrtale, 1 -14.0605 bcs-0904822, bcs-0624277, 1 -14.0605 language-unaware 1 -14.0605 seniorenr¨ader. 1 -14.0605 unanswered, 1 -14.0605 noam ordan 1 -14.0605 unresolved. 1 -14.0605 (∑ j∈t 1 -14.0605 max-rule-score. 1 -14.0605 +hits +v1 1 -14.0605 (ha*) 1 -14.0605 synecdoche 1 -14.0605 offer. 1 -14.0605 coreferentialresults, 1 -14.0605 28.15 1 -14.0605 durrell 1 -14.0605 (bifet 1 -14.0605 dw 1 -14.0605 (toefl). 1 -14.0605 usage; 1 -14.0605 redundancies. 1 -14.0605 non-idiomatic 1 -14.0605 relapse, 1 -14.0605 attractiveness 1 -14.0605 masashi shimbo 1 -14.0605 traditionallearningbased 1 -14.0605 “production” 1 -14.0605 capitals 1 -14.0605 database4 1 -14.0605 car? 1 -14.0605 codes; 1 -14.0605 charniak, 2000)’s 1 -14.0605 prepositional complement, 1 -14.0605 1900, omaha 1 -14.0605 2ghz 1 -14.0605 example? 1 -14.0605 θ(p1...pk) 1 -14.0605 ={1,2} 1 -14.0605 reproduce. 1 -14.0605 ramat gan, 1 -14.0605 nodecoalesce(s,q,q 1 -14.0605 sister. 1 -14.0605 [[[[n no] 1 -14.0605 describled 1 -14.0605 benefited. 1 -14.0605 virar enra 1 -14.0605 numerals, 1 -14.0605 flavors. 1 -14.0605 norel, 1 -14.0605 (s+p=total) 1 -14.0605 breakthroughs 1 -14.0605 .81, 1 -14.0605 counter-argument) 1 -14.0605 someone. 1 -14.0605 48.0% 1 -14.0605 collapsing’s 1 -14.0605 see! 1 -14.0605 ies, 3t, 1 -14.0605 them.2 1 -14.0605 rapid-response 1 -14.0605 bcubed 61.05% 1 -14.0605 third edition. 1 -14.0605 constructionare 1 -14.0605 summary] 1 -14.0605 nonprojectivity 1 -14.0605 nwords 1 -14.0605 play-by-play 1 -14.0605 whetertwo 1 -14.0605 kdd cup 1 -14.0605 (volk, 1 -14.0605 googleweb. 1 -14.0605 more-complex 1 -14.0605 (clia) 1 -14.0605 (7,154 1 -14.0605 satelite 1 -14.0605 91.86 1 -14.0605 word-to-sentence 1 -14.0605 reranker (es+pc+epr) 1 -14.0605 generalizer gcd. 1 -14.0605 dependency) 1 -14.0605 22-24, 1 -14.0605 not∇2,∇1 ne∇2 1 -14.0605 seco 1 -14.0605 −0.6 1 -14.0605 “better”. 1 -14.0605 ur-en. 1 -14.0605 english-tamil. 1 -14.0605 208) 1 -14.0605 1469 1 -14.0605 scanned. 1 -14.0605 best-performed 1 -14.0605 λs.λx.λy.λdi.λp.u1(x,y,di,nullp,s) 1 -14.0605 eachlink 1 -14.0605 handy 1 -14.0605 bama databases 1 -14.0605 [n], 1 -14.0605 18,941 1 -14.0605 optimized-diag. 1 -14.0605 serially 1 -14.0605 sergio horna 1 -14.0605 d.2 1 -14.0605 (muthukrishnan 1 -14.0605 enough). 1 -14.0605 trap. 1 -14.0605 1609 1 -14.0605 9%. 1 -14.0605 deferred 1 -14.0605 ��stlhqviz 1 -14.0605 stress. 1 -14.0605 null3. 1 -14.0605 single-argument 1 -14.0605 sub-classes. 1 -14.0605 likelier 1 -14.0605 factorings. 1 -14.0605 “just” 1 -14.0605 indb. 1 -14.0605 (sudo 1 -14.0605 4∼6. 1 -14.0605 nlg. 1 -14.0605 mead, 1965; 1 -14.0605 redo 1 -14.0605 0.73. 1 -14.0605  person  breads 1 -14.0605 onent 1 -14.0605 ‘assume 1 -14.0605 applythis 1 -14.0605 kernel). 1 -14.0605 thoroughly. 1 -14.0605 (qtb) (judge 1 -14.0605 chat lingo 1 -14.0605 avenue south, 1 -14.0605 approximated, 1 -14.0605 apriori. 1 -14.0605 dhillon 1 -14.0605 stacked. 1 -14.0605 .o. 1 -14.0605 externally. 1 -14.0605 terminator 1 -14.0605 backward. 1 -14.0605 [am-mod doivent] 1 -14.0605 1035 1 -14.0605 (synchronous 1 -14.0605 (ptsg) 1 -14.0605 hold-out validation 1 -14.0605 error-detection 1 -14.0605 〈z1,...,zk〉 1 -14.0605 amanda stent. 1 -14.0605 true). 1 -14.0605 knlu 1 -14.0605 (shannon, 1948) 1 -14.0605 content(nav instruction) 1 -14.0605 (greub, 1 -14.0605 conceptualised 1 -14.0605 twophase 1 -14.0605 mwe; 1 -14.0605 85721 {dhewlett,cohen}@cs.arizona.edu 1 -14.0605 ‘gold standard’ 1 -14.0605 waveforms. 1 -14.0605 inaugurated 1 -14.0605 (436 1 -14.0605 adrien lardilleux 1 -14.0605 gwb20 (22.8gb), 1 -14.0605 senses+words 1 -14.0605 direct. 1 -14.0605 words.4 1 -14.0605 prosperity 1 -14.0605 svo. 1 -14.0605 dislikes 1 -14.0605 act” 1 -14.0605 art” 1 -14.0605 1519 1 -14.0605 (raw 1 -14.0605 thementionsin a 1 -14.0605 steiner-tree 1 -14.0605 stenosis. 1 -14.0605 1,2,...,m, 1 -14.0605 pcs. 1 -14.0605 {shujin.jwn.1, 1 -14.0605 rule-types 1 -14.0605 these? 1 -14.0605 1 -14.0605 shift’s, 1 -14.0605 (khalilov 1 -14.0605 (s3 1 -14.0605 word-learning 1 -14.0605 sub-word. 1 -14.0605 hawaii. 1 -14.0605 texts-hypotheses 1 -14.0605 textually. 1 -14.0605 (mari˜no 1 -14.0605 pigs 1 -14.0605  y`m 1 -14.0605 (latech 1 -14.0605 “how” 1 -14.0605 authoring. 1 -14.0605 markables) 1 -14.0605 φ(b,h,x) 1 -14.0605 modelbasedon 1 -14.0605 31.86 1 -14.0605 convincingly 1 -14.0605 seattle zoo”, 1 -14.0605 franklin, 1 -14.0605 machine). 1 -14.0605 perspective). 1 -14.0605 contraints 1 -14.0605 67.83 1 -14.0605 ,ngramm 1 -14.0605 mighthavedifferentfeaturepreferences.sowetrain 1 -14.0605 .281 1 -14.0605 (dag). 1 -14.0605 (+0.55) 1 -14.0605 useles 1 -14.0605 disfluencies) 1 -14.0605 non-ironic 1 -14.0605 empathy 1 -14.0605 {c1,c2} 1 -14.0605 1–8. 1 -14.0605 (gui) 1 -14.0605 jj. 1 -14.0605 distractors’ 1 -14.0605 ˆtb 1 -14.0605 49.58% 1 -14.0605 quantitative, 1 -14.0605 sentences=sentencesplitting(t) 1 -14.0605 phoneme’s 1 -14.0605 reference-based 1 -14.0605 150,0001999: 1 -14.0605 [while] 1 -14.0605 .571 1 -14.0605 provisional 1 -14.0605 smu-sis (gottipati 1 -14.0605 3.1.1. 1 -14.0605 flags. 1 -14.0605 (iswc 1 -14.0605 sentences, paragraphs) 1 -14.0605 6-fold 1 -14.0605 density–based 1 -14.0605 independencies 1 -14.0605 99.7 1 -14.0605 schutze, 1 -14.0605 telephony. 1 -14.0605 location(s) 1 -14.0605 orthofeatures 1 -14.0605 the? 1 -14.0605 work.) 1 -14.0605 ecross. 1 -14.0605 srls 1 -14.0605 (shaywitz, 1 -14.0605 entailment) 1 -14.0605 (p1 +p2 1 -14.0605 “reverse” 1 -14.0605 ,.ie 1 -14.0605 pre-coreference 1 -14.0605 size”. 1 -14.0605 intra-relation 1 -14.0605 tracked, 1 -14.0605 *the 1 -14.0605 “iob” 1 -14.0605 28.69 56.21 1 -14.0605 3” 1 -14.0605 parse-augmentedword 1 -14.0605 slots; 1 -14.0605 nonsensical, 1 -14.0605 0.28. 1 -14.0605 measurable) 1 -14.0605 30.45% 36.8% 1 -14.0605 -ieren), 1 -14.0605 dencies 1 -14.0605 6.3) 1 -14.0605 economy. 1 -14.0605 trigger: 1 -14.0605 attempted, 1 -14.0605 stimulus: 1 -14.0605 bryan rink 1 -14.0605 (sfst). 1 -14.0605 non-agreeing 1 -14.0605 “relevant” 1 -14.0605 0.364. 1 -14.0605 “lexical” 1 -14.0605 (g). 1 -14.0605 216–226, 1 -14.0605 adv.denot.grad.neg 1 -14.0605 pestian 1 -14.0605 ml. 1 -14.0605 non-recursive 1 -14.0605 stars? 1 -14.0605 spread. 1 -14.0605 twet. 1 -14.0605 89.4 (77.7) 1 -14.0605 2. “shane 1 -14.0605 50.78 1 -14.0605 (26% 1 -14.0605 sourcetarget 1 -14.0605 no.91024009) 1 -14.0605 srl). 1 -14.0605 include; 1 -14.0605 w(i→j) 1 -14.0605 mtv unplugged 1 -14.0605 vacation). 1 -14.0605 eds. 1 -14.0605 fromnamesandnominalmentions,becausetheycan 1 -14.0605 poets 1 -14.0605 a1a2. 1 -14.0605 riester, 1 -14.0605 mode: 1 -14.0605 tine-srl-match 1 -14.0605 glues 1 -14.0605 purple. 1 -14.0605 property). 1 -14.0605 turn-level, 1 -14.0605 91.05% 50.67 1 -14.0605 λ1,λ2,λ3 1 -14.0605 br :={at(r,n2),b(o,at(r,n2)), 1 -14.0605 grafting (perkins 1 -14.0605 state-space, 1 -14.0605 (noreen, 1989; 1 -14.0605 82.72 1 -14.0605 hunspell1 1 -14.0605 hypotheses1 1 -14.0605 explode). 1 -14.0605 95.2% 1 -14.0605 zmert. 1 -14.0605 strict, 1 -14.0605 objects) 1 -14.0605 camera!), 1 -14.0605 bricht brechen 1 -14.0605 (mwes), 1 -14.0605 mabdulma@indiana.edu 1 -14.0605 guillaumin 1 -14.0605 isocat. 1 -14.0605 mcclosky. 1 -14.0605 domain? 1 -14.0605 election]. 1 -14.0605 converged {return 1 -14.0605 ={s 1 -14.0605 high-content 1 -14.0605 microcluster vlc/campus 1 -14.0605 †key 1 -14.0605 neurons 1 -14.0605 nullnull1nullnullnullnullnullnullqnull. 1 -14.0605 greece +digital 1 -14.0605 98005, 1 -14.0605 fellbaum’s 1 -14.0605 pre-grouping 1 -14.0605 91.8%. 1 -14.0605 pseudo-documents, 1 -14.0605 (“good”, “intermediate”, 1 -14.0605 tokenized, postagged 1 -14.0605 metagrammars. 1 -14.0605 duluth-2, 1 -14.0605 cx 1 -14.0605 (shape, 1 -14.0605 6(3-4):169-200. 1 -14.0605 diabetes mellitus 1 -14.0605 (api) 1 -14.0605 47%. 1 -14.0605 46.0% 1 -14.0605 hjalmarsson, 1 -14.0605 o(nm2) 1 -14.0605 o(log(1δ)) 1 -14.0605 parenleftbigglen(a,b) 2d 1 -14.0605 deverbial 1 -14.0605 levenshteindistance levenshteindistancebetweenthementions 1 -14.0605 “statement” 1 -14.0605 treetq 1 -14.0605 59.95%, 1 -14.0605 bm smt 1 -14.0605 no pets 1 -14.0605 ¯em1 1 -14.0605 wins. 1 -14.0605 http://lt3.hogent.be/ 1 -14.0605 network), 1 -14.0605 topically related, 1 -14.0605 perceptron; 1 -14.0605 craigslist 1 -14.0605 (sakai, 2007a). 1 -14.0605 uncommonly 1 -14.0605 same-sex 1 -14.0605 topic-unit 1 -14.0605 weaker, 1 -14.0605 51.14 1 -14.0605 familiarized 1 -14.0605 prot(x). 1 -14.0605 α; 1 -14.0605 decoding) 1 -14.0605 32.72 1 -14.0605 c&c; 1 -14.0605 news). 1 -14.0605 corpus.9 1 -14.0605 expression; 1 -14.0605 argmaxy∈yf(y). 1 -14.0605 1≤i≤n} 1 -14.0605 chi. 1 -14.0605 deux chambres 1 -14.0605 durme1,2 1department 1 -14.0605 meal. 1 -14.0605 underly 1 -14.0605 ruch 1 -14.0605 2without 1 -14.0605 rino. 1 -14.0605 2,747 1 -14.0605 8,300 1 -14.0605 middle-east 1 -14.0605 caching. 1 -14.0605 hood. 1 -14.0605 semantic. 1 -14.0605 “replacement” 1 -14.0605 door’. 1 -14.0605 alrjal ( 1 -14.0605 joseph rosenzweig. 1 -14.0605 90.62 1 -14.0605 inquirer lexicon4 1 -14.0605 tu egiten 1 -14.0605 afr 1 -14.0605 adept, 1 -14.0605 restaurant). 1 -14.0605 1379 1 -14.0605 ik(o,at(r,n2)),ik(o,at(r,n3)), b(o,ik(o,at(r,n1))),b(o,ik(o,at(r,n2))), 1 -14.0605 drive. 1 -14.0605 mittelfeld. 1 -14.0605 gold/gold 1 -14.0605 baltimore, md, 1 -14.0605 spellings. 1 -14.0605 midpoints 1 -14.0605 english-tojapanese 1 -14.0605 (testing) 1 -14.0605 cuny, 1 -14.0605 1.5gb 1 -14.0605 transmitting 1 -14.0605 madkour 1 -14.0605 0 1 -14.0605 setup; 1 -14.0605 zukunftsfonds 1 -14.0605 possessives. 1 -14.0605 18) 1 -14.0605 aaai bica 1 -14.0605 40.27 44.21 1 -14.0605 plane, spacecraft, 1 -14.0605 keet 1 -14.0605 selker. 1 -14.0605 handheld 1 -14.0605 partiality: 1 -14.0605 real callers, 1 -14.0605 costsensitive 1 -14.0605 “fair 1 -14.0605 3: burmah 1 -14.0605 n=20, 1 -14.0605 selectedgeneratedsentenceas 1 -14.0605 reliably, 1 -14.0605 bulgarian) 1 -14.0605 dang. 1 -14.0605 ontext. 1 -14.0605 unoptimized 1 -14.0605 rotten tomatoes 1 -14.0605 so-pmi 1 -14.0605 iargn. 1 -14.0605 valencies 1 -14.0605 s6. 1 -14.0605 guessing. 1 -14.0605 disputants: 1 -14.0605 tool3 1 -14.0605 sentence-semantics 1 -14.0605 eq.(14) 1 -14.0605 well-known adage, 1 -14.0605 they? 1 -14.0605 79.6% 1 -14.0605 (raaijmakers 1 -14.0605 necesitated 1 -14.0605 deana 1 -14.0605 det (determiner), 1 -14.0605 login 1 -14.0605 arepronouns;elsen 1 -14.0605 people”. 1 -14.0605 reclaim 1 -14.0605 symbolises 1 -14.0605 collier 1 -14.0605 o(nlog(n)) 1 -14.0605 90. 1 -14.0605 joining, 1 -14.0605 instruments. 1 -14.0605 kl fl* 1 -14.0605 61.78 1 -14.0605 (0.18) 1 -14.0605 seeming 1 -14.0605 unreservedly 1 -14.0605 scarcity, 1 -14.0605 higher). 1 -14.0605 michael ringgaard 1 -14.0605 hand-tuning 1 -14.0605 authentic. 1 -14.0605 96.51 1 -14.0605 blocking low-quality 1 -14.0605 mbr-sc-e/conjoin. 1 -14.0605 c1...cn. 1 -14.0605 cons. 1 -14.0605 grief 1 -14.0605 interests) 1 -14.0605 ascending factorials 1 -14.0605 pattvsm, 1 -14.0605 q. zhu. 1 -14.0605 58.78 1 -14.0605 yajuan l¨u 1 -14.0605 500k, 1 -14.0605 “to eat”) 1 -14.0605 an) 1 -14.0605 kaminski, 1 -14.0605 vector). 1 -14.0605 duchier 1 -14.0605 pman nman 1 -14.0605 (oot) 1 -14.0605 new, unseen, 1 -14.0605 splicing: 1 -14.0605 figure.) 1 -14.0605 p∈p, 1 -14.0605 dmin 1 -14.0605 malt. 1 -14.0605 wildcard. 1 -14.0605 gvr@students.iiit.ac.in 1 -14.0605 (councill 1 -14.0605 mentiondetectionmodel 1 -14.0605   1 -14.0605 part –labeled 1 -14.0605 revise, 1 -14.0605  pork  chop. 1 -14.0605 mad, sore) 1 -14.0605 st〉→〈[np 1 -14.0605 andpis 1 -14.0605 zanzotto@info.uniroma2.it 1 -14.0605 backtracks 1 -14.0605 syntactically-flexible 1 -14.0605 gobble. 1 -14.0605 performs. 1 -14.0605 1.0). 1 -14.0605 envelope 1 -14.0605 evens 1 -14.0605 y1. 1 -14.0605 spanmaps; 1 -14.0605 eq(4) 1 -14.0605 probabilities). 1 -14.0605 (mukherje 1 -14.0605 ∀s 1 -14.0605 gaussians, 1 -14.0605 sætre 1 -14.0605 higher-lower 1 -14.0605 k-factors. 1 -14.0605 counterexamples. 1 -14.0605 +/-10 1 -14.0605 mt(better): 1 -14.0605 {unplayful:1 serious:5 1 -14.0605 45.09 1 -14.0605 déjean 1 -14.0605 vagr, 1 -14.0605 ngram-r. 1 -14.0605 laboratoryofcomputationallinguistics(pekinguniversity)ministryofeducation,china 1 -14.0605 incby 1 -14.0605 montello 1 -14.0605 institut tecnol`ogic 1 -14.0605 mentioning. 1 -14.0605 soley 1 -14.0605 paolo rosso 1 -14.0605 user-validated 1 -14.0605 df∗,2). 1 -14.0605 58.54 1 -14.0605 titles), 1 -14.0605 27.92 57.66 1 -14.0605 pen. 1 -14.0605 lib 1 -14.0605 adapting percept, 1 -14.0605 cosmetic 1 -14.0605 software.6 1 -14.0605 5our 1 -14.0605 tightens 1 -14.0605 688 1 -14.0605 1btpd. 1 -14.0605 trick. 1 -14.0605 fgen, 1 -14.0605 fal etter 1 -14.0605 83.52%, 1 -14.0605 1.54 1.90 1 -14.0605 0.03. 1 -14.0605 pdtb-style 1 -14.0605 itsobjectiveistodeterminewhethertwomentions 1 -14.0605 best-scored 1 -14.0605 ‘cancer 1 -14.0605 .457 1 -14.0605 979 1 -14.0605 (lwlm) 1 -14.0605 68.28 1 -14.0605 variables{cij}for 1 -14.0605 1950) 1 -14.0605 attardi, 1 -14.0605 106. 1 -14.0605 xuen@brandeis.edu 1 -14.0605 core12 (repeated) 1 -14.0605 sentence-split 1 -14.0605 attribute-related 1 -14.0605 multi-layer, 1 -14.0605 utu 1 -14.0605 processes), 1 -14.0605 (contrary 1 -14.0605 [2,4]. 1 -14.0605 summary: 1 -14.0605 2002)), 1 -14.0605 greenwood 1 -14.0605 what, when, 1 -14.0605 eq(9). 1 -14.0605 modifier strings. 1 -14.0605 dual. 1 -14.0605 1they 1 -14.0605 back-translated 1 -14.0605 acetylation) 1 -14.0605 recalibration, 1 -14.0605 gui. 1 -14.0605 gray, 1 -14.0605 highlighting. 1 -14.0605 exemplify, 1 -14.0605 low-degree 1 -14.0605 ££day-month-part-of. 1 -14.0605 http:/ww.3.org/tr/owl-features/ 1 -14.0605 “phone” 1 -14.0605 liberty 1 -14.0605 (sogou-pinyin). 1 -14.0605 chart-parsing 1 -14.0605 activation(s). 1 -14.0605 99r70600, 1 -14.0605 downey, 1 -14.0605 knowledge resources#ablation 1 -14.0605 non-monotone. 1 -14.0605 ana cristina 1 -14.0605 translation-based lms 1 -14.0605 nats. 1 -14.0605 (nc). 1 -14.0605 heavily, 1 -14.0605 das zweite 1 -14.0605 sie... 1 -14.0605 words2, 1 -14.0605 +1 1 -14.0605 dedo), 1 -14.0605 (aw 1 -14.0605 stret journal 1 -14.0605 nd �2 1 -14.0605 rule-induction 1 -14.0605 prp-prplink.differentlinktypesmayhavediffer- 1 -14.0605 oe. 1 -14.0605 nullnullnullnullnullnullnullr 1 -14.0605 diagnosing 1 -14.0605 moore-lewis 35k 1 -14.0605 peak). 1 -14.0605 inches: 1 -14.0605 0.82. 1 -14.0605 (todman 1 -14.0605 rich´ard farkas 1 -14.0605 .w/ 1 -14.0605 stronggrowth” 1 -14.0605 feedbacks 1 -14.0605 rese 1 -14.0605 please ^‘comma’ 1 -14.0605 tutorials. 1 -14.0605 dataset5 1 -14.0605 ��hqviz 1 -14.0605 amplifier, 1 -14.0605 paired. 1 -14.0605 92.74 1 -14.0605 34.3% 1 -14.0605 1.5%, 1 -14.0605 i-gnssmm 1 -14.0605 frustratingly 1 -14.0605 (geraci 1 -14.0605 avoids. 1 -14.0605 seaghdha, 1 -14.0605 divergence). 1 -14.0605 (rai 1 -14.0605 tsc 1 -14.0605 case-insensitive. 1 -14.0605 qtb. 1 -14.0605 vbg, vbn, 1 -14.0605 1249 (.38) 1 -14.0605 fraction. 1 -14.0605 (2407 1 -14.0605 product/service 1 -14.0605 determing 1 -14.0605 sentimentality. 1 -14.0605 work! 1 -14.0605 pej 1 -14.0605 technologyassisted 1 -14.0605 (bergman, 1 -14.0605 non-poetic 1 -14.0605 λy.λx.loc(y,x)a1e1 1 -14.0605 agent, causer, 1 -14.0605 successfully. 1 -14.0605 kids. 1 -14.0605 randomization (ar) 1 -14.0605 {qing, 1 -14.0605 ke kaaran 1 -14.0605 approximate. 1 -14.0605 (bat) 1 -14.0605 overgenerate. 1 -14.0605 dialogue-specific 1 -14.0605 offset c(ˆwn−11 1 -14.0605 1ε 1 -14.0605 43.9% 1 -14.0605 k-nn, 1 -14.0605 vein. 1 -14.0605 helgadóttir, 1 -14.0605 envisioned. 1 -14.0605 sik 1 -14.0605 “transmit”. 1 -14.0605 doorway, 1 -14.0605 republic), p406/10/p259, 1 -14.0605 springer ccis 1 -14.0605 destroyed. 1 -14.0605 derivational. 1 -14.0605 4null2, 1 -14.0605 framesets. 1 -14.0605 linguisticaly 1 -14.0605 wer; 1 -14.0605 (27%). 1 -14.0605 1,...,t, 1 -14.0605 non-trivial; 1 -14.0605 nyc.com 1 -14.0605 domains’ 1 -14.0605 randomly selected. 1 -14.0605 (auli 1 -14.0605 system-building 1 -14.0605 lscript(y,yprime) 1 -14.0605 rec, p.800 1 -14.0605 within-conversation 1 -14.0605 d1. 1 -14.0605 (4a)). 1 -14.0605 discussion), 1 -14.0605 zp 1 -14.0605 bprime 1 -14.0605 1009 1 -14.0605 c(wn−11 1 -14.0605 “machine transliteration” 1 -14.0605 j]} 1 -14.0605 yprime) 1 -14.0605 (prevention). 1 -14.0605 93.2% 1 -14.0605 maritime greenwich, 1 -14.0605 request acknowledgment, 1 -14.0605 slms. 1 -14.0605 covariates, 1 -14.0605 34.3b 1 -14.0605 contact. 1 -14.0605 yj. 1 -14.0605 acceptability. 1 -14.0605 ghkm. 1 -14.0605 claire nédellec, 1 -14.0605 30.69 1 -14.0605 4note, 1 -14.0605 cancer’. 1 -14.0605 “construction request,” 1 -14.0605 (asuncion 1 -14.0605 c[i−1:e] 1 -14.0605 che rimane 1 -14.0605 (collocation). 1 -14.0605 taiwan. 1 -14.0605 50.77 43.25 1 -14.0605 competitively. 1 -14.0605 elixirfm 1 -14.0605 someone else’s 1 -14.0605 ’out-of-ten’ 1 -14.0605 87.64 1 -14.0605 precondition. 1 -14.0605 reusable. 1 -14.0605 henceforth. 1 -14.0605 + 1−pr)s 1 -14.0605 chi-square, 1 -14.0605 co-mentions 1 -14.0605 (francopoulo 1 -14.0605 cslu 1 -14.0605 277, 1 -14.0605 “shenme”. 1 -14.0605 post-editors 1 -14.0605 ri. 1 -14.0605 ineffective. 1 -14.0605 72.73 1 -14.0605 erg. 1 -14.0605 (slm). 1 -14.0605 nullnullnullnull1nullnullnullnull2null null nullnullnull. 1 -14.0605 (smt, 2011) 1 -14.0605 p(i)log p(i)q(i) 1 -14.0605 acknowledgements: 1 -14.0605 determinable 1 -14.0605 (preliminary 1 -14.0605 proportion) 1 -14.0605 boxer. 1 -14.0605 all-female 1 -14.0605 radio z-rank 1 -14.0605 enmeshed 1 -14.0605 (functional) 1 -14.0605 (29%). 1 -14.0605 non-expert, 1 -14.0605 ² 1 -14.0605 relevant). 1 -14.0605 medstract (castano 1 -14.0605 taskdependent 1 -14.0605 textbooks. 1 -14.0605 botstraped 1 -14.0605 prove. 1 -14.0605 aware. 1 -14.0605 reference-free 1 -14.0605 qa. 1 -14.0605 (paul, 1 -14.0605 xprime(sprime)) 1 -14.0605 φ). 1 -14.0605 atb, 1 -14.0605 pb. 1 -14.0605  overall 1 -14.0605 that-inclusion 1 -14.0605 150000) 1 -14.0605 85.62 1 -14.0605 s.s. 1 -14.0605 (1987). 1 -14.0605 parameters). 1 -14.0605 projected parsesby 1 -14.0605 heriberto.cuayahuitl@dfki.de 1 -14.0605 s,t. 1 -14.0605 dialect? 1 -14.0605 pathogen) 1 -14.0605 presentinfo 1 -14.0605 typelda 1 -14.0605 esr1. 1 -14.0605 corresponds, 1 -14.0605 parseval. 1 -14.0605 (0.11) 1 -14.0605 proportional. 1 -14.0605 sumlm. 1 -14.0605 (7b), 1 -14.0605 2,3,4,5, 1 -14.0605 awe 1 -14.0605 companies/publications. 1 -14.0605 49%. 1 -14.0605 interpretations). 1 -14.0605 (reaching sj) 1 -14.0605 (60873091; 61073140), 1 -14.0605 kprel. 1 -14.0605 threedifferentkindsof linktypes. 1 -14.0605 (r,dk) 1 -14.0605 target; 1 -14.0605 10001500 20002500 1 -14.0605 token). 1 -14.0605 unilateral 1 -14.0605 (milne08). 1 -14.0605 aij. 1 -14.0605 hierarchal 1 -14.0605 jun’ichi tsujii‡ 1 -14.0605 weight; 1 -14.0605 ai×j≈wi×khk×j 1 -14.0605 k∈a(x) eθk 1 -14.0605 (as). 1 -14.0605 [mary]x [came]y 1 -14.0605 mini-documents 1 -14.0605 visitors 1 -14.0605 mcguffey’s eclectic 1 -14.0605 (pca) 1 -14.0605 sharper 1 -14.0605 topic). 1 -14.0605 unsel) 1 -14.0605 gaizauskas. 1 -14.0605 (blake 1 -14.0605 non-pronominals 1 -14.0605 2.6, 1 -14.0605 doubling. 1 -14.0605 rate). 1 -14.0605 fraser, 1 -14.0605 category-based 1 -14.0605 a1...am 1 -14.0605 γ′. 1 -14.0605 broadcasts, 1 -14.0605 vignet. 1 -14.0605 g(〈s 1 -14.0605 gutenberg.) 1 -14.0605 o). 1 -14.0605 p(r). 1 -14.0605 glscript,k(x,y), 1 -14.0605 formulae. 1 -14.0605 −pp uninverted, 1 -14.0605 “die zeit”, 1 -14.0605 subcategories. 1 -14.0605 foreigners nur 1 -14.0605 altantawy 1 -14.0605 height−frontnessharmony 1 -14.0605 bioframenet, 1 -14.0605 overall). 1 -14.0605 (pargellis 1 -14.0605 application” 1 -14.0605 ’.’, 1 -14.0605 bakin 1 -14.0605 11–15). 1 -14.0605 space.) 1 -14.0605 training/evaluation routines. 1 -14.0605 sentenceinternal 1 -14.0605 jni 1 -14.0605 stroke. 1 -14.0605 s.stoyanchev@open.ac.uk 1 -14.0605 4null 1 -14.0605 savings. 1 -14.0605 hovy (2010: 1 -14.0605 one-to-many, 1 -14.0605 (68.3%). 1 -14.0605 gennadi lembersky 1 -14.0605 diseasedevelopmentdata. 1 -14.0605 shirt” 1 -14.0605 definitenpfeatures-trueifthegivennounphrase 1 -14.0605 multi-tasking dialogues, 1 -14.0605 rilof, 1 -14.0605 muccs 1 -14.0605 (alvez 1 -14.0605 1–5, 1 -14.0605 -2. 1 -14.0605 classifications). 1 -14.0605 webtlab team 1 -14.0605 pivot-oriented 1 -14.0605 hill. 1 -14.0605 tsuji, 1 -14.0605 madonna. 1 -14.0605 “transition” 1 -14.0605 polynomialtime 1 -14.0605 wedding gowns 1 -14.0605 ‘the hound 1 -14.0605 sl-based 1 -14.0605 [giza] 1 -14.0605 extent), 1 -14.0605 awakened 1 -14.0605 g18g367g381g400g286g282g855g898g396g349g448g286g396g856g286g449g374g856g1005g899 g75g393g286g374g855 1 -14.0605 dahl, 1 -14.0605 consensusbased 1 -14.0605 lexicographically). 1 -14.0605 (arg1, 1 -14.0605 7%, 1 -14.0605 singular). 1 -14.0605 matsubara, 1 -14.0605 shudo 1 -14.0605 that! 1 -14.0605 pennsylvania, philadelphia, 1 -14.0605 openie 1 -14.0605 proliferate 1 -14.0605 un-aligned 1 -14.0605 semafor, 1 -14.0605 71.65 +0.95 1 -14.0605 disco’2011 1 -14.0605 co-em 1 -14.0605 eula. 1 -14.0605 discriminating, 1 -14.0605 §5 1 -14.0605 (polarity) 1 -14.0605 “letter” 1 -14.0605 unfamiliar, 1 -14.0605 0750271 1 -14.0605 corpus? 1 -14.0605 qgstec 1 -14.0605 http://www.wikipedia.org/ 1 -14.0605 i-computation. 1 -14.0605 (people 1 -14.0605 enery 1 -14.0605 barge-in. 1 -14.0605 garrod 1 -14.0605 (∑ j∈t(ci,j+wpartial 1 -14.0605 33.4% 1 -14.0605 alignment). 1 -14.0605 not-sparse 1 -14.0605 (optical 1 -14.0605 hastings 1 -14.0605 bostonturnstileleftnp 1 -14.0605 perceptibility 1 -14.0605 operates. 1 -14.0605 icelandic volcanic 1 -14.0605 productdisplay (vi,vj)∈e 1 -14.0605 negatively, 1 -14.0605 biologist 1 -14.0605 emotion) 1 -14.0605 nondeterministically 1 -14.0605 kim, tomoko 1 -14.0605 verbaly 1 -14.0605 nummod 1 -14.0605 came. 1 -14.0605 preselected 1 -14.0605 vv. 1 -14.0605 jaccard (wj) 1 -14.0605 13.64 78.95 1 -14.0605 (1992)’s 1 -14.0605 non-opinionated 1 -14.0605 ∀ l x.[pred(l,“v 1 -14.0605 ga-extr, 1 -14.0605 comb. 1 -14.0605 “cleaner” 1 -14.0605 roiri@csail.mit.edu 1 -14.0605 korean serial 1 -14.0605 rank: 1 -14.0605 reinforcement. 1 -14.0605 institutions, 1 -14.0605 non-arguments. 1 -14.0605 nbadapted 1 -14.0605 egypt, 1 -14.0605 ! 1 -14.0605 lvcsr. 1 -14.0605 rumorous. 1 -14.0605 55.70 77.60 1 -14.0605 “vanilla” 1 -14.0605 sanae fujita, 1 -14.0605 ist, ¨ubt 1 -14.0605 shortcoming, 1 -14.0605 group-lasso. 1 -14.0605 passed. 1 -14.0605 jennifer stromer-galley 1 -14.0605 mount pinatubo 1 -14.0605 universitat d’alacant 1 -14.0605 5.5. 1 -14.0605 (a,b,c) 1 -14.0605 biocurators. 1 -14.0605 port menestheus’s 1 -14.0605 leibler, 1 -14.0605 imdb. 1 -14.0605 expressed) 1 -14.0605 ao. 1 -14.0605 (glenn 1 -14.0605 drink-pst-dec 'he 1 -14.0605 dkauchak@cs.pomona.edu 1 -14.0605 76.83 1 -14.0605 0w 1 -14.0605 domain(r2) 1 -14.0605 count(?s) 1 -14.0605 g.j.m.van.noord@rug.nl 1 -14.0605 (§4.2) 1 -14.0605 generally. 1 -14.0605 (p<.05) 1 -14.0605 nips 1 -14.0605 cheese. 1 -14.0605 llu´ıs padr´o 1 -14.0605 650-652-5806. 1 -14.0605 1314 1 -14.0605 fluctuated 1 -14.0605 1,2,3. 1 -14.0605 visualize). 1 -14.0605 languages? 1 -14.0605 next? 1 -14.0605 93.7%, 1 -14.0605 s.wubben@uvt.nl 1 -14.0605 .548 1 -14.0605 project: www.classic-project.org), 1 -14.0605 dixon’s 1 -14.0605 (in percent) 1 -14.0605 spos. 1 -14.0605 entity-wide 1 -14.0605 wirelessly. 1 -14.0605 state-splitting. 1 -14.0605 heaps 1 -14.0605 nv. 1 -14.0605 experimentalresultson developmentset 1 -14.0605 lcomb 1 -14.0605 alexander berman 1 -14.0605 v∈c. 1 -14.0605 governor; 1 -14.0605 meylan, 1 -14.0605 ‘≡’ 1 -14.0605 4,348 1 -14.0605 selectionprocedure.specially, 1 -14.0605 compilation. 1 -14.0605 unaccounted 1 -14.0605 }∩{t 1 -14.0605 robert. 1 -14.0605 cortes 1 -14.0605 post-filter. 1 -14.0605 1hmm. 1 -14.0605 linux, 1 -14.0605 59.78 1 -14.0605 147. 1 -14.0605 treebank2. 1 -14.0605 idealised 1 -14.0605 1-4% 1 -14.0605 statisticallysignificant. 1 -14.0605 zb. 1 -14.0605 supercomputer. 1 -14.0605 number 7398 1 -14.0605 karrowbothv(gprime1,gprime2) 1 -14.0605 multiwordnet (mwn) 1 -14.0605 1998b; ciravegna, 1 -14.0605 resources2. 1 -14.0605 famaf 1 -14.0605 summarize. 1 -14.0605 ilpbased 1 -14.0605 wasteful 1 -14.0605 1,284 1 -14.0605 ,t,li)} 1 -14.0605 posted. 1 -14.0605 “research” 1 -14.0605 /o/ 1 -14.0605 s1,...,sl. 1 -14.0605 β1) 1 -14.0605 quitea 1 -14.0605 dialogue-acts, 1 -14.0605 macedonian vase” 1 -14.0605 (57%). 1 -14.0605 stones 1 -14.0605 qe. 1 -14.0605 mod, pnmod, 1 -14.0605 1,2). 1 -14.0605 i.h. 1 -14.0605 (dorr, 1 -14.0605 tesla-tuned 1 -14.0605 (wfst). 1 -14.0605 overview). 1 -14.0605 pollack 1 -14.0605 *in 1 -14.0605 ifip 1 -14.0605 achievements. 1 -14.0605 distributes 1 -14.0605 stacking, 1 -14.0605 itgs 1 -14.0605 xprime∈n(f)\x µxprime→f) 1 -14.0605 75%, 1 -14.0605 examiner. 1 -14.0605 parsing-specific 1 -14.0605 chinese-english. 1 -14.0605 analysts’ 1 -14.0605 urduenglish, 1 -14.0605 acc prec 1 -14.0605 lower-casing 1 -14.0605 hierarchically. 1 -14.0605 51.9% 1 -14.0605 stts 1 -14.0605 )] n×1 1 -14.0605 able-bodied, 1 -14.0605 sameer pradhan 1 -14.0605 response). 1 -14.0605 daniel bauer 1 -14.0605 dimensions” 1 -14.0605 floor-grabbing 1 -14.0605 spinal 1 -14.0605 inflected; 1 -14.0605 0.40. 1 -14.0605 representationand 1 -14.0605 a2 fa3 1 -14.0605 hypothesis). 1 -14.0605 pado 1 -14.0605 are coreferentialwitha 1 -14.0605 hypothesis,ceprime(ng) 1 -14.0605 (pak 1 -14.0605 http://bionlp.utu.fi/. 1 -14.0605 motivated; 1 -14.0605 non-responses. 1 -14.0605 today! 1 -14.0605 gente 1 -14.0605 nonsarcastic 1 -14.0605 (undirected) 1 -14.0605 ,tt[k],rt) 1 -14.0605 re-defined 1 -14.0605 university of tübingen 1 -14.0605 lex) 1 -14.0605 bracketleftbig fla1b2,b1(γ)+fra1b2,b1(γ) 1 -14.0605 proceed. 1 -14.0605 multi-word, 1 -14.0605 severe. 1 -14.0605 si). 1 -14.0605 nucle, 1 -14.0605 business. 1 -14.0605 van landeghem1,2, 1 -14.0605 ez zuen 1 -14.0605 p1(x) 1 -14.0605 non-projectivity. 1 -14.0605 l1-paraphrase 1 -14.0605 indicted 1 -14.0605 detail: 1 -14.0605 1954), 1 -14.0605 tiáowén 1 -14.0605 shíshī 1 -14.0605 hyperlinks. 1 -14.0605 chris, 1 -14.0605 t(i) 1 -14.0605 right? 1 -14.0605 phrasepairs 1 -14.0605 kanazawa 1 -14.0605 (280) 1 -14.0605 going? 1 -14.0605 1109 1 -14.0605 priberam. 1 -14.0605 trades-off 1 -14.0605 exp.arg2 1 -14.0605 prepp 1 -14.0605 sender="#p2" addressee="#p1" 1 -14.0605 time”. 1 -14.0605 2.10 1 -14.0605 forest”. 1 -14.0605 danger–red, 1 -14.0605 turn-bidding 1 -14.0605 profiled 1 -14.0605 isu-based 1 -14.0605 tesla-f. 1 -14.0605 (no.20090001110047). 1 -14.0605 takipi. 1 -14.0605 punctuation). 1 -14.0605 tmp. 1 -14.0605 “bleu_d” 1 -14.0605 vidual 1 -14.0605 numb. 1 -14.0605 semcor. 1 -14.0605 24.39 68.01 1 -14.0605 pierre zweigenbaum 1 -14.0605 43.28 1 -14.0605 post-test. 1 -14.0605 faster? 1 -14.0605 enumeration. 1 -14.0605 pivots, 1 -14.0605 mwpro 1 -14.0605 (uima, 1 -14.0605 nounphraseinitssentence. 1 -14.0605 sub-samples 1 -14.0605 28). 1 -14.0605 experience). 1 -14.0605 2.67 6.05 1 -14.0605 h3t 1j4 1 -14.0605 imagability 1 -14.0605 h5. 1 -14.0605 harry.bunt@uvt.nl 1 -14.0605 tokenizers. 1 -14.0605 gere) 1 -14.0605 self-training) 1 -14.0605 vyas. 1 -14.0605 npv-trigram.} 1 -14.0605 sections). 1 -14.0605 propbank/verbnet 1 -14.0605 re-consider 1 -14.0605 complexities. 1 -14.0605 naturalness). 1 -14.0605 −choose word 1 -14.0605 fluctuations. 1 -14.0605 (§8). 1 -14.0605 86.91 1 -14.0605 h3. 1 -14.0605 variablefree 1 -14.0605 84.90% 1 -14.0605 group’s 1 -14.0605 tatoeba, 1 -14.0605 hashes. 1 -14.0605 were. 1 -14.0605 angbracketleftbig λy.λx.loc(y,x)∧state(x)a1e1 1 -14.0605 atemporal 1 -14.0605 10-words 1 -14.0605 manyto-many 1 -14.0605 instrument). 1 -14.0605 verification) 1 -14.0605 (europarl) 1 -14.0605 ⟨fs,t,a⟩, 1 -14.0605 concern; 1 -14.0605 )rvvxphwdo  1 -14.0605 europarl) 1 -14.0605 “v. 1 -14.0605 hatwidefx. 1 -14.0605 (slovak). 1 -14.0605 least. 1 -14.0605 vallejo. 1 -14.0605 hmm+lm 1 -14.0605 (1998, 1 -14.0605 medical). 1 -14.0605 (n-only) 1 -14.0605 deer 1 -14.0605 (pbt 1 -14.0605 uncurated 1 -14.0605 codified 1 -14.0605 •it 1 -14.0605 assessment: 1 -14.0605 weikum. 1 -14.0605 hlt/naacl 1 -14.0605 (indyk 1 -14.0605 (data, 1 -14.0605 yoi 1 -14.0605 “reduced” 1 -14.0605 .translation 1 -14.0605 (stubbs, 1 -14.0605 p@n 1 -14.0605 ad hoc, 1 -14.0605 yahā ṃ 1 -14.0605 item). 1 -14.0605 airlines. 1 -14.0605 farzindar, 1 -14.0605 zation 1 -14.0605 “jr.” 1 -14.0605 randomness. 1 -14.0605 ontology authoring, 1 -14.0605 rat. 1 -14.0605 kabale 1 -14.0605 writing-style 1 -14.0605 mharper@umd.edu 1 -14.0605 cfs. 1 -14.0605 sub-problems: 1 -14.0605 (mae). 1 -14.0605 'f cfc 1 -14.0605 gaze. 1 -14.0605 completed games. 1 -14.0605 soundbites 1 -14.0605 kenneth.church@jhu.edu 1 -14.0605 port-media. 1 -14.0605 10-docs 50-docs 1 -14.0605 kipper schuler 1 -14.0605 §5). 1 -14.0605 purposes) 1 -14.0605 ¬∃p.p 1 -14.0605 “edited” labellings 1 -14.0605 finalized 1 -14.0605 np↓-x↓ 1 -14.0605 n07 1 -14.0605 .554 1 -14.0605 faq 1 -14.0605 52k 1 -14.0605 iis-0713548. 1 -14.0605 83.32 1 -14.0605 concrete) 1 -14.0605 dpi. 1 -14.0605 resnik’s 1 -14.0605 (levi, 1 -14.0605 (upspeak 1 -14.0605 coercing 1 -14.0605 substring) 1 -14.0605 yuv ≤1 1 -14.0605 i] admired 1 -14.0605 test-52. 1 -14.0605 dragging 1 -14.0605 newsblaster 1 -14.0605 this; 1 -14.0605 snaptu 1 -14.0605 (+0.66) 1 -14.0605 visibility 1 -14.0605 signed-ranks 1 -14.0605 “breaking news” 1 -14.0605 votes, 1 -14.0605 weekends. 1 -14.0605 sophisticated. 1 -14.0605 restrictive, 1 -14.0605 immortal. 1 -14.0605 tsp. 1 -14.0605 [enters 1 -14.0605 themselves; 1 -14.0605 embedding. 1 -14.0605 mgalley@microsoft.com 1 -14.0605 (budzik 1 -14.0605 uh-huh a: 1 -14.0605 constituency-todependency 1 -14.0605 svm-score 1 -14.0605 (lms). 1 -14.0605 position). 1 -14.0605 spaces’ 1 -14.0605 (node 1 -14.0605 iis-0712764 1 -14.0605 character-by-character 1 -14.0605 12,500 1 -14.0605 ˆθ, 1 -14.0605 bored 1 -14.0605 35.7% 1 -14.0605 meaningful), 1 -14.0605 263-270. 1 -14.0605 clustering." 1 -14.0605 necesary. 1 -14.0605 goal). 1 -14.0605 (albrecht 1 -14.0605 allfragments 1 -14.0605 four-fold 1 -14.0605 beam-sizes, 1 -14.0605 [a1 services] 1 -14.0605 duc-07. 1 -14.0605 niloofar montazeri 1 -14.0605 35th 1 -14.0605 klaus tschira 1 -14.0605 students). 1 -14.0605 backof 1 -14.0605 pitched 1 -14.0605 empty-word 1 -14.0605 envisaged 1 -14.0605 exam. 1 -14.0605 lexpagerank 1 -14.0605 bracketings. 1 -14.0605 70. 1 -14.0605 score—number 1 -14.0605 (0.20) 1 -14.0605 effect). 1 -14.0605 boh 1 -14.0605 appropriateness consolidation: 1 -14.0605 tree.” 1 -14.0605 oravecz 1 -14.0605 (jespersen, 1 -14.0605 0.11. 1 -14.0605 corpusspecific 1 -14.0605 devtest. 1 -14.0605 d1,d2summationdisplay i,j=1 1 -14.0605 fourteenth 1 -14.0605 (conroy 1 -14.0605 he pounded 1 -14.0605 fluence 1 -14.0605 unobserved. 1 -14.0605 120. 1 -14.0605 trading. 1 -14.0605 divergence) 1 -14.0605 characteristic. 1 -14.0605 0.27. 1 -14.0605 explodes. 1 -14.0605 pls, 1 -14.0605 0.02. 1 -14.0605 pedagogy. 1 -14.0605 multi-authored 1 -14.0605 (204) “speling 1 -14.0605 malay-english 1 -14.0605 (nishimura 1 -14.0605 bi-parsed 1 -14.0605 trend). 1 -14.0605 explantation 1 -14.0605 far away. 1 -14.0605 speculations. 1 -14.0605 zipf (1935) 1 -14.0605 modification) 1 -14.0605 positional. 1 -14.0605 “a”. 1 -14.0605 1,...,r 1 -14.0605 roser morante, 1 -14.0605 [causal] 1 -14.0605 rivers? 1 -14.0605 directive. 1 -14.0605 text-to-graphics 1 -14.0605 mistranslation 1 -14.0605 48.94% 1 -14.0605 argumentof 1 -14.0605 depart. 1 -14.0605 slotted 1 -14.0605 multi-dir. 1 -14.0605 1..k: 1 -14.0605 {a1, 1 -14.0605 authortopic 1 -14.0605 b1 b2a 1 -14.0605 moot. 1 -14.0605 type-token 1 -14.0605 39–47, 1 -14.0605 residential steak 1 -14.0605 ˆy1:t 1 -14.0605 100k cipher 1 -14.0605 biasness 1 -14.0605 jbaldrid@mail.utexas.edu 1 -14.0605 argamon. 1 -14.0605 3.6). 1 -14.0605 2.5.1). 1 -14.0605 dg882>b 1 -14.0605 answer). 1 -14.0605 alternately. 1 -14.0605 eral 1 -14.0605 gone. 1 -14.0605 “m” 1 -14.0605 test-takers’ 1 -14.0605 gnc, 1 -14.0605 (2000)). 1 -14.0605 words4. 1 -14.0605 pc(o), 1 -14.0605 macnish, 1 -14.0605 (jimeno-yepes 1 -14.0605 brevity penalty) 1 -14.0605 directories 1 -14.0605 (birke 1 -14.0605 visualisation, 1 -14.0605 48%. 1 -14.0605 harmonizing 1 -14.0605 alexis nasr 1 -14.0605 matthias eck, 1 -14.0605 1one 1 -14.0605 (1996; 1 -14.0605 +1.6 1 -14.0605 outperformed, 1 -14.0605 s∗t 1 -14.0605 ted-based 1 -14.0605 relevance+readability 1 -14.0605 30.03 26.91 1 -14.0605 parsimony, 1 -14.0605 semanticallycoherent 1 -14.0605 exists) 1 -14.0605 utilized, 1 -14.0605 cea. 1 -14.0605 show.v. 1 -14.0605 f1-score, 1 -14.0605 θdk 1 -14.0605 televised 1 -14.0605 móra 1 -14.0605 sub-tree. 1 -14.0605 nonalphanumeric 1 -14.0605 47.00 1 -14.0605 2(t2)d 3(t2) 1 -14.0605 mean(andstandarddeviation)ofmaximum 1 -14.0605 replaced. 1 -14.0605 constitution 1 -14.0605 (kalmar, 1 -14.0605 uplug 1 -14.0605 bugliesi, lamma, 1 -14.0605 kbp track. 1 -14.0605 developing. 1 -14.0605 randlm’s 1 -14.0605 schmitt 1 -14.0605 generalizer gud. 1 -14.0605 lems 1 -14.0605 delves 1 -14.0605 co-occurs, 1 -14.0605 l(u). 1 -14.0605 kept). 1 -14.0605 spandana gella 1 -14.0605 λx.state(x)) 1 -14.0605 negotiations). 1 -14.0605 pos4 pos5 1 -14.0605 uni. 1 -14.0605 algebraically 1 -14.0605 the)?,sg 1 -14.0605 cycling 1 -14.0605 f=1 (te,f)ne,f 1 -14.0605 reminding 1 -14.0605 errorprone 1 -14.0605 p(+nullnullp,iarg0,ci,iarg1,cjnull) 1 -14.0605 subj vobj 1 -14.0605 1-ter[%] 1b 1 -14.0605 succesful 1 -14.0605 3there 1 -14.0605 classifiers). 1 -14.0605 disagrement. 1 -14.0605 01.00 p.m 1 -14.0605 option: hire 1 -14.0605 giga 5-gr 1 -14.0605 selectively. 1 -14.0605 trapezoids 1 -14.0605 desktop. 1 -14.0605 thatwi 1 -14.0605 nyt. 1 -14.0605 , 1 -14.0605 (burger 1 -14.0605 jancura 1 -14.0605 acuracy. 1 -14.0605 qui appartient 1 -14.0605 b(b) d(d) 1 -14.0605 anaphoric. 1 -14.0605 [0,1]). 1 -14.0605 decisions) 1 -14.0605 setings. 1 -14.0605 text-specific 1 -14.0605 (skut 1 -14.0605 letters/digits/apostrophe, 1 -14.0605 −.528 1 -14.0605 (m)) 1 -14.0605 6they 1 -14.0605 ip-node. 1 -14.0605 ηι 1 -14.0605 right-endoc. 1 -14.0605 concordances. 1 -14.0605 994 1 -14.0605 monument’s 1 -14.0605 “dislike” 1 -14.0605 hartu (“to 1 -14.0605 ∗correspondence author: 1 -14.0605 crosssource 1 -14.0605 munson 1 -14.0605 4,060 1 -14.0605 (shabdanjali) 1 -14.0605 (lit.: ‘memory 1 -14.0605 incongruous 1 -14.0605 delalleau, 1 -14.0605 pa(w) pb(w) 1 -14.0605 widths, 1 -14.0605 acl’06. 1 -14.0605 441–449, 1 -14.0605 if-then 1 -14.0605 t2j+1 1 -14.0605 105–110, 1 -14.0605 mwn. 1 -14.0605 associates (jdpa) 1 -14.0605 )q 1 -14.0605 nudge 1 -14.0605 (zero 1 -14.0605 pl-mrfs 1 -14.0605 falls. 1 -14.0605 (tsujimura, 1 -14.0605 〈wk,wprimek〉 1 -14.0605 mentionboundaries areshownintable1andtable2. 1 -14.0605 (it’s 1 -14.0605 recursions. 1 -14.0605 plaintext:d 1 -14.0605 result) 1 -14.0605 appeasement emboldened 1 -14.0605 organisation. 1 -14.0605 wellrepresented 1 -14.0605 1−f. 1 -14.0605 6,312 1 -14.0605 labs. 1 -14.0605 num sg, 1 -14.0605 force. 1 -14.0605 “head-percolation” 1 -14.0605 correct4. 1 -14.0605 sub-parts, 1 -14.0605 94.2% 1 -14.0605 infrequently. 1 -14.0605 cardinal, 1 -14.0605 1/3. 1 -14.0605 ters 1 -14.0605 95.70% 1 -14.0605 spans) 1 -14.0605 theirs), 1 -14.0605 unprocessed, 1 -14.0605 resolvers, 1 -14.0605 joshua), 1 -14.0605 “solvable” 1 -14.0605 http://www.cs.sfu.ca/ anoop/students/jbirke/. 1 -14.0605 s0 5] 1 -14.0605 vextendsinglevextendsingle∂l(ws) ∂wi 1 -14.0605 cherry pie, 1 -14.0605 “green” 1 -14.0605 variabilities 1 -14.0605 administered. 1 -14.0605 saers et 1 -14.0605 by; near; 1 -14.0605 (§3.2), 1 -14.0605 sparkling 1 -14.0605 gratefully acknowledged. 1 -14.0605 coercive, 1 -14.0605 /b2(x 1 -14.0605 functions). 1 -14.0605 (mwu) 1 -14.0605 gimpel). 1 -14.0605 mortar\w2 attacks\h 1 -14.0605 (1993, 1 -14.0605 oft-observed 1 -14.0605 phenotype. 1 -14.0605 81.48% 1 -14.0605 bard. 1 -14.0605 luo, 1 -14.0605 bootstrappers 1 -14.0605 first-best 1 -14.0605 seven. 1 -14.0605 models.4 1 -14.0605 69.78 1 -14.0605 2004: 1 -14.0605 coalescence. 1 -14.0605 firstname.lastname@imag.fr 1 -14.0605 live. 1 -14.0605 (perez-tellez 1 -14.0605 hit-3 (bidirectional 1 -14.0605 (brickley 1 -14.0605 fellbaum (ed.), 1 -14.0605 sarah. 1 -14.0605 eng1, 1 -14.0605 overfitting; 1 -14.0605 dearth 1 -14.0605 (eurospeech). 1 -14.0605 demote 1 -14.0605 pie y.) 1 -14.0605 ntcir-6. 1 -14.0605 63.14 1 -14.0605 30.82 32.54 1 -14.0605 onementionis 1 -14.0605 24.16 32.04 1 -14.0605 15we 1 -14.0605 phonetic-, 1 -14.0605 porly 1 -14.0605 88–97, 1 -14.0605 induces. 1 -14.0605 complementizer-choice 1 -14.0605 sparing 1 -14.0605 1–3. 1 -14.0605 expedite 1 -14.0605 human-rated 1 -14.0605 1950 1 -14.0605 miscellaneous. 1 -14.0605 2due 1 -14.0605 distribution’s 1 -14.0605 (0,0,...,0) 1 -14.0605 ratkiewicz 1 -14.0605 syntax syntaxrelaxed 1 -14.0605 (fort 1 -14.0605 satisfies. 1 -14.0605 (w.r.t. 1 -14.0605 rhyme. 1 -14.0605 1 -14.0605 mis-tagged 1 -14.0605 funds. 1 -14.0605 hindsight, 1 -14.0605 simlch(a,b) 1 -14.0605 spaces: 1 -14.0605 thrun, 1 -14.0605 (bcubed, 1 -14.0605 voluminous 1 -14.0605 include. 1 -14.0605 +pp uninverted, 1 -14.0605 beast 1 -14.0605 (otherwise 1 -14.0605 team). 1 -14.0605 hypertexify 1 -14.0605 designers, 1 -14.0605 recall/precision/f-score 1 -14.0605 formulated, 1 -14.0605 i0, 1 -14.0605 ?”, 1 -14.0605 se?, 1 -14.0605 fang, 1 -14.0605 svrrec 1 -14.0605 apostrophes, 1 -14.0605 model-hmmp 1 -14.0605 75k 1 -14.0605 furnished 1 -14.0605 splitter2. 1 -14.0605 unity. 1 -14.0605 calvin 1 -14.0605 (sfc) 1 -14.0605 (16% 1 -14.0605 stein. 1 -14.0605 maise’s codebase 1 -14.0605 mt-eval 1 -14.0605 )ising 1 -14.0605 top-1000 1 -14.0605 decide. 1 -14.0605 protects 1 -14.0605 10−8. 1 -14.0605 (') 1 -14.0605 uncertainties. 1 -14.0605 desempenho[performance] melhore[improves] 1 -14.0605 minimal-rule 1 -14.0605 ]⇒σg[w1, 1 -14.0605 )φ. 1 -14.0605 rover (fiscus, 1 -14.0605 senna. 1 -14.0605 edge-flips. 1 -14.0605 (2. 1 -14.0605 talk; 1 -14.0605 ch(r) 1 -14.0605 10−2, 10−3, 1 -14.0605 growdiag-final-and 1 -14.0605 (35.45%) 1 -14.0605 ‹ /f/then 1 -14.0605 s(a). 1 -14.0605 0.1786 1 -14.0605 positive–negative 1 -14.0605 corpus- 1 -14.0605 417-422. 1 -14.0605 hr0011-06-c-0022 (subcontract 1 -14.0605 18,085 1 -14.0605 titov@mmci.uni-saarland.de 1 -14.0605 nounphraseare 1 -14.0605 (dreyer, 1 -14.0605 no-remediation 1 -14.0605 phrasetranslation 1 -14.0605 sigir2002. 1 -14.0605 32–42, 1 -14.0605 1≤j≤n. 1 -14.0605 excellatach, 1 -14.0605 blogger.com 1 -14.0605 (mitkov 1 -14.0605 ktfk 1 -14.0605 (1978). 1 -14.0605 -tuned 1 -14.0605 72.83 1 -14.0605 (±se). 1 -14.0605 occ(t2). 1 -14.0605 shaikh 1 -14.0605 (minimal) 1 -14.0605 misclassifications. 1 -14.0605 speaker/addressee 1 -14.0605 collocations) 1 -14.0605 {nn, jj} 1 -14.0605 measures). 1 -14.0605 chua, 1 -14.0605 roulette 1 -14.0605 mert). 1 -14.0605 query). 1 -14.0605 nonterm 1 -14.0605 v tns 1 -14.0605 1986. 1 -14.0605 sigurd 1 -14.0605 lscript1-regularized 1 -14.0605 additionallm. 1 -14.0605 laboratoire informatique 1 -14.0605 hlt/emnlp. 1 -14.0605 lego mindstorms 1 -14.0605 tiebreaking. 1 -14.0605 supersenses. 1 -14.0605 features.) 1 -14.0605 ccomp 1 -14.0605 conversation? 1 -14.0605 wsj∞. 1 -14.0605 cross-lingual/multilingual 1 -14.0605 therapists. 1 -14.0605 atmosphere. 1 -14.0605 v1dvt2 1 -14.0605 edward.grefenstette@cs.ox.ac.uk 1 -14.0605 picturable 1 -14.0605 acl-07 1 -14.0605 (bryl 1 -14.0605 log-score 1 -14.0605 inc. 2065 1 -14.0605 (2ab), 1 -14.0605 algo. 1 -14.0605 gid 1 -14.0605 innateness 1 -14.0605 bavarian), 1 -14.0605 zhai. 1 -14.0605 ewtm. 1 -14.0605 america). 1 -14.0605 r(wm,wc) 1 -14.0605 {0,...,n 1 -14.0605 “punctuation 1 -14.0605 tmargin 1 -14.0605 trec-2006 1 -14.0605 {r}= constructrules,σ,a(x,ξ) 1 -14.0605 delimiter. 1 -14.0605 positioning, 1 -14.0605 flp. 1 -14.0605 (mext, japan), 1 -14.0605 1,1 ,...,s 1 -14.0605 one-by-one. 1 -14.0605 (dcs), 1 -14.0605 (pccg) 1 -14.0605 parsers), 1 -14.0605 documents.4 1 -14.0605 (gabani 1 -14.0605 0.23. 1 -14.0605 kearns, 1 -14.0605 (200,000 1 -14.0605 gramars 1 -14.0605 hdp. 1 -14.0605 outputted 1 -14.0605 chromosome. 1 -14.0605 analyzing. 1 -14.0605 -0.62 1 -14.0605 urement 1 -14.0605 waibel. 1 -14.0605 6.3. 1 -14.0605 ‘clust-7-tgt-slow’ 1 -14.0605 1,...,m. 1 -14.0605 “walked” 1 -14.0605 “generally known” 1 -14.0605 uk suresh@cs.york.ac.uk 1 -14.0605 sense-appropriate, 1 -14.0605 burnin 1 -14.0605 document.” 1 -14.0605 1...t: 1 -14.0605 no-mention 1 -14.0605 0.29, 1 -14.0605 hash+scroll, 1 -14.0605 7,628 1 -14.0605 kappa. 1 -14.0605 aircraft” “apache 1 -14.0605 67.14 1 -14.0605 andq-queries. 1 -14.0605 wordnet,propbankand 1 -14.0605 prachya boonkwan, 1 -14.0605 syntactified 1 -14.0605 arranged. 1 -14.0605 let o(q) = 1 -14.0605 kbp2010. 1 -14.0605 (+tb) 1 -14.0605 94.2 1.060 1 -14.0605 matchings, 1 -14.0605 2×2×2×2×2 1 -14.0605 set-1. 1 -14.0605 new york, springer. 1 -14.0605 r.l. 1 -14.0605 parallelizable 1 -14.0605 organisers, 1 -14.0605 confirmation ok. 1 -14.0605 corrector 1 -14.0605 justify.v 1 -14.0605 demonstrably 1 -14.0605 (aso) 1 -14.0605 wii−n+1 1 -14.0605 prof-of-concept 1 -14.0605 entered, 1 -14.0605 facts), 1 -14.0605 plain, 1 -14.0605 r´ep´eter agent 1 -14.0605 tokenroles 1 -14.0605 (bird, 1 -14.0605 adj-nn v-subj 1 -14.0605 kuˇcerov´a 1 -14.0605 to−∞. 1 -14.0605 resources, aclijcnlp 1 -14.0605 re-rankings 1 -14.0605 ru) 1 -14.0605 (1973). 1 -14.0605 jailed. 1 -14.0605 bacterium-type 1 -14.0605 given). 1 -14.0605 (pekar 1 -14.0605 summaries? 1 -14.0605 “es gibt” 1 -14.0605 nadya.’) 1 -14.0605 stimulus-response 1 -14.0605 (gruhl 1 -14.0605 surprisal-based 1 -14.0605 (of). 1 -14.0605 38.91 1 -14.0605 (n,x) 1 -14.0605 57%. 1 -14.0605 model.counts(); 1 -14.0605 0.80. 1 -14.0605 automaticsource 1 -14.0605 u2. 1 -14.0605 lexicalized. 1 -14.0605 bpngs 1 -14.0605 (rbf 1 -14.0605 filestore, 1 -14.0605 andψ 1 -14.0605 instantaneously 1 -14.0605 grill. 1 -14.0605 comparisons: 1 -14.0605 intimately 1 -14.0605 (atem 1 -14.0605 74.32% 1 -14.0605 stanford-granularity 1 -14.0605 advisors, 1 -14.0605 74–83, 1 -14.0605 ,. 1 -14.0605 “% urban” 1 -14.0605 (p, 1 -14.0605 posts/responses 1 -14.0605 log-inverse 1 -14.0605 14m 1 -14.0605 carpol 1 -14.0605 monotone. 1 -14.0605 (l¨u 1 -14.0605 (current 1 -14.0605 prescribe 1 -14.0605 deviation). 1 -14.0605 of yamanashi 1 -14.0605 reacting 1 -14.0605 discontinuous) 1 -14.0605 anyway? 1 -14.0605 case-link 1 -14.0605 (td). 1 -14.0605 (wcn), 1 -14.0605 ]”). 1 -14.0605 (burges, 1 -14.0605 60.4% 1 -14.0605 single-step 1 -14.0605 (bottom-right). 1 -14.0605 rapoport, 1 -14.0605 york, yo10 1 -14.0605 memory-and-retrieval 1 -14.0605 wordsimilarity-353 1 -14.0605 pre-authored 1 -14.0605 45.37 9.0825 1 -14.0605 21.50 751k 1 -14.0605 parts-ofspeech 1 -14.0605 schemes; 1 -14.0605 fingerprint, 1 -14.0605 altun 1 -14.0605 “standard” 1 -14.0605 accidental. 1 -14.0605 declarativelyspecified 1 -14.0605 ωj. 1 -14.0605 tools). 1 -14.0605 catvar (categorial 1 -14.0605 832 1 -14.0605 z(fk) 1 -14.0605 towels 1 -14.0605 parser1 1 -14.0605 error-rate. 1 -14.0605 (dice: f(5,185) 1 -14.0605 cliques, 1 -14.0605 pra2d. 1 -14.0605 change” 1 -14.0605 simp. 1 -14.0605 0.19. 1 -14.0605 pn. 1 -14.0605 0.17. 1 -14.0605 (pc) 1 -14.0605 source-specific 1 -14.0605 “are”. 1 -14.0605 evaluationresultsondevelopmentdatasetwithoutgoldmentionboundaries 1 -14.0605 positioning. 1 -14.0605 ‘b’ 1 -14.0605 connective’s 1 -14.0605 60.84 1 -14.0605 beds. 1 -14.0605 downspeak) 1 -14.0605 instant. 1 -14.0605 !arg2 1 -14.0605 matcher (gtm) 1 -14.0605 goodwin, 1 -14.0605 deviation) 1 -14.0605 monica: 1 -14.0605 4e 1 -14.0605 srebrenica massacre 1 -14.0605 bilingual. 1 -14.0605 reverser, 1 -14.0605 thatcomplementizer 1 -14.0605 regularizers. 1 -14.0605 congestion 1 -14.0605 understudy) 1 -14.0605 chander 1 -14.0605 1≤n≤5. 1 -14.0605 want. 1 -14.0605 dep(ti,?). 1 -14.0605 696-735. 1 -14.0605 melm, feat-i 1 -14.0605 pitch. 1 -14.0605 posttest. 1 -14.0605 lowerorder 1 -14.0605 ,...,w 1 -14.0605 fβ=1 roov 1 -14.0605 (laf) 1 -14.0605 questiontype 1 -14.0605 verb? 1 -14.0605 subscription. 1 -14.0605 te rijden/nar 1 -14.0605 punctuation? 1 -14.0605 (finin, 1 -14.0605 ∼ dir(γ) 1 -14.0605 adult-like 1 -14.0605 renaming. 1 -14.0605 7:3 1 -14.0605 misidentify 1 -14.0605 compression) 1 -14.0605 pricing 1 -14.0605 captures, 1 -14.0605 msd ncmsn 1 -14.0605 normalised. 1 -14.0605 “?p 1 -14.0605 chloride... 1 -14.0605 licheng fang 1 -14.0605 revision: 1 -14.0605 -0.012 1 -14.0605 male,female,neuter, 1 -14.0605 original; 1 -14.0605 terp. 1 -14.0605 ˇcov´a,and panevov 1 -14.0605 subordinators 1 -14.0605 sst, sst+bow, 1 -14.0605 inconsistently. 1 -14.0605 +simc-lda-a 1 -14.0605 decoding), 1 -14.0605 33.53 1 -14.0605 experimentalresultsshow 1 -14.0605 comparison), 1 -14.0605 etymology 1 -14.0605 µij(vi,vj) 1 -14.0605 卣敯 ‘former 1 -14.0605 -.50 -.12 1 -14.0605 (among others), 1 -14.0605 fik. 1 -14.0605 503-511. 1 -14.0605 unnatural. 1 -14.0605 x1,...,xn. 1 -14.0605 88.67 1 -14.0605 other_symbol 1 -14.0605 6= end(w¡1) 1 -14.0605 potent 1 -14.0605 personality. 1 -14.0605 radicals 1 -14.0605 f-measure-augmented 1 -14.0605 mles 1 -14.0605 r1),(r2, r1−1)]: 1 -14.0605 all.1 1 -14.0605 task1 usesontonotes2 1 -14.0605 capitalised 1 -14.0605 apfel l¨asst 1 -14.0605 of coling’2010, 1 -14.0605 bengali. 1 -14.0605 categories), 1 -14.0605 [bjr08]: 1 -14.0605 observed; 1 -14.0605 pejoration 1 -14.0605 backstrom 1 -14.0605 modale, 1 -14.0605 45. 1 -14.0605 metrics’ 1 -14.0605 (84.9%), 1 -14.0605 (http://www.ecdf.ed.ac.uk/). 1 -14.0605 concession). 1 -14.0605 1520 2235 1 -14.0605 unlike, 1 -14.0605 highest-quality 1 -14.0605 αs 1 -14.0605 opennlp.maxent3 1 -14.0605 {τk}k−1k=1 1 -14.0605 speaker; 1 -14.0605 giga, 1 -14.0605 97.59 1 -14.0605 prohibited). 1 -14.0605 s´anchez 1 -14.0605 readymades. 1 -14.0605 -0.008 1 -14.0605 multi-core 1 -14.0605 complied 1 -14.0605 tremendously. 1 -14.0605 lawyer’s 1 -14.0605 o’leary, 1 -14.0605 attributions 1 -14.0605 [2009], 1 -14.0605 helpfulness) 1 -14.0605 (hawley 1 -14.0605 ¨ag. 1 -14.0605 ina. 1 -14.0605 ��smq 1 -14.0605 (automatically) 1 -14.0605 chip, 1 -14.0605 continuity. 1 -14.0605 emanuele pianta, 1 -14.0605 yarowsky. 1 -14.0605 constructrule. 1 -14.0605 τι είναι 1 -14.0605 subdialogues. 1 -14.0605 illegal, 1 -14.0605 perceptron) 1 -14.0605 maximised. 1 -14.0605 (role) 1 -14.0605 awarenes. 1 -14.0605 saj10a). 1 -14.0605 boundaries”meansa morebroadconceptthan 1 -14.0605 identifiable. 1 -14.0605 moriyama, 1 -14.0605 jurgens 1 -14.0605 inter-argument 1 -14.0605 lik elihood 1 -14.0605 discouraging 1 -14.0605 isomorphismbetween 1 -14.0605 (oov 1 -14.0605 janyce m. 1 -14.0605 lexically-based 1 -14.0605 pos-labeled 1 -14.0605 (4.4%) 1 -14.0605 content-oriented 1 -14.0605 slowly. 1 -14.0605 digit 4got, 1 -14.0605 (lrec2010), 1 -14.0605 escherichia coli 1 -14.0605 (2011)), 1 -14.0605 svm one-vs-all: 1 -14.0605 responders. 1 -14.0605 out-of-wikipedia 1 -14.0605 srl-aware, 1 -14.0605 obscures 1 -14.0605 rom(x,v1)] 1 -14.0605 inv, 1 -14.0605 “log-linear” 1 -14.0605 assume. 1 -14.0605 grammar-specific 1 -14.0605 2k wud], 1 -14.0605 participant-dependent 1 -14.0605 “mamma mia” 1 -14.0605 center (ftw) 1 -14.0605 two-tailed pearson 1 -14.0605 t-cap 1 -14.0605 qj. 1 -14.0605 ge-task, 1 -14.0605 independence. 1 -14.0605 tanaka. 1 -14.0605 abstract). 1 -14.0605 coreferencelayerin 1 -14.0605 available.8 1 -14.0605 da? 1 -14.0605 wordw. 1 -14.0605 faction 1 -14.0605 pointless 1 -14.0605 sentence-level, 1 -14.0605 francisco 49ers 1 -14.0605 opment 1 -14.0605 p1–p3 1 -14.0605 220. 1 -14.0605 cyc. 1 -14.0605 well-approximated 1 -14.0605 friday. 1 -14.0605 set.4 1 -14.0605 visual, 1 -14.0605 ipus. 1 -14.0605 tree-like 1 -14.0605 participle. 1 -14.0605 d⇒ 1 -14.0605 swapna gottipati 1 -14.0605 (2006b). 1 -14.0605 bi¯albuh˘ a¯ari 1 -14.0605 33.99 1 -14.0605 dualist. 1 -14.0605 truth setst(ϕ) 1 -14.0605 centripetal 1 -14.0605 84%, 1 -14.0605 3.2.1. 1 -14.0605 case-factor 1 -14.0605 1328 1 -14.0605 “crime” 1 -14.0605 you” 1 -14.0605 generatepre-clusters. 1 -14.0605 p(x). 1 -14.0605 tx 76203 1 -14.0605 (0.13%) 1 -14.0605 tail). 1 -14.0605 vandalized 1 -14.0605 expense. 1 -14.0605 liblinear. 1 -14.0605 five, 1 -14.0605 korea busan, 1 -14.0605 turns). 1 -14.0605 morning. 1 -14.0605 choose. 1 -14.0605 sentenceedgeflags 1 -14.0605 (1993)) 1 -14.0605 tig, 1 -14.0605 cgs 1 -14.0605 (2.2%) 1 -14.0605 zscores. 1 -14.0605 (5%) 1 -14.0605 “pass”). 1 -14.0605 (si 1 -14.0605 word-aligned source-parsed 1 -14.0605 implausible). 1 -14.0605 gaming (dr 1 -14.0605 89.78 1 -14.0605 n◦ 247762 1 -14.0605 retranslating 1 -14.0605 0.91). 1 -14.0605 treebanking. 1 -14.0605 25-26. 1 -14.0605 sentences1 1 -14.0605 (implicitly) 1 -14.0605 bionlp09st. 1 -14.0605 individualization 1 -14.0605 condensation. 1 -14.0605 sentencealigned 1 -14.0605 apoptosis 1 -14.0605 (matrix 1 -14.0605 nicolas anelka 1 -14.0605 regulations. 1 -14.0605 sumarize, 1 -14.0605 vectori2l(vectori2)− 1 -14.0605 〉,...,〈s 1 -14.0605 jennifer balogh, 1 -14.0605 1993)). 1 -14.0605 urdu’s 1 -14.0605 61). 1 -14.0605 (inversion 1 -14.0605 1n. 1 -14.0605 77.78% 1 -14.0605 coordhypermero attrieventran.nran.jran.v 1 -14.0605 120–127, 1 -14.0605 demonstratives, 1 -14.0605 (zhuang 1 -14.0605 word/lexeme 1 -14.0605 (69.3). 1 -14.0605 ne). 1 -14.0605 sus dientes 1 -14.0605 breakthrough 1 -14.0605 2708 1 -14.0605 87.32%. 1 -14.0605 kno 1 -14.0605 ap-adapted. 1 -14.0605 canonicalizing 1 -14.0605 unstable. 1 -14.0605 867 1 -14.0605 bon−1(wprime). 1 -14.0605 • entitytypefeatures-thenamedentitytypeforthe 1 -14.0605 believe. 1 -14.0605 avg-source 1 -14.0605 abdessamad 1 -14.0605 ∈x. 1 -14.0605 journalism. 1 -14.0605 (cu-2b) 1 -14.0605 word/character 1 -14.0605 “mission 4636” 1 -14.0605 pmi-score 1 -14.0605 (ctpr) 1 -14.0605 5.0, 1 -14.0605 contemplated 1 -14.0605 site). 1 -14.0605 hawaien 1 -14.0605 datapoint, 1 -14.0605 desc(?z) 1 -14.0605 selfgenerated 1 -14.0605 aids. 1 -14.0605 questioned. 1 -14.0605 datasets.8 1 -14.0605 ho-peng, 1 -14.0605 erk, 1 -14.0605 (vrm) 1 -14.0605 tianyi pavilion 1 -14.0605 and2© 1 -14.0605 (⊗) 1 -14.0605 demonstrator 1 -14.0605 flip. 1 -14.0605 sign-test 1 -14.0605 christian federmann 1 -14.0605 vázquez. 1 -14.0605 raw-toclean 1 -14.0605 le. 1 -14.0605 41.74 1 -14.0605 wods, 1 -14.0605 crp, 1 -14.0605 innocuous 1 -14.0605 lowdimensional 1 -14.0605 opposite: 1 -14.0605 fluctuate around 1 -14.0605 (murthy 1 -14.0605 indecent’. 1 -14.0605 rom,bos]), 1 -14.0605 relational. 1 -14.0605 s–v 1 -14.0605 107–115, 1 -14.0605 depends. 1 -14.0605 divorced 1 -14.0605 damani. 1 -14.0605 holdings. 1 -14.0605 opencyc. 1 -14.0605 75,000 1 -14.0605 populations, 1 -14.0605 (ǫ, ǫ) 1 -14.0605 30.61 1 -14.0605 s100 s101 1 -14.0605 lvc recognition, 1 -14.0605 950 1 -14.0605 balance, 1 -14.0605 nf-ka.. 1 -14.0605 +det 1 -14.0605 widest 1 -14.0605 fixation. 1 -14.0605 spaces; 1 -14.0605 ’, ’was’, 1 -14.0605 full-word 1 -14.0605 ppalli ha-yess-ta. 1 -14.0605 0904684. 1 -14.0605 ’good’ 1 -14.0605 (sukkarieh 1 -14.0605 “fix” 1 -14.0605 5890 1 -14.0605 leter. 1 -14.0605 65.83 1 -14.0605 exercise) 1 -14.0605 acronym-long-form 1 -14.0605 ppis. 1 -14.0605 986 1 -14.0605 lda). 1 -14.0605 null] 1 -14.0605 ho ‘be’, 1 -14.0605 imballanced 1 -14.0605 3http://www.natcorp.ox.ac.uk/ 1 -14.0605 (hernault 1 -14.0605 (2007),and 1 -14.0605 aij 1 -14.0605 round”. 1 -14.0605 (2007a). 1 -14.0605 languageindependent; 1 -14.0605 necessarily, 1 -14.0605 1.8) 49.6 1 -14.0605 errorannotated 1 -14.0605 inuit 1 -14.0605 (9.8gb). 1 -14.0605 www.classic-project.org). 1 -14.0605 non-arguments 1 -14.0605 out-forest. 1 -14.0605 formfunction 1 -14.0605 3.27 1 -14.0605 brule 1 -14.0605 ‘enter 1 -14.0605 home meds 1 -14.0605 h133g080047. 1 -14.0605 localization:t2 theme:t1 1 -14.0605 nm(d) 1 -14.0605 10-million-word 1 -14.0605 irregular. 1 -14.0605 content? 1 -14.0605 estimators. 1 -14.0605 /\$ \s∗/ 1 -14.0605 dir(α). 1 -14.0605 littman’s 1 -14.0605 64–73, 1 -14.0605 ˆnullnull 1 -14.0605 slope. 1 -14.0605 1–6 1 -14.0605 54–62, 1 -14.0605 risky, 1 -14.0605 2002a; 1 -14.0605 trnka, m., 1 -14.0605 randomise 1 -14.0605 ’123’, ’form’: 1 -14.0605 val. 1 -14.0605 (smaller 1 -14.0605 simon. 1 -14.0605 deletes, 1 -14.0605 (complete 1 -14.0605 wmt07 1 -14.0605 holder. 1 -14.0605 “single sample” 1 -14.0605 1992. 1 -14.0605 (1996)). 1 -14.0605 “de mello’s 1 -14.0605 link). 1 -14.0605 “(definec” 1 -14.0605 phrase/rule 1 -14.0605 1 -14.0605 ino(n2), 1 -14.0605 0.6%. 1 -14.0605 (8)) 1 -14.0605 inter-subject 1 -14.0605 glish 1 -14.0605 “groupie”. 1 -14.0605 (saga). 1 -14.0605 logn). 1 -14.0605 contest. 1 -14.0605 58.67 1 -14.0605 o’donnell 1 -14.0605 36.4m 1 -14.0605 agreed. 1 -14.0605 genabith. 1 -14.0605 standard-form 1 -14.0605 floor). 1 -14.0605 ameliorated 1 -14.0605 4096 1 -14.0605 unlike psg, 1 -14.0605 thistask 1 -14.0605 news2011 1 -14.0605 null(nullnullw) null(nullnullnullt)null 1 -14.0605 ph. 1 -14.0605 clusters.) 1 -14.0605 in), 1 -14.0605 argmaxy∈yprime 1 -14.0605 92.00 1 -14.0605 synchronous, multistack 1 -14.0605 infusing 1 -14.0605 52.96 1 -14.0605 collin f., 1 -14.0605 incoming/outgoing 1 -14.0605 twenty-three 1 -14.0605 mturk: 1 -14.0605 empronouns. 1 -14.0605 m4 m5, 1 -14.0605 nathan bodenstab† 1 -14.0605 conservatively. 1 -14.0605 gangemi 1 -14.0605 supervision; 1 -14.0605 irregularity, 1 -14.0605 949 1 -14.0605 kitchen. 1 -14.0605 concerted 1 -14.0605 david kauchak 1 -14.0605 please^rb bring^vb 1 -14.0605 germanc-gs. 1 -14.0605 ‘direction’, 1 -14.0605 (velldal, 1 -14.0605 •these 1 -14.0605 designators. 1 -14.0605 ne) 1 -14.0605 date? 1 -14.0605 bill, cheesecake, 1 -14.0605 routine, 1 -14.0605 billdol@microsoft.com 1 -14.0605 world-wide 1 -14.0605 muslim. 1 -14.0605 ns prompt .001 1 -14.0605 ...], 1 -14.0605 stutterheim, 1 -14.0605 intuit 1 -14.0605 o(v2) 1 -14.0605 (s2net) 1 -14.0605 sfs 1 -14.0605 content-based, network-based, 1 -14.0605 0.04. 1 -14.0605 2001b; harkema, 1 -14.0605 definitions). 1 -14.0605 withstand 1 -14.0605 (2010b): 1 -14.0605 exemplars. 1 -14.0605 prevents, 1 -14.0605 schubert, 190) 1 -14.0605 frames: 1 -14.0605 interchangeable hashtags, 1 -14.0605 0.116, 1 -14.0605 0≤s≤sprime, 0≤u≤uprime, 1 -14.0605 concept-value 1 -14.0605 intertwined 1 -14.0605 maxdifference. 1 -14.0605 256-263. 1 -14.0605 coin. 1 -14.0605 supertags. 1 -14.0605 freya (damljanovic 1 -14.0605 dt,u(r) 1 -14.0605 0.5n [nac 1 -14.0605 sequitur’s 1 -14.0605 mei, 1 -14.0605 electroencephalography (eeg) 1 -14.0605 currency side 1 -14.0605 “i”. 1 -14.0605 fox. 1 -14.0605 base”, 1 -14.0605 fa8750-09-c-0172, 1 -14.0605 48.2%. 1 -14.0605 module) 1 -14.0605 mineditedcost=systemmax; mineditedsug=“” 1 -14.0605 markus dreyer, 1 -14.0605 wg. 1 -14.0605 afterward 1 -14.0605 il 60208 1 -14.0605 wsyn. 1 -14.0605 of ranlp, 2005. 1 -14.0605 sanjika hewavitharana, 1 -14.0605 (ramakrishnan 1 -14.0605 19-51. 1 -14.0605 computers) 1 -14.0605 tiziana ligorio, 1 -14.0605 neut. 1 -14.0605 “forest” 1 -14.0605 ich m¨ochte 1 -14.0605 moses+del 0.6046 1 -14.0605 cade 1 -14.0605 5-tuple (σ, 1 -14.0605 fhmm-based 1 -14.0605 reuse. 1 -14.0605 syntbip-monods. 1 -14.0605 bullshit. 1 -14.0605 “votes” 1 -14.0605 csubi cobjj 1 -14.0605 1203 1 -14.0605 fe). 1 -14.0605 topicexist(e 1 -14.0605 none; 1 -14.0605 stopwords). 1 -14.0605 imation 1 -14.0605 transmissions 1 -14.0605 2,277 1 -14.0605 (6–8) 1 -14.0605 563 53.46 1 -14.0605 60.5%. 1 -14.0605 appearance. 1 -14.0605 (prob.) 1 -14.0605 60903138. 1 -14.0605 goaltest(word 1 -14.0605 450 (+/20) 1 -14.0605 rabiner, l. r. 1 -14.0605 mmaction: 1 -14.0605 cmejrek 1 -14.0605 studer 1 -14.0605 hr0011-06c-0022 1 -14.0605 am-fm: 1 -14.0605 @usernames 1 -14.0605 brian roark† 1 -14.0605 5gm+4gm 1 -14.0605 neighbors: 1 -14.0605 discredit 1 -14.0605 prototypes. 1 -14.0605 2,3,4 1 -14.0605 positives). 1 -14.0605 cross-topic] 1 -14.0605 galen, 1 -14.0605 (widdows, 1 -14.0605 993–1022. 1 -14.0605 self-training (assvm) 1 -14.0605 inapropriate 1 -14.0605 1-87. 1 -14.0605 first). 1 -14.0605 output? 1 -14.0605 .67, 1 -14.0605 h−1) 1 -14.0605 misidentifies 1 -14.0605 deployed, 1 -14.0605 integraldisplay p(θ) 1 -14.0605 discus 1 -14.0605 idiom, 1 -14.0605 9014 1 -14.0605 melissa patchan 1 -14.0605 niche 1 -14.0605 pre-annotate 1 -14.0605 riloff’s 1 -14.0605 susan windisch 1 -14.0605 =〈pi1,...,pik−1〉 1 -14.0605 hearingimpaired 1 -14.0605 candidate). 1 -14.0605 textbased 1 -14.0605 arbitrariness 1 -14.0605 81.00 1 -14.0605 distance/similarity 1 -14.0605 fact-based 1 -14.0605 oleksandr kolomiyets, 1 -14.0605 (fang, 1 -14.0605 tsuchida 1 -14.0605 pairings. 1 -14.0605 wex12@cs.pitt.edu 1 -14.0605 rada, wiebe 1 -14.0605 meronymy/holonymy 1 -14.0605 neat. 1 -14.0605 srilm) 1 -14.0605 cation. 1 -14.0605 20ng. 1 -14.0605 32.00 1 -14.0605 ∗⇒ ubα 1 -14.0605 660 rosedale 1 -14.0605 102–111, 1 -14.0605 rlms. 1 -14.0605 #unks 1 -14.0605 improved? 1 -14.0605 uni-gram, 1 -14.0605 sions 1 -14.0605 atom pred(l1, 1 -14.0605 self-translated 1 -14.0605 (ist-249119), 1 -14.0605 symmetry. 1 -14.0605 pieces. 1 -14.0605 vrij 1 -14.0605 approximate: 1 -14.0605 1(c). 1 -14.0605 [shaping]. 1 -14.0605 trigraphs, 1 -14.0605 99.8 1 -14.0605 property-denoting 1 -14.0605 dni/ini 1 -14.0605 avenue lowell, 1 -14.0605 intractably 1 -14.0605 61.57% 1 -14.0605 paris-est, ligm 1 -14.0605 58.72 1 -14.0605 “hmm” 1 -14.0605 corpora-based 1 -14.0605 minimised 1 -14.0605 import export, 1 -14.0605 length-related 1 -14.0605 resolution; 1 -14.0605 strongly, 1 -14.0605 may. 1 -14.0605 1958) 1 -14.0605 o.. 1 -14.0605 id’s. 1 -14.0605 p(p i,pj) 1 -14.0605 amended 1 -14.0605 document– 1 -14.0605 satisfied). 1 -14.0605 run). 1 -14.0605 peculiarity 1 -14.0605 (ngd) (cilibrasi 1 -14.0605 nakano, 1 -14.0605 veteran democratic 1 -14.0605 redaction. 1 -14.0605 string-to-tree, 1 -14.0605 fixations, 1 -14.0605 pushed. 1 -14.0605 r_bo mbl 1 -14.0605 facilities!! 1 -14.0605 instrumental. 1 -14.0605 yt−1 1 -14.0605 s× ×np 1 -14.0605 hands. 1 -14.0605 overseas. 1 -14.0605 prf-based 1 -14.0605 non-occurrence 1 -14.0605 (argmax 1 -14.0605 platoon 1 -14.0605 .re/ 1 -14.0605 (wt 1 -14.0605 accustomed 1 -14.0605 j na. 1 -14.0605 zahlreiche 1 -14.0605 fine-to-coarse 1 -14.0605 palo alto, 1 -14.0605 decomposition; 1 -14.0605 invitation. 1 -14.0605 regulations, 1 -14.0605 semi-automatically. 1 -14.0605 multilingually 1 -14.0605 0.700.75 0.800.85 1 -14.0605 (4.3) 1 -14.0605 和 (hé) 1 -14.0605 hak. 1 -14.0605 cxy. 1 -14.0605 55.83 1 -14.0605 2009-04-19: 1 -14.0605 forever. 1 -14.0605 dmitriy.dligach@colorado.edu 1 -14.0605 dfus 1 -14.0605 consonant: 1 -14.0605 78712 joeraii@cs.utexas.edu 1 -14.0605 multi-operator 1 -14.0605 pbmt(s,σ,a). 1 -14.0605 misspell 1 -14.0605 discussion: 1 -14.0605 c&c. 1 -14.0605 loria/cnrs lorraine. 1 -14.0605 dld. 1 -14.0605 (today 1 -14.0605 david weir 1 -14.0605 you! 1 -14.0605 wuensch 1 -14.0605 cascade, 1 -14.0605 yunyaoli@us.ibm.com 1 -14.0605 d6 1 -14.0605 “sempos” 1 -14.0605 (finch 1 -14.0605 better-performing 1 -14.0605 87.2% 1 -14.0605 brain. 1 -14.0605 (1 −α) 1 -14.0605 coffee-acc 1 -14.0605 parse: 1 -14.0605 dar duro 1 -14.0605 confidence.” 1 -14.0605 parenthesis, 1 -14.0605 didn’t. 1 -14.0605 rec f-sc 1 -14.0605 close encounters 1 -14.0605 ewn. 1 -14.0605 conclusion). 1 -14.0605 liscombe 1 -14.0605 sentencealignment 1 -14.0605 spain cgomezr@udc.es 1 -14.0605 description; 1 -14.0605 controlled. 1 -14.0605 sell sold 1 -14.0605 wolff, 1 -14.0605 craig saunders, 1 -14.0605 fragment). 1 -14.0605 {−1,0,1}. 1 -14.0605 g396g349g448g286g396 g862g396g349g448g349g287g396g286g863g396g349g448g286g396g856g286g449g374g856g1005 1 -14.0605 satisfaction”. 1 -14.0605 turktest). 1 -14.0605 opinion-oriented 1 -14.0605 multiple r-squared 1 -14.0605 defrayal, defrayment: 1 -14.0605 α2) 1 -14.0605 behavior). 1 -14.0605 pranav anand, 1 -14.0605 bayesian decipherment, 1 -14.0605 leaflets, 1 -14.0605 mellon. 1 -14.0605 sara kiesler 1 -14.0605 probabilistic. 1 -14.0605 w∗, 1 -14.0605 (background) 1 -14.0605 5,800 1 -14.0605 unsurprising, 1 -14.0605 re-de 1 -14.0605 (2bc). 1 -14.0605 3.28 1 -14.0605 (diab, 1 -14.0605 uncorrelated 1 -14.0605 martinez. 1 -14.0605 old hungarian 1 -14.0605 {kapil,kathy}@cs.columbia.edu 1 -14.0605 nwire, 1 -14.0605 punyakanok, 1 -14.0605 killed). 1 -14.0605 expected) 1 -14.0605 [p loss] 1 -14.0605 moved, 1 -14.0605 gray. 1 -14.0605 gopubmed (doms 1 -14.0605 (nakano 1 -14.0605 farmers. 1 -14.0605 infrequently, 1 -14.0605 1320 1 -14.0605 non-associative 1 -14.0605 environmental. 1 -14.0605 blanks 1 -14.0605 rescored, 1 -14.0605 sender. 1 -14.0605 construe 1 -14.0605 psccuny 1 -14.0605 small.8 1 -14.0605 ‘?’ 1 -14.0605 development). 1 -14.0605 c. l.; 1 -14.0605 φi(sourceword, targetword, 1 -14.0605 resulting, 1 -14.0605 personalize 1 -14.0605 forced-choice 1 -14.0605 imprisoned 1 -14.0605 quake. 1 -14.0605 intonation, 1 -14.0605 bilotti 1 -14.0605 nested: 1 -14.0605 tokenisation. 1 -14.0605 rank-correlation 1 -14.0605 s116s97 s103 1 -14.0605 requester 1 -14.0605 0.68. 1 -14.0605 fa8750-09-c-179. 1 -14.0605 s(w(black hole), 1 -14.0605 claim: 1 -14.0605 p−10≤pprime pprime≤p+10 1 -14.0605 aissen, 1 -14.0605 sushi 1 -14.0605 pra2d 1 -14.0605 peers. 1 -14.0605 1 56124 1 -14.0605 annotationpatern 1 -14.0605 (2.3) 1 -14.0605 nadya=ka ıntızar 1 -14.0605 56.61 1 -14.0605 (initially 1 -14.0605 omitted) 1 -14.0605 pst, 1 -14.0605 samaas, 1 -14.0605 wmtsmall model. 1 -14.0605 bag-of-words-based 1 -14.0605 3,3the 7,7regular 1 -14.0605 qºa mahrwn 1 -14.0605 interlocutors. 1 -14.0605 curves: 1 -14.0605 english/czech 1 -14.0605 trig(x) 1 -14.0605 trustable 1 -14.0605 g1850 g1846g3013 1 -14.0605 (mlp). 1 -14.0605 (gellerstam, 1 -14.0605 browser cookie 1 -14.0605 algorithms). 1 -14.0605 mystery. 1 -14.0605 intrusions 1 -14.0605 47.99 1 -14.0605 anova, 1 -14.0605 lhs. 1 -14.0605 linktypebasedpre-clusterpairmodelforcoreferenceresolution yangsong†, 1 -14.0605 f¨urstenau 1 -14.0605 schemespecific 1 -14.0605 medstractplus, 1 -14.0605 (berant 1 -14.0605 bos [cue1] 1 -14.0605 verb-specific. 1 -14.0605 subtitles. 1 -14.0605 (development) 1 -14.0605 disgust (bellegarda, 1 -14.0605 clarity). 1 -14.0605 attentionintheresearchcommunityforalongtime. 1 -14.0605 (maintaining 1 -14.0605 idf, 1 -14.0605 •cprimen 1 -14.0605 (-1.8, 1 -14.0605 author; 1 -14.0605 standpoint. 1 -14.0605 ered 1 -14.0605 damianos karakos 1 -14.0605 erroneously. 1 -14.0605 disastrous 1 -14.0605 (mochihashi 1 -14.0605 hill’s 1 -14.0605 prices. 1 -14.0605 http://www.internetworldstats.com/stats.htm 1 -14.0605 2any 1 -14.0605 38.9% 1 -14.0605 menjalani kehidupan 1 -14.0605 (golan, 1 -14.0605 talkamatic ab. 1 -14.0605 sbml/celldesigner pathways 1 -14.0605 {mr} 1 -14.0605 ficar triste 1 -14.0605 up? 1 -14.0605 gray) 1 -14.0605 dyslexia. 1 -14.0605 nelson, 1 -14.0605 yj.k 1 -14.0605 subderivation 1 -14.0605 under searn. 1 -14.0605 set-subset 1 -14.0605 noisy-seed) .148 1 -14.0605 predicted opi=2.3) 1 -14.0605 extrinsically, 1 -14.0605 94.84 1 -14.0605 v.good 1 -14.0605 run-time. 1 -14.0605 shorthand, 1 -14.0605 publics. 1 -14.0605 ‘/’ 0.333333 -14.0605 moens, 0.6 -14.0605 depu 0.6 -14.0605 satire 0.142857 -14.0607 an annotation 0.444444 -14.0612 flowers 0.5 -14.0613 expressives, 0.5 -14.0613 .46 0.5 -14.0613 lane 0.5 -14.0613 wang. 0.5 -14.0613 of). 0.666667 -14.0616 (blackstone 0.666667 -14.0616 summarizing, 0.666667 -14.0616 prose. 0.666667 -14.0616 receivers 0.666667 -14.0616 egy 0.666667 -14.0616 svr-base. 0.666667 -14.0616 (~) 0.666667 -14.0616 (2a)). 0.666667 -14.0616 63.78 0.666667 -14.0616 (hopcroft 0.666667 -14.0616 possible). 0.666667 -14.0616 virtual nonterminals. 0.666667 -14.0616 17-19, 0.666667 -14.0616 pace. 0.666667 -14.0616 lemma-based 0.666667 -14.0616 67.78 0.666667 -14.0616 tensored 0.666667 -14.0616 74.67 0.666667 -14.0616 55.00 0.666667 -14.0616 attack) 0.666667 -14.0616 humanproduced 0.666667 -14.0616 plag 0.666667 -14.0616 kong. 0.666667 -14.0616 pdas. 0.666667 -14.0616 (farhadi 0.666667 -14.0616 54–63, 0.666667 -14.0616 inclusive. 0.666667 -14.0616 news-trained 0.666667 -14.0616 ζ(z) 0.666667 -14.0616 ssa. 0.666667 -14.0616 s.t 0.666667 -14.0616 4a. 0.666667 -14.0616 perceptrons. 0.666667 -14.0616 (frequent 0.666667 -14.0616 42.67 0.666667 -14.0616 20082010. 0.666667 -14.0616 delaying 0.666667 -14.0616 dutch ’6 0.666667 -14.0616 61%. 0.666667 -14.0616 (2003)’s 0.666667 -14.0616 prohibited. 0.666667 -14.0616 86.0% 0.666667 -14.0616 penn. 0.666667 -14.0616 minconf. 0.666667 -14.0616 recorded action. 0.666667 -14.0616 aggregation. 0.666667 -14.0616 93 94 0.666667 -14.0616 exert 0.666667 -14.0616 scientists. 0.666667 -14.0616 .95, 0.666667 -14.0616 (bharati 0.666667 -14.0616 glance. 0.666667 -14.0616 similes. 0.666667 -14.0616 becomes. 0.666667 -14.0616 credence 0.666667 -14.0616 rearrangement 0.666667 -14.0616 58.61 0.666667 -14.0616 ratio). 0.666667 -14.0616 (25 0.666667 -14.0616 hd. 0.666667 -14.0616 linguistically–driven 0.666667 -14.0616 gerund. 0.666667 -14.0616 loosing 0.666667 -14.0616 cold. 0.666667 -14.0616 applications). 0.666667 -14.0616 schinner 0.666667 -14.0616 shafto 0.666667 -14.0616 -0.61 0.666667 -14.0616 lsat 0.666667 -14.0616 ruminants 0.666667 -14.0616 lh3(null) 0.666667 -14.0616 compositions, 0.666667 -14.0616 0null0001) 0.666667 -14.0616 not, no, 0.666667 -14.0616 92.73 0.666667 -14.0616 antibiotic 0.666667 -14.0616 forthcoming 0.666667 -14.0616 reduction). 0.666667 -14.0616 michigan. 0.666667 -14.0616 nist). 0.666667 -14.0616 violated, 0.666667 -14.0616 watermarked. 0.666667 -14.0616 genre? 0.666667 -14.0616 capital. 0.666667 -14.0616 non-zero. 0.666667 -14.0616 fixedness. 0.666667 -14.0616 byte 0.666667 -14.0616 finkelstein 0.666667 -14.0616 grouped, 0.666667 -14.0616 experiencer, 0.666667 -14.0616 visibly 0.666667 -14.0616 opponents, 0.666667 -14.0616 (1995)). 0.666667 -14.0616 combination) 0.666667 -14.0616 conj. 0.666667 -14.0616 (jourde 0.666667 -14.0616 +0.1 0.666667 -14.0616 existed. 0.666667 -14.0616 0.20. 0.666667 -14.0616 jobs. 0.666667 -14.0616 1.1m 0.666667 -14.0616 kim. 0.666667 -14.0616 3for example, 0.666667 -14.0616 .008 0.666667 -14.0616 biscuits. 0.666667 -14.0616 one-toone 0.666667 -14.0616 classes; 0.666667 -14.0616 2004a). 0.666667 -14.0616 (4)). 0.666667 -14.0616 djk 0.666667 -14.0616 wbow. 0.666667 -14.0616 rap 0.666667 -14.0616 murata 0.666667 -14.0616 letter choices 0.666667 -14.0616 unconfident 0.666667 -14.0616 rigau. 0.666667 -14.0616 44.4% 0.666667 -14.0616 tokenizations 0.666667 -14.0616 greatly, 0.666667 -14.0616 pruning) 0.666667 -14.0616 meaning? 0.666667 -14.0616 users). 0.666667 -14.0616 (nederhof 0.666667 -14.0616 (helbig, 0.666667 -14.0616 metric). 0.666667 -14.0616 reconstruction. 0.666667 -14.0616 codeword. 0.666667 -14.0616 (tl) 0.666667 -14.0616 aε. 0.666667 -14.0616 85.0% 0.666667 -14.0616 f(y∗). 0.666667 -14.0616 clairlib, 0.666667 -14.0616 0.1◦ 0.666667 -14.0616 blood cells 0.666667 -14.0616 90.89 0.666667 -14.0616 (felzenszwalb 0.666667 -14.0616 tac. 0.666667 -14.0616 66.99 0.666667 -14.0616 sec./sent. 0.666667 -14.0616 spell-checker 0.666667 -14.0616 macquarie, 0.666667 -14.0616 resnik (2008). 0.666667 -14.0616 event-event 0.666667 -14.0616 αh 0.666667 -14.0616 restrictors 0.666667 -14.0616 ages. 0.666667 -14.0616 perpetrator, 0.666667 -14.0616 (kuo 0.666667 -14.0616 till now, 0.666667 -14.0616 69%, 0.666667 -14.0616 lower-cased, 0.666667 -14.0616 weiss 0.666667 -14.0616 98%, 0.666667 -14.0616 aggressively. 0.666667 -14.0616 reviewer. 0.666667 -14.0616 phones) 0.666667 -14.0616 77.73 0.666667 -14.0616 socialist 0.666667 -14.0616 unambiguously. 0.666667 -14.0616 questions; 0.666667 -14.0616 polygons. 0.666667 -14.0616 (except stop 0.666667 -14.0616 ...si−1) 0.666667 -14.0616 segmentation; 0.666667 -14.0616 prompts. 0.666667 -14.0616 øj. 0.666667 -14.0616 huffaker 0.666667 -14.0616 carry. 0.666667 -14.0616 ftib. 0.666667 -14.0616 morphological operations, 0.666667 -14.0616 angle. 0.666667 -14.0616 cross-entropy. 0.666667 -14.0616 confounder, 0.666667 -14.0616 long; 0.666667 -14.0616 39.08 0.666667 -14.0616 15), 0.666667 -14.0616 product moment 0.666667 -14.0616 lowfrequency 0.666667 -14.0616 (arabic, 0.666667 -14.0616 priate 0.666667 -14.0616 electronics. 0.666667 -14.0616 interesting) 0.666667 -14.0616 (sc). 0.666667 -14.0616 208), 0.666667 -14.0616 501 0.666667 -14.0616 tseng, 0.666667 -14.0616 belief-updating 0.666667 -14.0616 92.0% 0.666667 -14.0616 59.8% 0.666667 -14.0616 iterations; 0.666667 -14.0616 see table 2). 0.666667 -14.0616 inspections 0.666667 -14.0616 computationally. 0.666667 -14.0616 current, 0.666667 -14.0616 90.93 0.666667 -14.0616 τ.f 0.666667 -14.0616 offline. 0.666667 -14.0616 preprocessed, 0.666667 -14.0616 sempos tags 0.666667 -14.0616 finite 0.666667 -14.0616 credit. 0.666667 -14.0616 ∈c. 0.666667 -14.0616 out-domain 0.666667 -14.0616 appointed 0.666667 -14.0616 4.95 0.666667 -14.0616 bates, 0.666667 -14.0616 to). 0.666667 -14.0616 degenerates 0.666667 -14.0616 before’ 0.666667 -14.0616 (hvs) 0.666667 -14.0616 walkers 0.666667 -14.0616 (l), 0.666667 -14.0616 83.99 0.666667 -14.0616 42.86 0.666667 -14.0616 nonconvex 0.666667 -14.0616 argumentation. 0.666667 -14.0616 imp. 0.666667 -14.0616 cross-validation). 0.666667 -14.0616 lehr 0.666667 -14.0616 mellish 0.666667 -14.0616 /! 0.666667 -14.0616 (9.8gb), 0.666667 -14.0616 0.07). 0.666667 -14.0616 station. 0.666667 -14.0616 co-hyponymy 0.666667 -14.0616 p.c.). 0.666667 -14.0616 (agar, 0.666667 -14.0616 cross correlation. 0.666667 -14.0616 worse). 0.666667 -14.0616 grf621008, 0.666667 -14.0616 algorithm1. 0.666667 -14.0616 িব 0.666667 -14.0616 reservation. 0.666667 -14.0616 ubiub & 0.666667 -14.0616 joke. 0.666667 -14.0616 (darwish 0.666667 -14.0616 59.91 0.666667 -14.0616 p-level≤0.10 0.666667 -14.0616 high-pmi 0.666667 -14.0616 0.52. 0.666667 -14.0616 558 0.666667 -14.0616 unplugged. 0.666667 -14.0616 flagging 0.666667 -14.0616 604 0.666667 -14.0616 0.836 0.666667 -14.0616 cky, 0.666667 -14.0616 bleu_w. 0.666667 -14.0616 (trento), 0.666667 -14.0616 clarification. 0.666667 -14.0616 gothenburg. 0.666667 -14.0616 sight. 0.666667 -14.0616 86.14 0.666667 -14.0616 notation). 0.666667 -14.0616 1998)). 0.666667 -14.0616 79.89 0.666667 -14.0616 bestsinglesystem. 0.666667 -14.0616 clang. 0.666667 -14.0616 ulisse. 0.666667 -14.0616 70.8% 0.666667 -14.0616 conllsharedtask. 0.666667 -14.0616 impair 0.666667 -14.0616 potter 0.666667 -14.0616 offset. 0.666667 -14.0616 you.” 0.666667 -14.0616 85.73 0.666667 -14.0616 u(c). 0.666667 -14.0616 legislation. 0.666667 -14.0616 boruta 0.666667 -14.0616 fstructure 0.666667 -14.0616 viable. 0.666667 -14.0616 summarization? 0.666667 -14.0616 partially, 0.666667 -14.0616 82.62 0.666667 -14.0616 specialists. 0.666667 -14.0616 :λg.λf.λx.g(x)∧f(x)a1〈e,t〉1 0.666667 -14.0616 ipu 0.666667 -14.0616 superlatives. 0.666667 -14.0616 janssen 0.666667 -14.0616 66.74 0.666667 -14.0616 hier mod 0.666667 -14.0616 1the annotations 0.666667 -14.0616 wavgf 0.666667 -14.0616 unregularized 0.666667 -14.0616 n(w) 0.666667 -14.0616 other packages 0.666667 -14.0616 2.4, 0.666667 -14.0616 massive, 0.666667 -14.0616 outputs) 0.666667 -14.0616 instructor, 0.666667 -14.0616 consecutive, 0.666667 -14.0616 (deneefe 0.666667 -14.0616 database). 0.666667 -14.0616 42.08 0.666667 -14.0616 30.00 0.666667 -14.0616 korean). 0.666667 -14.0616 phrases), 0.666667 -14.0616 parse-augmented 0.666667 -14.0616 bat. 0.666667 -14.0616 controversial, 0.666667 -14.0616 .946/.947 0.666667 -14.0616 ontonotesdata. 0.666667 -14.0616 author). 0.666667 -14.0616 (4.33 + 0.666667 -14.0616 wnd 0.666667 -14.0616 (cerini 0.666667 -14.0616 utsumi, 0.666667 -14.0616  crumbs. 0.666667 -14.0616 posts) 0.666667 -14.0616 proportions). 0.666667 -14.0616 shook 0.666667 -14.0616 cannot, 0.666667 -14.0616 lexical, morpho-syntactic 0.666667 -14.0616 heard. 0.666667 -14.0616 e-rater. 0.666667 -14.0616 keyphrase. 0.666667 -14.0616 (lgc) 0.666667 -14.0616 maxpaths, 0.666667 -14.0616 pragmatic neighborhood 0.666667 -14.0616 sense). 0.666667 -14.0616 informed. 0.666667 -14.0616 alan r. 0.666667 -14.0616 procrustes alignment 0.666667 -14.0616 (de-en). 0.666667 -14.0616 forbus 0.666667 -14.0616 globe, 0.666667 -14.0616 let’s suppose 0.666667 -14.0616 94.0% 0.666667 -14.0616 hf. 0.666667 -14.0616 uninterrupted 0.666667 -14.0616 more-than-a-day 0.666667 -14.0616 notions. 0.666667 -14.0616 41.78 0.666667 -14.0616 rembanchs@i2r.a-star.edu.sg 0.666667 -14.0616 −1). 0.666667 -14.0616 (aer) 0.666667 -14.0616 mobility 0.666667 -14.0616 asigns 0.666667 -14.0616 customers. 0.666667 -14.0616 remained. 0.666667 -14.0616 toolkit1. 0.666667 -14.0616 raes 0.666667 -14.0616 most-true, 0.666667 -14.0616 meaningoriented 0.666667 -14.0616 extend. 0.666667 -14.0616 real-time, 0.666667 -14.0616 (large) 0.666667 -14.0616 navigation strategy, 0.666667 -14.0616 features); 0.666667 -14.0616 ins(zuv) 0.666667 -14.0616 5.4.2 0.666667 -14.0616 x”. 0.666667 -14.0616 puting 0.666667 -14.0616 cafarella 0.666667 -14.0616 candidate; 0.666667 -14.0616 female) 0.666667 -14.0616 spanish (es), 0.666667 -14.0616 locations). 0.666667 -14.0616 james r. 0.666667 -14.0616 football. 0.666667 -14.0616 apprentices. 0.666667 -14.0616 haydn on 0.666667 -14.0616 (katz-brown 0.666667 -14.0616 leftbranching 0.666667 -14.0616 reevaluate 0.666667 -14.0616 undirected-cky 0.666667 -14.0616 x2. 0.666667 -14.0616 out). 0.666667 -14.0616 smoothed priors 0.666667 -14.0616 segmentor, 0.666667 -14.0616 enforced. 0.666667 -14.0616 nissim, 0.666667 -14.0616 gi, 0.666667 -14.0616 3an 0.666667 -14.0616 brockett. 0.666667 -14.0616 overfit. 0.666667 -14.0616 wordsall 0.666667 -14.0616 optimisation. 0.666667 -14.0616 test10. 0.666667 -14.0616 onlyall 0.666667 -14.0616 remarkable. 0.666667 -14.0616 gc 0.666667 -14.0616 reverse. 0.666667 -14.0616 wightman 0.666667 -14.0616 branches, 0.666667 -14.0616 german accent 0.666667 -14.0616 instanceinstance 0.666667 -14.0616 6.1.1 0.666667 -14.0616 (st). 0.666667 -14.0616 high.a 0.666667 -14.0616 net. 0.666667 -14.0616 speech-trained 0.666667 -14.0616 caja (you 0.666667 -14.0616 spends 0.666667 -14.0616 “a.zampolli” 0.666667 -14.0616 insects 0.666667 -14.0616 dq. 0.666667 -14.0616 @p” (e.g. 0.666667 -14.0616 onull. 0.666667 -14.0616 appendix b 0.666667 -14.0616 poor, negative, 0.666667 -14.0616 pathological cases 0.666667 -14.0616 language-model 0.666667 -14.0616 preterminals. 0.666667 -14.0616 distracting 0.666667 -14.0616 v3. 0.666667 -14.0616 word-byword 0.666667 -14.0616 found). 0.666667 -14.0616 setiawan 0.666667 -14.0616 inter-correlations 0.666667 -14.0616 sub-genres 0.666667 -14.0616 63.3% 0.666667 -14.0616 tanh 0.666667 -14.0616 t∗. 0.666667 -14.0616 version). 0.666667 -14.0616 detect inconsistencies 0.666667 -14.0616 himself. 0.666667 -14.0616 esg 0.666667 -14.0616 ctpr. 0.666667 -14.0616 less). 0.666667 -14.0616 html5 0.666667 -14.0616 continuum. 0.666667 -14.0616 overload. 0.666667 -14.0616 glf 0.666667 -14.0616 41.14 0.666667 -14.0616 ezekiel. 0.666667 -14.0616 (num. 0.666667 -14.0616 sbar. 0.666667 -14.0616 (contractor 0.666667 -14.0616 path) 0.666667 -14.0616 holistically 0.666667 -14.0616 sports. 0.666667 -14.0616 separable. 0.666667 -14.0616 decoration 0.666667 -14.0616 preceding, 0.666667 -14.0616 24) 0.666667 -14.0616 nts 0.666667 -14.0616 (2009)]. 0.666667 -14.0616 ing. 0.666667 -14.0616 traces. 0.666667 -14.0616 something). 0.666667 -14.0616 ginter 0.666667 -14.0616 1,235 0.666667 -14.0616 strtm. 0.666667 -14.0616 important) 0.666667 -14.0616 198). 0.666667 -14.0616 .001; 0.666667 -14.0616 oblivious 0.666667 -14.0616 qian 0.666667 -14.0616 kink 0.666667 -14.0616 assoc(.) 0.666667 -14.0616 81.91 0.666667 -14.0616 αt. 0.666667 -14.0616 variables; 0.666667 -14.0616 m} 0.666667 -14.0616 (moore, 0.666667 -14.0616 viterbi-like, 0.666667 -14.0616 tech. 0.666667 -14.0616 1957) 0.666667 -14.0616 ipu-final 0.666667 -14.0616 mvm−100m−0.1−0.01 0.666667 -14.0616 do), 0.666667 -14.0616 acom 0.666667 -14.0616 bowen 0.666667 -14.0616 kneser 0.666667 -14.0616 (kulkarni 0.666667 -14.0616 rare: 0.666667 -14.0616 subject). 0.666667 -14.0616 revsimilarfj. 0.666667 -14.0616 memory overhead 0.666667 -14.0616 netflix 0.666667 -14.0616 duplication. 0.666667 -14.0616 electromagnetic 0.666667 -14.0616 80.8% 0.666667 -14.0616 (pmi). 0.196078 -14.0621 effect. 0.5 -14.0621 tolerates 0.4 -14.0625 9000 0.6 -14.0625 73.26 0.6 -14.0625 478 0.666667 -14.0627 4.3% 0.666667 -14.0627 s-factor, 0.666667 -14.0627 volunteered 0.666667 -14.0627 61.25%, 0.666667 -14.0627 adam l. 0.666667 -14.0627 qlf 0.666667 -14.0627 below 2−20 0.666667 -14.0627 2-4, 0.666667 -14.0627 rearranging 0.666667 -14.0627 (mcp) 0.666667 -14.0627 51.61 0.666667 -14.0627 wmt-10 0.666667 -14.0627 it appearsin 0.666667 -14.0627 sender’s 0.666667 -14.0627 too) 0.666667 -14.0627 ad-hoc rules 0.666667 -14.0627 ‘discourse 0.666667 -14.0627 co university 0.666667 -14.0627 linearisation 0.666667 -14.0627 graders, 0.666667 -14.0627 0.407 0.666667 -14.0627 21.5 0.666667 -14.0627 petit 0.666667 -14.0627 sugesting 0.666667 -14.0627 homoplasy, 0.666667 -14.0627 hollingshead, 0.666667 -14.0627 0.666667 -14.0627 10-fold crossvalidation. 0.666667 -14.0627 1150 0.666667 -14.0627 versus: 0.666667 -14.0627 transductive svms 0.666667 -14.0627 through. 0.666667 -14.0627 (sp). 0.666667 -14.0627 (magnanti 0.666667 -14.0627 alryad 0.666667 -14.0627 i am ... 0.666667 -14.0627 particle verbs 0.666667 -14.0627 highest-scoring alignment 0.666667 -14.0627 constraint function: 0.666667 -14.0627 inter-category 0.666667 -14.0627 new’ 0.666667 -14.0627 top-down semantics 0.666667 -14.0627 learning (conll) 0.666667 -14.0627 09 0.666667 -14.0627 29.89 0.666667 -14.0627 woeparse. 0.666667 -14.0627 avatars, 0.666667 -14.0627 (proportion 0.666667 -14.0627 enron. 0.666667 -14.0627 jackendoff 0.666667 -14.0627 machine-produced 0.666667 -14.0627 content). 0.666667 -14.0627 seriously. 0.666667 -14.0627 28.96 0.666667 -14.0627 field lscript 0.666667 -14.0627 r-w reftl 0.666667 -14.0627 smooth. 0.666667 -14.0627 f-ratio 0.666667 -14.0627 bigrams+ 0.666667 -14.0627 early-switching lateen 0.666667 -14.0627 72 28 0.666667 -14.0627 course titles 0.666667 -14.0627 sources” 0.666667 -14.0627 resumed 0.666667 -14.0627 comparable). 0.666667 -14.0627 ucsc 0.666667 -14.0627 earth. 0.666667 -14.0627 (2a). 0.666667 -14.0627 wna, 0.666667 -14.0627 =vector0} 0.666667 -14.0627 negation; 0.666667 -14.0627 plan” 0.666667 -14.0627 dictcorp 0.666667 -14.0627 8gb 0.666667 -14.0627 deemter, 0.666667 -14.0627 merits. 0.666667 -14.0627 28.99 0.666667 -14.0627 meta-judge 0.666667 -14.0627 sped 0.666667 -14.0627 italk 0.666667 -14.0627 koehn, 2007; 0.666667 -14.0627 são paulo 0.666667 -14.0627 expressive, 0.666667 -14.0627 55.84 0.666667 -14.0627 64.67 0.666667 -14.0627 human user: 0.666667 -14.0627 59.72 0.666667 -14.0627 vmeasure 0.666667 -14.0627 jump distances 0.666667 -14.0627 standard lda. 0.666667 -14.0627 cotton. 0.666667 -14.0627 39.61 0.666667 -14.0627 p(z). 0.666667 -14.0627 component), 0.666667 -14.0627 surfacelevel 0.666667 -14.0627 (allfeatures) 0.666667 -14.0627 52.62 0.666667 -14.0627 rte3 dataset 0.666667 -14.0627 coreferenceresolutionsystems. 0.666667 -14.0627 unknown targets. 0.666667 -14.0627 92.89 0.666667 -14.0627 (kumar & 0.666667 -14.0627 67.5% 0.666667 -14.0627 continual 0.666667 -14.0627 separate. 0.5 -14.0629 (message 0.444444 -14.0634 word-pos 0.6 -14.0635 rsvm 0.5 -14.0638 sophia 0.666667 -14.0639 ccg parsing, 0.666667 -14.0639 citation-based summarization. 0.666667 -14.0639 impart 0.666667 -14.0639 hand-tuned 0.666667 -14.0639 revsimilarfi 0.666667 -14.0639 oracle reranker 0.666667 -14.0639 internal modification 0.666667 -14.0639 subsumption, 0.666667 -14.0639 web-scale ngram 0.666667 -14.0639 top-level interpolation 0.666667 -14.0639 16,358 0.666667 -14.0639 subjective nouns 0.666667 -14.0639 +kt) 0.666667 -14.0639 pa rse 0.666667 -14.0639 buyers 0.666667 -14.0639 generic biblical 0.666667 -14.0639 stevenson. 0.666667 -14.0639 j. och 0.666667 -14.0639 cros-domain 0.666667 -14.0639 (xi,xj), 0.666667 -14.0639 discourse studies, 0.666667 -14.0639 each smdp 0.666667 -14.0639 (etl) 0.666667 -14.0639 59.62 0.666667 -14.0639 vocative, 0.666667 -14.0639 translation spotting. 0.666667 -14.0639 unitex 0.666667 -14.0639 2010), usfd 0.666667 -14.0639 0.611 0.666667 -14.0639 critique 0.666667 -14.0639 jirs 0.666667 -14.0639 mental modeling 0.666667 -14.0639 file: 0.666667 -14.0639 null-gram 0.666667 -14.0639 class-label correlation 0.666667 -14.0639 koester 0.666667 -14.0639 “incorrect 0.666667 -14.0639 desde tu 0.666667 -14.0639 [object] 0.666667 -14.0639 cluto’s 0.666667 -14.0639 (rink 0.666667 -14.0639 sets), 0.666667 -14.0639 dembe 0.666667 -14.0639 barnard 0.666667 -14.0639 multilingual subjective 0.666667 -14.0639 discreet / 0.259259 -14.064 fellowship 0.4 -14.0645 surgery 0.6 -14.0645 (explained 0.444444 -14.0649 baltic 0.666667 -14.065 74.3% 0.666667 -14.065 sch¨utze (1998) 0.666667 -14.065 particulars 0.666667 -14.065 argument) 0.666667 -14.065 government purposes 0.666667 -14.065 inequality constraints 0.666667 -14.065 triggerless 0.666667 -14.065 null5, 0.666667 -14.065 technical challenges 0.666667 -14.065 comparable fora 0.666667 -14.065 research direction, 0.666667 -14.065 reformulations, 0.666667 -14.065 collaboration group 0.666667 -14.065 feng (2009) 0.666667 -14.065 0.585 0.666667 -14.065 57.8% 0.666667 -14.065 zipfian distribution 0.666667 -14.065 her peer 0.666667 -14.065 wmpbleu 0.666667 -14.065 74.6% 0.666667 -14.065 person (e.g. 0.6 -14.0655 others: 0.6 -14.0655 anderson 0.384615 -14.0656 alessandro 0.666667 -14.0661 members) 0.666667 -14.0661 participating groups 0.666667 -14.0661 official policies, 0.666667 -14.0661 social relationships, 0.666667 -14.0661 icc 0.666667 -14.0661 84.8% 0.666667 -14.0661 starting points. 0.666667 -14.0661 3.1 tokenisation 0.666667 -14.0661 thereof; 0.666667 -14.0661 0.9 0.91 0.666667 -14.0661 kernels) 0.666667 -14.0661 5-grams, 0.666667 -14.0661 2003; mohammad 0.666667 -14.0661 zig-zag 0.666667 -14.0661 sentential evidence 0.666667 -14.0661 2084 0.666667 -14.0661 visual cues. 0.666667 -14.0661 noisy-and, 0.666667 -14.0661 correspon 0.666667 -14.0661 triggered, 0.666667 -14.0661 (97%) 0.666667 -14.0661 leo wanner, 0.666667 -14.0661 straight lines 0.25 -14.0663 any new 0.5 -14.0663 elena 0.5 -14.0663 0.710 0.5 -14.0663 figure2 0.357143 -14.0664 rhs 0.6 -14.0665 figure3 0.6 -14.0665 1002 0.333333 -14.0666 reviewers. 0.196078 -14.0667 using multiple 0.357143 -14.067 versa, 0.259259 -14.0671 image. 0.666667 -14.0672 id track 0.666667 -14.0672 usp provides 0.666667 -14.0672 casacuberta, 0.666667 -14.0672 table 13: 0.666667 -14.0672 57.9% 0.666667 -14.0672 roll-out 0.666667 -14.0672 72.8% 0.666667 -14.0672 bonnie webber 0.666667 -14.0672 multiple decoders, 0.666667 -14.0672 special care 0.666667 -14.0672 (ren) 0.666667 -14.0672 chord sets 0.666667 -14.0672 0.523 0.6 -14.0675 shots 0.5 -14.068 98.3 0.666667 -14.0683 ji (2009) 0.666667 -14.0683 triage application 0.666667 -14.0683 201 2 0.666667 -14.0683 feature vector template 0.666667 -14.0683 (ppt) 0.666667 -14.0683 tightening procedure 0.666667 -14.0683 of... 0.666667 -14.0683 0.685 0.666667 -14.0683 {λk}k−1k=1 0.666667 -14.0683 j. r. 0.666667 -14.0683 pos ngrams, 0.666667 -14.0683 profitable, 0.666667 -14.0683 functions fs 0.666667 -14.0683 0.957 0.25 -14.0684 another interesting 0.6 -14.0685 insect 0.6 -14.0685 ‘in’ 0.357143 -14.0688 paris, 0.666667 -14.0694 restatement 0.666667 -14.0694 23.3% 0.666667 -14.0694 external knowledge. 0.666667 -14.0694 retrieval (clir) 0.666667 -14.0694 generating instructions 0.666667 -14.0694 abstracts away 0.666667 -14.0694 868 0.666667 -14.0694 dative, 0.666667 -14.0694 businesses, 0.666667 -14.0694 paraphrase harvesting 0.666667 -14.0694 (or concepts) 0.666667 -14.0694 0.850 0.666667 -14.0694 785 0.666667 -14.0694 exposure, 0.666667 -14.0694 switching variable 0.666667 -14.0694 independent) 0.666667 -14.0694 information-rich 0.0652174 -14.0695 word, 0.5 -14.0696 all-edges 0.5 -14.0705 550 0.666667 -14.0706 summarizing decisions 0.666667 -14.0706 turn completions 0.666667 -14.0706 machine translation), 0.666667 -14.0706 previous posts. 0.666667 -14.0706 (mira) 0.666667 -14.0706 roth, 1999). 0.666667 -14.0706 49.54 0.666667 -14.0706 2) retrieve 0.666667 -14.0706 (natural language 0.0778816 -14.0708 score. 0.444444 -14.0709 dépecer 0.333333 -14.0711 chocolate 0.333333 -14.0711 bitext. 0.5 -14.0713 non-negative, 0.5 -14.0713 average precision, 0.169231 -14.0714 the u.s. 0.666667 -14.0717 rouge-be 0.666667 -14.0717 98.8 0.666667 -14.0717 projects like 0.666667 -14.0717 mcmurray, 0.666667 -14.0717 2×105 0.666667 -14.0717 charniak, 2008), 0.666667 -14.0717 sound units 0.666667 -14.0717 quite simple. 0.666667 -14.0717 ap+gmf2 0.666667 -14.0717 creoleenglish 0.666667 -14.0717 0.884 0.666667 -14.0717 is-a relations 0.666667 -14.0717 yeah, 0.4 -14.0719 faust+all 0.384615 -14.0721 w1 w2 0.135922 -14.0722 importantly, 0.444444 -14.0724 much richer 0.4 -14.0726 jari 0.209302 -14.0727 has some 0.666667 -14.0728 34.4% 0.666667 -14.0728 experimentations 0.666667 -14.0728 68.8% 0.666667 -14.0728 dimplek 0.666667 -14.0728 (in of) 0.666667 -14.0728 0.567 0.666667 -14.0728 jackendoff, 0.666667 -14.0728 82.7% 0.666667 -14.0728 rule “dp→d 0.666667 -14.0728 opens many 0.5 -14.073 influenza, 0.4 -14.0732 apis 0.6 -14.0735 rashmi 0.6 -14.0735 isceoof 0.357143 -14.0736 2009a). 0.666667 -14.0739 uc, 0.666667 -14.0739 harrington, 0.666667 -14.0739 ψ(z−1)i 0.666667 -14.0739 (42) 0.666667 -14.0739 workers must 0.666667 -14.0739 (sall) 0.666667 -14.0739 arora, 0.666667 -14.0739 logreg 0.666667 -14.0739 ordinarily 0.666667 -14.0739 24.0% 0.666667 -14.0739 0.809 0.666667 -14.0739 semeval-2010 task 13: 0.666667 -14.0739 saw, 0.666667 -14.0739 plan-correction can 0.5 -14.0747 could support 0.666667 -14.0751 strongly influence 0.666667 -14.0751 recovery, 0.666667 -14.0751 management acts 0.666667 -14.0751 duc2007 0.666667 -14.0751 hr-scfg 0.666667 -14.0751 2θ 0.666667 -14.0751 louisiana 0.666667 -14.0751 archeological 0.666667 -14.0751 currently exist 0.666667 -14.0751 kidnapped 0.666667 -14.0751 (quattoni 0.666667 -14.0751 ×v →r 0.666667 -14.0751 offset) 0.666667 -14.0751 expand sets 0.666667 -14.0751 10,417 0.666667 -14.0762 universities, 0.666667 -14.0762 lo, 0.666667 -14.0762 gold event, 0.666667 -14.0762 (tp) 0.666667 -14.0762 nominalised 0.666667 -14.0762 0.877 0.666667 -14.0762 ogy 0.666667 -14.0762 supervised summaries. 0.666667 -14.0762 conventional hiero 0.666667 -14.0762 7with 0.666667 -14.0762 triggering behavior 0.6 -14.0766 adjective: 0.285714 -14.0768 generates an 0.666667 -14.0773 1128 0.666667 -14.0773 2-fold cv, 0.666667 -14.0773 (yessenalina 0.666667 -14.0773 26.23 0.666667 -14.0773 krek, 0.666667 -14.0773 0.537 0.666667 -14.0773 clickthrough 0.4 -14.078 tsinghua 0.5 -14.0781 -null 0.5 -14.0781 rome, 0.5 -14.0781 [conceived,entity 0.444444 -14.0784 italics. 0.666667 -14.0785 shatkay, 0.666667 -14.0785 (carpenter 0.666667 -14.0785 precedence relation 0.666667 -14.0785 0.778 0.666667 -14.0785 tac-kbp2010 0.666667 -14.0785 0.892 0.666667 -14.0785 overlappings 0.666667 -14.0785 mel-cepstrum, 0.666667 -14.0785 stories: 0.666667 -14.0785 synonymy: 0.666667 -14.0785 l+ 0.666667 -14.0785 obliged 0.6 -14.0786  0.6 -14.0786 browse, 0.5 -14.0789 venir 0.666667 -14.0796 0.900 0.666667 -14.0796 us-based 0.666667 -14.0796 (rec.) 0.666667 -14.0796 0.609 0.666667 -14.0796 13.2% 0.666667 -14.0796 ante 0.666667 -14.0796 (instances) 0.666667 -14.0796 38.22 0.666667 -14.0796 k landauer 0.666667 -14.0796 60.55 0.666667 -14.0796 tire 0.666667 -14.0796 (functional 0.666667 -14.0796 .\ 0.666667 -14.0796 anr project 0.6 -14.0796 filled [slot1,slot2,slot3,slot4], 0.151899 -14.0796 step, we 0.28 -14.0798 were: 0.2 -14.0803 mention. 0.0447917 -14.0804 compute 0.209302 -14.0805 this hypothesis 0.666667 -14.0807 0.244 0.666667 -14.0807 1θ 0.666667 -14.0807 barzilay, 2008a; 0.666667 -14.0807 bing zhao 0.666667 -14.0807 short-hand 0.666667 -14.0807 clarifications, 0.666667 -14.0807 1493 0.666667 -14.0807 normal distribution-based 0.666667 -14.0807 previously established 0.666667 -14.0807 lp/ilp 0.235294 -14.0807 basis. 0.6 -14.0807 p( 0.444444 -14.0807 (active 0.242424 -14.0813 each language. 0.4 -14.0814 retrieve all 0.5 -14.0815 venkataraman 0.6 -14.0817 1996): 0.666667 -14.0819 reflections 0.666667 -14.0819 favouring 0.666667 -14.0819 alcohol 0.666667 -14.0819 more), 0.666667 -14.0819 [slot2, 0.666667 -14.0819 this subsection 0.666667 -14.0819 distractors selection 0.666667 -14.0819 tra 0.666667 -14.0819 gildea (2001) 0.666667 -14.0819 landmarks, 0.3 -14.0822 two components 0.5 -14.0823 52.12 0.444444 -14.0829 gigabytes 0.666667 -14.083 (b) example 0.666667 -14.083 1343 0.666667 -14.083 warner 0.666667 -14.083 sharedtaskbest 0.666667 -14.083 0.294 0.666667 -14.083 uid 0.666667 -14.083 spanconstrained 0.666667 -14.083 (sarkar, 0.666667 -14.083 mention strings 0.666667 -14.083 43.56 0.666667 -14.083 75.8% 0.666667 -14.083 scheme2 0.333333 -14.0835 overall, our 0.6 -14.0837 jaguar 0.666667 -14.0841 life: 0.666667 -14.0841 2.90 0.666667 -14.0841 personalities, 0.666667 -14.0841 uk: 0.666667 -14.0841 giuglea 0.666667 -14.0841 85.85 0.666667 -14.0841 baa-no 0.666667 -14.0841 fairon 0.666667 -14.0841 0.332 0.666667 -14.0841 baa-t 0.666667 -14.0841 crust 0.666667 -14.0841 syntactic construction, 0.25 -14.0841 first term 0.333333 -14.0846 ali 0.333333 -14.0846 (eisenstein 0.0852713 -14.0847 experiment, 0.6 -14.0847 segmenting named 0.6 -14.0847 junior 0.6 -14.0847 -al 0.4 -14.0848 ivan 0.5 -14.0849 i)) 0.444444 -14.0852 soldiers 0.666667 -14.0853 newman, 0.666667 -14.0853 very comon 0.666667 -14.0853 74.75% 0.666667 -14.0853 significant opportunities 0.666667 -14.0853 specialists, 0.666667 -14.0853 delimiters 0.666667 -14.0853 sales, 0.666667 -14.0853 62.42 0.666667 -14.0853 agreement; 0.666667 -14.0853 moorelewis 0.666667 -14.0853 96.9% 0.666667 -14.0853 letter-based 0.666667 -14.0853 lvg 0.666667 -14.0853 a\b 0.666667 -14.0864 electronic dictionary 0.666667 -14.0864 81.82 0.666667 -14.0864 regulon-operon 0.666667 -14.0864 89.8% 0.666667 -14.0864 trend-contrasting, 0.666667 -14.0864 university’s 0.666667 -14.0864 ao’s 0.666667 -14.0864 (relative) 0.666667 -14.0864 (rte-5) 0.666667 -14.0864 34.37 0.666667 -14.0864 genre) 0.666667 -14.0864 design decisions. 0.666667 -14.0864 (wan, 0.5 -14.0866 ctb7 0.6 -14.0868 kill, 0.666667 -14.0875 subcellular localization 0.666667 -14.0875 processing steps, 0.666667 -14.0875 future works. 0.666667 -14.0875 landmark, 0.666667 -14.0875 b::n 0.666667 -14.0875 (ddtk) 0.666667 -14.0875 test-278 0.5 -14.0875 han, 0.666667 -14.0887 dissertation, 0.666667 -14.0887 (s9), 0.666667 -14.0887 78.9% 0.666667 -14.0887 stk-ct 0.666667 -14.0887 metres 0.666667 -14.0887 absolute scores, 0.666667 -14.0887 ist project 0.666667 -14.0887 δ2 0.666667 -14.0887 (projective) 0.666667 -14.0887 tick 0.666667 -14.0887 barto 0.666667 -14.0887 argentina 0.666667 -14.0887 decoder: 0.666667 -14.0887 39.71 0.666667 -14.0887 resistant 0.285714 -14.089 attaching 0.444444 -14.089 melbourne 0.228571 -14.0891 third, we 0.666667 -14.0898 eo-jeol 0.666667 -14.0898 (content words 0.666667 -14.0898 subclassifier 0.666667 -14.0898 39.24 0.666667 -14.0898 string-valued 0.666667 -14.0898 less prone 0.666667 -14.0898 freezing 0.666667 -14.0898 (avg 0.666667 -14.0898 (nemeskey 0.666667 -14.0898 behaviours 0.666667 -14.0898 (lv 0.666667 -14.0898 0.434 0.6 -14.0899 “it”, 0.6 -14.0899 matches: 0.5 -14.09 mumbai, 0.6 -14.0909 illustration: 0.666667 -14.091 geo:population(y 0.666667 -14.091 tilburg, 0.666667 -14.091 sinha, 0.666667 -14.091 0.279 0.666667 -14.091 96.52 0.666667 -14.091 combiner 0.666667 -14.091 guyon 0.666667 -14.091 +0.51 0.666667 -14.091 golding 0.666667 -14.091 final submission 0.666667 -14.091 keller, 2008; 0.444444 -14.0913 gmbh 0.6 -14.0919 metaphors, 0.666667 -14.0921 (e2), 0.666667 -14.0921 ppts 0.666667 -14.0921 (ogi) 0.666667 -14.0921 0.388 0.666667 -14.0921 ‘standard’ 0.666667 -14.0921 95.5% 0.666667 -14.0921 krijgslaan 0.666667 -14.0921 ganchev, 0.666667 -14.0921 through” 0.666667 -14.0921 domain” 0.666667 -14.0921 dumb 0.666667 -14.0921 φ3 0.384615 -14.093 incorporate prior 0.666667 -14.0933 x(cq1)j 0.666667 -14.0933 32.36 0.666667 -14.0933 0.555 0.666667 -14.0933 3.24 0.666667 -14.0933 cookies 0.666667 -14.0933 have) 0.666667 -14.0933 (sim) 0.666667 -14.0933 4.11 0.666667 -14.0933 unverifiable 0.666667 -14.0933 1.88 0.666667 -14.0933 digitization 0.666667 -14.0933 indefinitely 0.666667 -14.0933 1429 0.666667 -14.0933 abreviations 0.666667 -14.0933 raised, 0.666667 -14.0933 grammar’ 0.5 -14.0935 488 0.444444 -14.0936 (3); 0.25 -14.094 2011 workshop, 0.6 -14.094 unrecognized 0.666667 -14.0944 yd, 0.666667 -14.0944 build distributional 0.666667 -14.0944 d-64289 0.666667 -14.0944 70.29% and 65.23% 0.666667 -14.0944 debate stance 0.666667 -14.0944 problem” 0.666667 -14.0944 bottom, 0.666667 -14.0944 977 0.666667 -14.0944 geometry 0.666667 -14.0944 l2f/inesc-id lisbon 0.666667 -14.0944 (csr) 0.666667 -14.0944 labeling) 0.666667 -14.0944 vinea 0.666667 -14.0944 11in 0.4 -14.0944 fazer 0.122951 -14.0947 side. 0.6 -14.095 e(z) 0.6 -14.095 n] 0.130841 -14.0951 database. 0.444444 -14.0951 jin 0.5 -14.0952 ζ 0.5 -14.0952 examiner 0.666667 -14.0956 rent 0.666667 -14.0956 non-punctuation) 0.666667 -14.0956 messages), 0.666667 -14.0956 x10 0.666667 -14.0956 copulative 0.666667 -14.0956 terminology: 0.666667 -14.0956 (martin, 0.666667 -14.0956 coverbal 0.666667 -14.0956 tsunami 0.666667 -14.0956 students) 0.666667 -14.0956 exhaust 0.666667 -14.0956 msec 0.666667 -14.0956 lifestyle 0.666667 -14.0956 27.49 0.4 -14.0958 ecur 0.285714 -14.0959 individual sentences 0.444444 -14.0959 lapata (2008). 0.6 -14.0961 dirk 0.5 -14.0961 feature-view 0.0590406 -14.0962 pairs of 0.4 -14.0965 0.0010 0.666667 -14.0967 ,w 0.666667 -14.0967 57.45 0.666667 -14.0967 nielsen, 0.666667 -14.0967 waxmonsky, 0.666667 -14.0967 unassigned 0.666667 -14.0967 ‘beautiful’ 0.666667 -14.0967 gentleman 0.666667 -14.0967 (cons) 0.666667 -14.0967 passes, 0.666667 -14.0967 sys+maj 0.666667 -14.0967 2525 0.666667 -14.0967 0.849 0.5 -14.0969 entailed, 0.6 -14.0971 harvard 0.6 -14.0971 creators 0.195652 -14.0976 we exploit 0.5 -14.0978 xiang 0.5 -14.0978 0.8) 0.666667 -14.0979 untuned 0.666667 -14.0979 abend, 0.666667 -14.0979 0.553 0.666667 -14.0979 zero-base 0.666667 -14.0979 39.29 0.666667 -14.0979 refrence 0.666667 -14.0979 93.3% 0.666667 -14.0979 mid-frequency 0.6 -14.0982 possibletreatment() 0.5 -14.0987 “> 0.5 -14.0987 unary, 0.666667 -14.099 semistructured 0.666667 -14.099 5820 0.666667 -14.099 agnew 0.666667 -14.099 night” 0.666667 -14.099 u4: 0.666667 -14.099 comprehending 0.666667 -14.099 human-judged 0.666667 -14.099 (beň 0.666667 -14.099 λ-wasp 0.666667 -14.099 exhaustive, 0.666667 -14.099 (bjorne 0.666667 -14.099 levi’s 0.666667 -14.099 82.8% 0.3 -14.0992 models perform 0.6 -14.0992 (namely 0.5 -14.0995 bilmes 0.5 -14.0995 m10 0.1 -14.0996 p r 0.384615 -14.0997 zhu. 0.384615 -14.0997 lee. 0.3 -14.0998 and dorr, 0.333333 -14.1001 id’s 0.666667 -14.1002 leftor 0.666667 -14.1002 65.87 0.666667 -14.1002 26.64 0.666667 -14.1002 wp$ 0.666667 -14.1002 ⌋ 0.666667 -14.1002 84.1% 0.666667 -14.1002 0.497 0.666667 -14.1002 4.15 0.666667 -14.1002 9.76 0.666667 -14.1002 wiki’s 0.666667 -14.1002 skip-2 0.666667 -14.1002 0.366 0.666667 -14.1002 learninga semanticparser 0.666667 -14.1002 (mccord, 0.6 -14.1002 198, 0.5 -14.1004 poll 0.216216 -14.1006 force research 0.3 -14.1008 subsetsqequal 0.222222 -14.1009 al., 2011), 0.0549273 -14.1009 we do 0.666667 -14.1013 french-speaking 0.666667 -14.1013 0.342 0.666667 -14.1013 1346 0.666667 -14.1013 seasons 0.666667 -14.1013 49.7% 0.666667 -14.1013 3170 0.666667 -14.1013 state-of-the-art methods. 0.666667 -14.1013 allan 0.666667 -14.1013 inland 0.666667 -14.1013 t:±1 0.666667 -14.1013 psss 0.666667 -14.1013 space-separated 0.666667 -14.1013 (1996a) 0.666667 -14.1013 his) 0.666667 -14.1013 (e1), 0.666667 -14.1013 congruence 0.666667 -14.1013 1342 0.444444 -14.1013 argmaxynull 0.5 -14.1013 (tm) 0.25 -14.1017 (zollmann 0.5 -14.1021 inference algorithm. 0.5 -14.1021 girls. 0.666667 -14.1025 1978, 0.666667 -14.1025 yannick 0.666667 -14.1025 sidner’s 0.666667 -14.1025 respected 0.666667 -14.1025 αn 0.666667 -14.1025 caselli 0.666667 -14.1025 imperative, 0.0703125 -14.1025 process. 0.4 -14.1027 no-unary 0.5 -14.103 contacts 0.5 -14.103 [null], 0.666667 -14.1037 gtm, 0.666667 -14.1037 features.2 0.666667 -14.1037 996 0.666667 -14.1037 highdegree 0.666667 -14.1037 salomaa, 0.666667 -14.1037 comprehends 0.666667 -14.1037 i(n) 0.666667 -14.1037 (lprime,mprime) 0.666667 -14.1037 inc, 0.666667 -14.1037 wn4 0.666667 -14.1037 lombardo, 0.666667 -14.1037 40.7 0.666667 -14.1037 wrapped, 0.666667 -14.1037 ⌊maxlength+12 0.666667 -14.1037 37.70 0.666667 -14.1037 ologreg+bowinit 0.666667 -14.1037 (1-b) 0.666667 -14.1048 subject) 0.666667 -14.1048 maccartney, 0.666667 -14.1048 wmt08e 0.666667 -14.1048 pursuit 0.666667 -14.1048 decode, 0.666667 -14.1048 iiit-hyderabad 0.666667 -14.1048 4.59 0.666667 -14.1048 37.94 0.666667 -14.1048 long-forms: 0.666667 -14.1048 gardent, 0.666667 -14.1048 saurí, 0.666667 -14.1048 595 0.666667 -14.1048 coisa 0.666667 -14.1048 suffixes) 0.181818 -14.1054 maturity 0.6 -14.1055 ead documents 0.5 -14.1056 democratic 0.666667 -14.106 representationof 0.666667 -14.106 ‘tonight’ 0.666667 -14.106 derbyshire 0.666667 -14.106 4.75 0.666667 -14.106 apd 0.666667 -14.106 tragedy 0.666667 -14.106 implementation) 0.666667 -14.106 “give me”, 0.666667 -14.106 71.4% 0.666667 -14.106 ascertained 0.666667 -14.106 1756 0.666667 -14.106 slicing 0.666667 -14.106 sea, 0.666667 -14.106 learnt, 0.666667 -14.106 1149 0.666667 -14.106 26.3% 0.666667 -14.106 subconscious 0.666667 -14.106 (devault 0.666667 -14.106 nakov, 0.666667 -14.106 27.58 0.666667 -14.106 0.319 0.3 -14.106 maybe 0.4 -14.1062 (coling 0.5 -14.1065 georgila, 0.444444 -14.1067 state-of-the-art approaches 0.666667 -14.1071 hardly, 0.666667 -14.1071 alen, 0.666667 -14.1071 20.66 0.666667 -14.1071 bjoerne 0.666667 -14.1071 (8, 0.666667 -14.1071 cluster), 0.666667 -14.1071 boom 0.666667 -14.1071 ungrammaticalities 0.666667 -14.1071 er, 0.666667 -14.1071 ita-timebank 0.666667 -14.1071 constitutive 0.666667 -14.1071 a2ev 0.666667 -14.1071 58.39 0.666667 -14.1071 (fei 0.666667 -14.1071 33.01 0.666667 -14.1071 swbd 0.666667 -14.1071 part-factored 0.666667 -14.1071 27.6% 0.666667 -14.1071 winners, 0.5 -14.1073 three experiments. 0.6 -14.1075 (dfg) 0.0640177 -14.1078 observe 0.5 -14.1082 image-centered 0.5 -14.1082 tok 0.5 -14.1082 1790, 0.5 -14.1082 puniyani 0.666667 -14.1083 units) 0.666667 -14.1083 predicate> 0.666667 -14.1083 96.8% 0.666667 -14.1083 22.52 0.666667 -14.1083 seven, 0.666667 -14.1083 algorithm), 0.666667 -14.1083 left-attach 0.666667 -14.1083 eng→eng1 0.666667 -14.1083 (coarse) 0.666667 -14.1083 smoothing) 0.666667 -14.1083 frequencies: 0.666667 -14.1083 activation, 0.666667 -14.1083 dundee, 0.666667 -14.1083 0.988, 0.666667 -14.1083 probing, 0.666667 -14.1083 clause/sentence 0.666667 -14.1083 laplace 0.666667 -14.1083 $1 0.666667 -14.1083 m(n−k 0.3 -14.1086 qun 0.6 -14.1086 gramar 0.5 -14.1091 529 0.666667 -14.1095 0.355 0.666667 -14.1095 θf 0.666667 -14.1095 ‘valid 0.666667 -14.1095 swear 0.666667 -14.1095 ghahramani 0.666667 -14.1095 (sdss) 0.666667 -14.1095 36.46 0.666667 -14.1095 3m 0.666667 -14.1095 h3: 0.666667 -14.1095 harmonised 0.666667 -14.1095 -1.80 0.666667 -14.1095 “best 0.666667 -14.1095 pnorm) 0.6 -14.1096 interest: 0.666667 -14.1106 remain, 0.666667 -14.1106 interpretation), 0.666667 -14.1106 chu-carroll 0.666667 -14.1106 c:±2 0.666667 -14.1106 pix[] 0.666667 -14.1106 490 0.666667 -14.1106 32.32 0.666667 -14.1106 partsof-speech 0.666667 -14.1106 46.34 0.666667 -14.1106 0.169 0.666667 -14.1106 companies; 0.666667 -14.1106 mixtures, 0.666667 -14.1106 (4)–(6) 0.666667 -14.1106 (sample 0.666667 -14.1106 φrev :={}; 0.666667 -14.1106 88.7% 0.666667 -14.1106 micro-macro 0.6 -14.1107 co-occurrence-based 0.5 -14.1108 length; 0.5 -14.1108 million) 0.4 -14.1111 mattress 0.15 -14.1112 translation system. 0.3 -14.1113 only requires 0.444444 -14.1114 final version 0.6 -14.1117 607 0.666667 -14.1118 evanston, 0.666667 -14.1118 f5,f6,f7,f10,f11 0.666667 -14.1118 indictment 0.666667 -14.1118 function” 0.666667 -14.1118 (1)–(3) 0.666667 -14.1118 63.37 0.666667 -14.1118 demands, 0.666667 -14.1118 inertia) 0.666667 -14.1118 knott 0.666667 -14.1118 diagrams, 0.666667 -14.1118 cats/word 0.666667 -14.1118 quantifiers, e.g. 0.666667 -14.1118 dieting 0.666667 -14.1118 performance, including 0.666667 -14.1118 pipeline: 0.666667 -14.1118 αfragment 0.666667 -14.1118 1173 0.666667 -14.1118 hand) 0.666667 -14.1118 w(x) 0.666667 -14.1118 0.688 0.5 -14.1126 investigating how 0.5 -14.1126 large-scale, 0.666667 -14.113 low-tech 0.666667 -14.113 39.15 0.666667 -14.113 784 0.666667 -14.113 l(w)i 0.666667 -14.113 762 0.666667 -14.113 66.31 0.666667 -14.113 56.94 0.666667 -14.113 gold-clean 0.666667 -14.113 a3ev 0.666667 -14.113 todman 0.666667 -14.113 α< 0.666667 -14.113 inputting 0.666667 -14.113 68.40 0.666667 -14.113 otherfeatures 0.666667 -14.113 schütze, 0.666667 -14.113 pi(u2, 0.666667 -14.113 (billion) 0.178571 -14.1137 object. 0.666667 -14.1141 time-series 0.666667 -14.1141 99.06% 0.666667 -14.1141 0.675 0.666667 -14.1141 59.17 0.666667 -14.1141 19.6% 0.666667 -14.1141 news-test2009 0.666667 -14.1141 787 0.666667 -14.1141 training/development 0.666667 -14.1141 disentangle 0.666667 -14.1141 buried 0.666667 -14.1141 2051 0.666667 -14.1141 (yes/no); 0.666667 -14.1141 (1998); 0.666667 -14.1141 navigates 0.666667 -14.1141 thousands, 0.666667 -14.1141 (submitted) 0.666667 -14.1141 inhibition 0.666667 -14.1141 (wierzbicka, 0.666667 -14.1141 22.56 0.222222 -14.1141 carlo 0.357143 -14.1143 < 1 0.5 -14.1144 monsieur 0.357143 -14.1149 (l 0.5 -14.1152 n; 0.666667 -14.1153 argument mapping, 0.666667 -14.1153 bleu/pinc 0.666667 -14.1153 increases: 0.666667 -14.1153 +0.14 0.666667 -14.1153 entitywords 0.666667 -14.1153 mp1234ibm1 0.666667 -14.1153 55,520 0.666667 -14.1153 smola, 0.666667 -14.1153 43.69 0.666667 -14.1153 annotators) 0.666667 -14.1153 0.992 0.666667 -14.1153 sampo pysalo, 0.666667 -14.1153 bagwc_wc 0.666667 -14.1153 9.1% 0.666667 -14.1153 bngs 0.666667 -14.1153 threshold; 0.666667 -14.1153 hyperparameters tuned 0.666667 -14.1153 low-pmi 0.666667 -14.1153 component) 0.666667 -14.1153 web’s 0.666667 -14.1153 diamond 0.357143 -14.1155 ward, 0.0859375 -14.1159 the paper 0.6 -14.116 (structured) 0.5 -14.1161 -e 0.5 -14.1161 plane, 0.666667 -14.1165 0.505 0.666667 -14.1165 weights), 0.666667 -14.1165 false-negatives 0.666667 -14.1165 multi-ft 0.666667 -14.1165 hopelessly 0.666667 -14.1165 1.18 0.666667 -14.1165 0.608 0.666667 -14.1165 tanenhaus 0.666667 -14.1165 4,011 0.666667 -14.1165 restaurants, 0.666667 -14.1165 (syntactic) 0.666667 -14.1165 s7 0.666667 -14.1165 nr2,3↓ 0.666667 -14.1165 751 0.666667 -14.1165 1893, 0.666667 -14.1165 (gold, 0.666667 -14.1165 bunescu, 0.666667 -14.1165 attainment 0.666667 -14.1165 attitude, 0.666667 -14.1165 sdss, 0.666667 -14.1165 head-percolation 0.666667 -14.1165 parameter(s) 0.666667 -14.1165 32.4% 0.666667 -14.1165 expression) 0.666667 -14.1165 0.651 0.666667 -14.1165 ’*’ 0.666667 -14.1165 20.45 0.666667 -14.1165 (object 0.5 -14.117 0.694 0.666667 -14.1177 vandeghinste 0.666667 -14.1177 solution) 0.666667 -14.1177 kitsuregawa 0.666667 -14.1177 (frege, 0.666667 -14.1177 60.34 0.666667 -14.1177 indonesia 0.666667 -14.1177 judicious 0.666667 -14.1177 iteration; 0.666667 -14.1177 emission, 0.666667 -14.1177 (gross, 0.666667 -14.1177 (encouraging 0.666667 -14.1177 +token-seg 0.666667 -14.1177 vbz, 0.666667 -14.1177 retailers 0.666667 -14.1177 thelen 0.666667 -14.1177 p@70 0.666667 -14.1177 hidden-markov 0.666667 -14.1177 framework’s 0.5 -14.1179 (e), 0.5 -14.1179 fbk-irst 0.5 -14.1179 lisboa, 0.210526 -14.1186 (johansson 0.4 -14.1188 mining system 0.5 -14.1188 16.5% 0.666667 -14.1189 m¨uller, 0.666667 -14.1189 λj 0.666667 -14.1189 programmer, 0.666667 -14.1189 bcs 0.666667 -14.1189 (multilingual) 0.666667 -14.1189 langkilde 0.666667 -14.1189 prp) 0.666667 -14.1189 spanish-french 0.666667 -14.1189 non-regular 0.666667 -14.1189 gender), 0.666667 -14.1189 topbs, 0.666667 -14.1189 coefficient: 0.666667 -14.1189 auc, 0.666667 -14.1189 52.52 0.666667 -14.1189 st., 0.666667 -14.1189 636 0.666667 -14.1189 36.28 0.666667 -14.1189 (ldc2003e14) 0.666667 -14.1189 prospect 0.666667 -14.1189 stage; 0.666667 -14.1189 bisani 0.666667 -14.1189 88.46 0.666667 -14.1189 temporal) 0.666667 -14.1189 86.68 0.666667 -14.1189 0.092 0.666667 -14.1189 linearized 0.666667 -14.1189 (15% 0.666667 -14.1189 65.7% 0.166667 -14.1192 by: 0.357143 -14.1193 soroa, 0.4 -14.1195 klaus 0.5 -14.1197 carbonell 0.5 -14.1197 3(a) 0.5 -14.1197 ictclas 0.5 -14.1197 e-to-f 0.169231 -14.1199 (3) where 0.666667 -14.12 because: 0.666667 -14.12 random) 0.666667 -14.12 sb, 0.666667 -14.12 sonderegger 0.666667 -14.12 〈/s〉 0.666667 -14.12 annotators; 0.666667 -14.12 non-sentiment 0.666667 -14.12 k=2 0.666667 -14.12 “part” 0.666667 -14.12 aur´elien 0.666667 -14.12 administration, 0.666667 -14.12 judgment) 0.666667 -14.12 value’ 0.666667 -14.12 0.621 0.666667 -14.12 questionbank 0.666667 -14.12 2.39 0.666667 -14.12 q(rlscriptk) 0.666667 -14.12 jiabao 0.666667 -14.12 well-defined, 0.666667 -14.12 french) 0.666667 -14.12 inter-sentence 0.444444 -14.12 stress information 0.444444 -14.12 0.5). 0.6 -14.1202 (12), 0.6 -14.1202 34.05 0.6 -14.1202 (ed) 0.272727 -14.1204 ponzetto 0.444444 -14.1208 kind. 0.141304 -14.1209 vocabulary. 0.0694087 -14.1209 it. 0.666667 -14.1212 backgrounds, 0.666667 -14.1212 simard 0.666667 -14.1212 (mccoy 0.666667 -14.1212 (reordering) 0.666667 -14.1212 (rca) 0.666667 -14.1212 seats 0.666667 -14.1212 products” 0.666667 -14.1212 rch 0.666667 -14.1212 (turney, 2000; 0.666667 -14.1212 refigure 0.666667 -14.1212 †computer science 0.666667 -14.1212 blandijnberg 0.666667 -14.1212 cfl 0.666667 -14.1212 bda-manhattan 0.666667 -14.1212 complications, 0.666667 -14.1212 27.97 0.6 -14.1213 weaver 0.6 -14.1213 0.666667 -14.1427 out-of 0.666667 -14.1427 (gwb50), 0.666667 -14.1427 renormalizing 0.666667 -14.1427 slowly, 0.666667 -14.1427 8.7m 0.666667 -14.1427 ns, 0.666667 -14.1427 78.7% 0.666667 -14.1427 e.g 0.666667 -14.1427 7.16 0.666667 -14.1427 54.22 0.666667 -14.1427 (ccg) 0.666667 -14.1427 domicile 0.666667 -14.1427 (purandare 0.666667 -14.1427 intervention, 0.666667 -14.1427 whole-sentence 0.666667 -14.1427 constituent-todependency 0.666667 -14.1427 nascent 0.666667 -14.1427 obie 0.6 -14.1427 i} 0.6 -14.1427 “size” 0.141304 -14.1429 implementation of the 0.5 -14.1429 0.535 0.6 -14.1438 yt, 0.5 -14.1438 s2), 0.5 -14.1438 18.4% 0.666667 -14.1439 suppressing 0.666667 -14.1439 n-stem 0.666667 -14.1439 edges), 0.666667 -14.1439 worries 0.666667 -14.1439 flex 0.666667 -14.1439 test” 0.666667 -14.1439 0.750 0.666667 -14.1439 mishne 0.666667 -14.1439 [that 0.666667 -14.1439 speed-ups 0.666667 -14.1439 popel 0.666667 -14.1439 →uttfinal 0.666667 -14.1439 65.28 0.666667 -14.1439 88.10 0.666667 -14.1439 seconds), 0.666667 -14.1439 “soft” 0.666667 -14.1439 atterer 0.666667 -14.1439 19.9 0.666667 -14.1439 mediating 0.666667 -14.1439 weed 0.666667 -14.1439 represented: 0.666667 -14.1439 predicate), 0.666667 -14.1439 adheres 0.666667 -14.1439 multiple-system 0.666667 -14.1439 investors, 0.666667 -14.1439 rater2 0.666667 -14.1439 homage 0.666667 -14.1439 in-line 0.666667 -14.1439 user-created 0.4 -14.1439 verb senses 0.6 -14.1449 ai,j 0.6 -14.1449 sulfur 0.6 -14.1449 link: 0.666667 -14.1451 n2) 0.666667 -14.1451 revertible 0.666667 -14.1451 entailment-related 0.666667 -14.1451 qualifier 0.666667 -14.1451 quantities, 0.666667 -14.1451 7.7% 0.666667 -14.1451 length), 0.666667 -14.1451 42.7 0.666667 -14.1451 {ghannema, 0.666667 -14.1451 0.136 0.666667 -14.1451 physicians 0.666667 -14.1451 dialect, 0.666667 -14.1451 beloved 0.666667 -14.1451 turk: 0.666667 -14.1451 897 0.666667 -14.1451 non-argument 0.666667 -14.1451 5000) 0.666667 -14.1451 23.35 0.666667 -14.1451 url) 0.666667 -14.1451 (hm 0.666667 -14.1451 47.06 0.666667 -14.1451 toolbox3 0.666667 -14.1451 interference, 0.666667 -14.1451 m2maligner 0.666667 -14.1451 parent-sibling 0.666667 -14.1451 prescribing 0.666667 -14.1451 lennar 0.666667 -14.1451 (rc) 0.666667 -14.1451 subjective/objective 0.666667 -14.1451 subject), 0.666667 -14.1451 novelty, 0.666667 -14.1451 tn, 0.5 -14.1456 corporation, 0.5 -14.1456 carvalho, 0.0846774 -14.1457 our experiments, 0.444444 -14.1462 sanity 0.666667 -14.1463 correct” 0.666667 -14.1463 (erg; 0.666667 -14.1463 derrick 0.666667 -14.1463 trainig 0.666667 -14.1463 2.1), 0.666667 -14.1463 symbols) 0.666667 -14.1463 50.85 0.666667 -14.1463 interspersed 0.666667 -14.1463 non-opinion 0.666667 -14.1463 minds 0.666667 -14.1463 ‘q’, 0.666667 -14.1463 regulondependence 0.666667 -14.1463 0.405 0.666667 -14.1463 0.072 0.666667 -14.1463 l(yprime,y) 0.666667 -14.1463 interrelationships 0.666667 -14.1463 wordtopic 0.5 -14.1474 christine 0.5 -14.1474 93.55 0.5 -14.1474 (ahn 0.5 -14.1474 0.403 0.5 -14.1474 banea, 0.5 -14.1474 ama 0.666667 -14.1475 andv 0.666667 -14.1475 ϕi 0.666667 -14.1475 believe) 0.666667 -14.1475 (lattice 0.666667 -14.1475 source-syntax-augmented 0.666667 -14.1475 package5 0.666667 -14.1475 cheung 0.666667 -14.1475 (wolf 0.666667 -14.1475 recombined 0.666667 -14.1475 51.04 0.666667 -14.1475 4grams 0.666667 -14.1475 (lch) 0.666667 -14.1475 takahashi 0.666667 -14.1475 (td) 0.666667 -14.1475 (cooper, 0.666667 -14.1475 fatty 0.666667 -14.1475 usingthe 0.666667 -14.1475 0.933 0.666667 -14.1475 deficits 0.666667 -14.1475 word-form, 0.666667 -14.1475 posf 0.666667 -14.1475 0.171 0.666667 -14.1475 2.41 0.666667 -14.1475 minimally, 0.233333 -14.1476 can you 0.666667 -14.1487 pre-stored 0.666667 -14.1487 later) 0.666667 -14.1487 incorrect; 0.666667 -14.1487 46.26 0.666667 -14.1487 sub-gradient 0.666667 -14.1487 991 0.666667 -14.1487 wcbr 0.666667 -14.1487 correlative 0.666667 -14.1487 weakened 0.666667 -14.1487 0.612 0.666667 -14.1487 synthdriver 0.666667 -14.1487 part” 0.666667 -14.1487 52.63 0.666667 -14.1487 yoshida 0.666667 -14.1487 mcenery 0.666667 -14.1487 mappable 0.666667 -14.1487 contentwindow20 0.666667 -14.1487 inferable 0.666667 -14.1487 veterans 0.666667 -14.1487 specialist, 0.666667 -14.1487 51.95% 0.666667 -14.1487 1hmm-lm 0.666667 -14.1487 14,856 0.357143 -14.1494 xa 0.4 -14.1497 intraand 0.139785 -14.15 is determined 0.666667 -14.15 1512 0.666667 -14.15 understanding) 0.666667 -14.15 lsi 0.666667 -14.15 verbs); 0.666667 -14.15 i;i;ic1 0.666667 -14.15 60.87 0.666667 -14.15 section; 0.666667 -14.15 deviating 0.666667 -14.15 clues; 0.666667 -14.15 60736014 0.666667 -14.15 (application 0.666667 -14.15 “hidden” 0.666667 -14.15 æ« 0.666667 -14.15 browse) 0.666667 -14.15 partnered 0.666667 -14.15 careful, 0.666667 -14.15 apostrophes 0.666667 -14.15 43.4% 0.666667 -14.15 81.4% 0.666667 -14.15 wsk 0.666667 -14.15 vertically, 0.666667 -14.15 ocupations 0.666667 -14.15 i←0 0.666667 -14.15 pharaoh 0.666667 -14.15 ginzburg 0.666667 -14.15 yngve 0.666667 -14.15 watch, 0.666667 -14.15 qqueries 0.5 -14.1502 ‘i 0.6 -14.1504 ˆyi 0.25 -14.1507 roark et 0.666667 -14.1512 rebuttals, 0.666667 -14.1512 innovate 0.666667 -14.1512 levels) 0.666667 -14.1512 mloc, 0.666667 -14.1512 27.25 0.666667 -14.1512 recal) 0.666667 -14.1512 (patchan 0.666667 -14.1512 60.2% 0.666667 -14.1512 if-then-else 0.666667 -14.1512 78.3% 0.666667 -14.1512 dsa 0.666667 -14.1512 (oracle) 0.666667 -14.1512 584 0.666667 -14.1512 krenn, 0.666667 -14.1512 774 0.666667 -14.1512 checkers, 0.666667 -14.1512 (alshawi, 0.666667 -14.1512 smuggling 0.666667 -14.1512 anyhow 0.666667 -14.1512 cess 0.6 -14.1515 reporting, 0.148148 -14.1519 (blei et 0.357143 -14.152 identi 0.5 -14.152 (ptk) 0.5 -14.152 0.246 0.666667 -14.1524 noun1-ed 0.666667 -14.1524 instance), 0.666667 -14.1524 47.33 0.666667 -14.1524 ings 0.666667 -14.1524 issuing 0.666667 -14.1524 (venkatapathy 0.666667 -14.1524 upright 0.666667 -14.1524 applies, 0.666667 -14.1524 (fe) 0.666667 -14.1524 revoke 0.666667 -14.1524 notwithstanding, 0.666667 -14.1524 #173 0.666667 -14.1524 null u25(x,y,di,α) 0.666667 -14.1524 monotonous, 0.666667 -14.1524 counterintuitive 0.666667 -14.1524 2300 0.666667 -14.1524 30.22 0.666667 -14.1524 penalty” 0.666667 -14.1524 transcriptionevent 0.666667 -14.1524 inflection’ 0.666667 -14.1524 um), 0.666667 -14.1524 0.348 0.666667 -14.1524 θ1, 0.666667 -14.1524 english-russian, 0.666667 -14.1524 gaithersburg, 0.666667 -14.1524 56.07 0.666667 -14.1524 phrasetable 0.666667 -14.1524 contains: 0.666667 -14.1524 etree 0.666667 -14.1524 (42 0.666667 -14.1524 p(f 0.6 -14.1526 holy 0.6 -14.1526 f(u,v) 0.6 -14.1526 addressee’s 0.333333 -14.1529 indices. 0.5 -14.1529 itg-based 0.126126 -14.153 it seems 0.666667 -14.1536 73.8% 0.666667 -14.1536 generating, 0.666667 -14.1536 spherical 0.666667 -14.1536 lowe, 0.666667 -14.1536 lish 0.666667 -14.1536 46.60 0.666667 -14.1536 minor, 0.666667 -14.1536 (22 0.666667 -14.1536 sub-entities, 0.666667 -14.1536 demonstractive 0.666667 -14.1536 vigor 0.666667 -14.1536 disproportionately 0.666667 -14.1536 971 0.666667 -14.1536 tags.1 0.666667 -14.1536 24.30 0.666667 -14.1536 teresniak 0.666667 -14.1536 (only) 0.666667 -14.1536 rprimed, 0.666667 -14.1536 fraudulent 0.666667 -14.1536 versions: 0.666667 -14.1536 ‘change 0.666667 -14.1536 ts) 0.666667 -14.1536 “artifact” 0.666667 -14.1536 wuebker 0.666667 -14.1536 pairwise, 0.666667 -14.1536 3-tuple 0.666667 -14.1536 omd, 0.666667 -14.1536 invocations 0.666667 -14.1536 region) 0.666667 -14.1536 (applying 0.666667 -14.1536 searle 0.5 -14.1538 vous 0.0665025 -14.1542 2005). 0.5 -14.1547 78712 0.5 -14.1547 #1: 0.666667 -14.1548 history-length/class 0.666667 -14.1548 draft, 0.666667 -14.1548 ta, 0.666667 -14.1548 sqlite 0.666667 -14.1548 rcv1, 0.666667 -14.1548 reviewed, 0.666667 -14.1548 active/passive 0.666667 -14.1548 springhouse 0.666667 -14.1548 6,821 0.666667 -14.1548 (treating 0.666667 -14.1548 seniority 0.666667 -14.1548 incompleteness 0.666667 -14.1548 dead, 0.666667 -14.1548 said”, 0.666667 -14.1548 kep 0.666667 -14.1548 (32%) 0.666667 -14.1548 clustering), 0.666667 -14.1548 (svms), 0.666667 -14.1548 “real 0.666667 -14.1548 (manual) 0.666667 -14.1548 ide, 0.666667 -14.1548 with), 0.666667 -14.1548 89.70 0.666667 -14.1548 list2 0.666667 -14.1548 (yangarber 0.666667 -14.1548 non-projective, 0.666667 -14.1548 over-fitting, 0.666667 -14.1548 topic-sensitive 0.666667 -14.1548 utiyama 0.666667 -14.1548 x(x 0.6 -14.1548 doug 0.216216 -14.1549 our experiment 0.3 -14.1559 long-forms 0.6 -14.1559 medt-3 0.666667 -14.156 tuned, 0.666667 -14.156 gtran 0.666667 -14.156 participants) 0.666667 -14.156 0.964 0.666667 -14.156 standby 0.666667 -14.156 40.39 0.666667 -14.156 64.39 0.666667 -14.156 landscape) 0.666667 -14.156 mayfield 0.666667 -14.156 she/he 0.666667 -14.156 mullen, 0.666667 -14.156 c:±3 0.666667 -14.156 interconnected, 0.666667 -14.156 nikolay 0.666667 -14.156 bcub, 0.666667 -14.156 53.54 0.666667 -14.156 edited” 0.666667 -14.156 perkins, 0.666667 -14.156 dell’orletta 0.666667 -14.156 cumulatively 0.666667 -14.156 “go” 0.666667 -14.156 plsr 0.666667 -14.156 18.3% 0.666667 -14.156 offline, 0.666667 -14.156 safer 0.666667 -14.156 57.55 0.666667 -14.156 (blatz 0.666667 -14.156 95.1% 0.666667 -14.156 identifying, 0.666667 -14.156 english-hindi, 0.666667 -14.156 8,008 0.666667 -14.156 usualy 0.666667 -14.156 1323 0.666667 -14.156 neubau 0.444444 -14.1568 responders 0.2 -14.157 however, their 0.363636 -14.1571 fingerprints 0.666667 -14.1573 gillette 0.666667 -14.1573 (hoi 0.666667 -14.1573 0.805 0.666667 -14.1573 scaling, 0.666667 -14.1573 linnea 0.666667 -14.1573 24.79 0.666667 -14.1573 antonymous, 0.666667 -14.1573 10 0.666667 -14.1783 (overall) 0.666667 -14.1783 markable> 0.666667 -14.1783 generalpurpose 0.666667 -14.1783 meat, 0.666667 -14.1783 szpektor, 0.666667 -14.1783 wizard, 0.363636 -14.1787 factorized 0.5 -14.1788 t2: 0.666667 -14.1795 reasonably, 0.666667 -14.1795 without, 0.666667 -14.1795 rnn 0.666667 -14.1795 test-train 0.666667 -14.1795 75.41 0.666667 -14.1795 village 0.666667 -14.1795 wolfe, 0.666667 -14.1795 forest; 0.666667 -14.1795 32.35 0.666667 -14.1795 (svm 0.666667 -14.1795 inasmuch 0.666667 -14.1795 710 0.666667 -14.1795 0.818 0.666667 -14.1795 max-derivation 0.666667 -14.1795 +0.24 0.666667 -14.1795 infinity 0.666667 -14.1795 “parts” 0.666667 -14.1795 journey 0.666667 -14.1795 naaman 0.666667 -14.1795 wasted 0.666667 -14.1795 1,2,...,k 0.666667 -14.1795 deadjectival 0.666667 -14.1795 w3’s 0.666667 -14.1795 geting 0.666667 -14.1795 revised, 0.666667 -14.1795 ascribe 0.666667 -14.1795 same-as 0.666667 -14.1795 officers 0.666667 -14.1795 “stop” 0.666667 -14.1795 (cdc) 0.5 -14.1798 0.131 0.5 -14.1807 pennsylvania, 0.5 -14.1807 61.6% 0.5 -14.1807 usprimev 0.666667 -14.1808 0.767 0.666667 -14.1808 o’neill 0.666667 -14.1808 wsj-test 0.666667 -14.1808 chiticariu 0.666667 -14.1808 clasifiers 0.666667 -14.1808 reinterpret 0.666667 -14.1808 ratified 0.666667 -14.1808 (sequences 0.666667 -14.1808 (minimum 0.666667 -14.1808 grice 0.666667 -14.1808 late, 0.666667 -14.1808 bayesum 0.666667 -14.1808 crf, crf-d, 0.666667 -14.1808 (essentially, 0.666667 -14.1808 properties; 0.666667 -14.1808 tsur, 0.666667 -14.1808 inclusion, 0.666667 -14.1808 (68%) 0.666667 -14.1808 walks, 0.666667 -14.1808 predj 0.666667 -14.1808 deutsch 0.666667 -14.1808 637 0.666667 -14.1808 0.163 0.666667 -14.1808 (input) 0.666667 -14.1808 factor), 0.666667 -14.1808 ‘_n_’ 0.666667 -14.1808 right-headed 0.666667 -14.1808 4.77 0.333333 -14.1814 promise for 0.666667 -14.1821 ctb7. 0.666667 -14.1821 win/loss 0.666667 -14.1821 498 0.666667 -14.1821 aachen, 0.666667 -14.1821 10.92 0.666667 -14.1821 25.07 0.666667 -14.1821 24617-2 0.666667 -14.1821 (fox, 0.666667 -14.1821 79.4% 0.666667 -14.1821 its peak 0.666667 -14.1821 autonomic nerve 0.666667 -14.1821 seretan 0.666667 -14.1821 however,it 0.666667 -14.1833 certified 0.666667 -14.1833 here; 0.666667 -14.1833 3to 0.666667 -14.1833 2.51 0.666667 -14.1833 s3, 0.666667 -14.1833 tao 0.666667 -14.1833 32.80 0.666667 -14.1833 annotator) 0.666667 -14.1833 largest, 0.666667 -14.1833 93.33 0.666667 -14.1833 triggers) 0.666667 -14.1833 0.498 0.666667 -14.1833 b-bit 0.666667 -14.1833 japanese) 0.666667 -14.1833 non-partners 0.666667 -14.1833 (ui) 0.666667 -14.1833 (stsg) 0.666667 -14.1833 i,0 0.666667 -14.1833 reverts 0.666667 -14.1833 91.04% 0.666667 -14.1833 τv 0.666667 -14.1833 somers 0.666667 -14.1833 0.751 0.666667 -14.1833 row); 0.666667 -14.1833 2.83 0.666667 -14.1833 minnesota, 0.5 -14.1835 (lang, 0.5 -14.1835 english/arabic, 0.169492 -14.1836 is treated 0.121739 -14.184 what we 0.5 -14.1845 whittaker 0.5 -14.1845 cake 0.5 -14.1845 kill 0.666667 -14.1846 khalil 0.666667 -14.1846 cybulska, 0.666667 -14.1846 surf, 0.666667 -14.1846 cotraining, 0.666667 -14.1846 andr 0.666667 -14.1846 ’05 0.666667 -14.1846 finishes, 0.666667 -14.1846 commodity 0.666667 -14.1846 2-21, 0.666667 -14.1846 hori 0.666667 -14.1846 603 0.666667 -14.1846 335, 0.666667 -14.1846 english-specific 0.666667 -14.1846 english-hindi 0.666667 -14.1846 acts) 0.666667 -14.1846 (typically, 0.666667 -14.1846 envdb 0.666667 -14.1846 nonisomorphism 0.666667 -14.1846 (dr) 0.666667 -14.1846 petrakis 0.666667 -14.1846 floorleftsfloorright 0.666667 -14.1846 1028 0.666667 -14.1858 (§4.1), 0.666667 -14.1858 408 0.666667 -14.1858 structure), 0.666667 -14.1858 kite 0.666667 -14.1858 bartender 0.666667 -14.1858 (ortiz-mart´ınez 0.666667 -14.1858 threatening 0.666667 -14.1858 log) 0.666667 -14.1858 specifier 0.666667 -14.1858 vectorx 0.666667 -14.1858 (introduced 0.666667 -14.1858 cpu, 0.666667 -14.1858 a11 0.666667 -14.1858 0.959 0.666667 -14.1858 42.50 0.666667 -14.1858 is-a) 0.666667 -14.1858 hovy’s 0.666667 -14.1858 (pscfg) 0.666667 -14.1858 krippendorff, 0.666667 -14.1858 gt(lscript) 0.666667 -14.1858 entity-role 0.666667 -14.1858 kushmerick, 0.666667 -14.1858 rom 0.666667 -14.1858 58.43 0.666667 -14.1858 methods’ 0.666667 -14.1858 concatenations, 0.666667 -14.1858 technique: 0.13 -14.186 this study, 0.5 -14.1864 ¯a 0.294118 -14.1868 400 600 0.4 -14.1868 lscript2 0.3125 -14.1869 424 0.666667 -14.1871 diwa 0.666667 -14.1871 0.037 0.666667 -14.1871 87.63 0.666667 -14.1871 ‘p’ 0.666667 -14.1871 max-ent 0.666667 -14.1871 0.515 0.666667 -14.1871 “before” 0.666667 -14.1871 8see 0.666667 -14.1871 250,000 0.666667 -14.1871 verses, 0.666667 -14.1871 nullt, 0.666667 -14.1871 factorings 0.666667 -14.1871 2008)’s 0.666667 -14.1871 non-sparse 0.666667 -14.1871 np al$yx 0.666667 -14.1871 operationally 0.666667 -14.1871 macro-averages 0.666667 -14.1871 55.6% 0.666667 -14.1871 non-annotated 0.666667 -14.1871 proficiency, 0.666667 -14.1871 buyer 0.666667 -14.1871 football, 0.666667 -14.1871 bristol 0.666667 -14.1871 counteract 0.666667 -14.1871 sellers 0.666667 -14.1871 str: 0.666667 -14.1871 1-million-sentence 0.103896 -14.1872 presents a 0.0410959 -14.1872 for computational linguistics 0.5 -14.1873 τ(σ) 0.5 -14.1873 hcr 0.5 -14.1873 everybody 0.205128 -14.1875 earlier. 0.4 -14.1876 (false 0.666667 -14.1883 nonbinary 0.666667 -14.1883 biemiler 0.666667 -14.1883 *”, 0.666667 -14.1883 khan 0.666667 -14.1883 pattern” 0.666667 -14.1883 readable, 0.666667 -14.1883 lehnert 0.666667 -14.1883 (soderland 0.666667 -14.1883 23.86 0.666667 -14.1883 speakers; 0.666667 -14.1883 files), 0.666667 -14.1883 phrasess 0.666667 -14.1883 196), 0.666667 -14.1883 unnecessarily, 0.666667 -14.1883 else0)·c1(p) 0.666667 -14.1883 comparative, 0.666667 -14.1883 (koch, 0.666667 -14.1883 85k 0.666667 -14.1883 911 0.666667 -14.1883 assertions, 0.666667 -14.1883 62.3% 0.666667 -14.1883 o(nst) 0.666667 -14.1883 random-walk 0.666667 -14.1883 nao 0.666667 -14.1883 setto 0.666667 -14.1883 (3)⇐⇒ 0.666667 -14.1883 95.49 0.666667 -14.1883 curran’s 0.5 -14.1892 correctly tokenized 0.5 -14.1892 30.52 0.666667 -14.1896 “concept” 0.666667 -14.1896 lists), 0.666667 -14.1896 post-annotation 0.666667 -14.1896 fukumoto 0.666667 -14.1896 θlq 0.666667 -14.1896 cregular, 0.666667 -14.1896 “skills” 0.666667 -14.1896 n), 0.666667 -14.1896 (montague, 0.666667 -14.1896 a*, 0.666667 -14.1896 autoemo 0.666667 -14.1896 downwards 0.666667 -14.1896 nonnegligible 0.666667 -14.1896 52.42 0.666667 -14.1896 submodel 0.666667 -14.1896 =1), 0.666667 -14.1896 2.03 0.666667 -14.1896 dirichlet(α) 0.666667 -14.1896 56%, 0.666667 -14.1896 51.3 0.666667 -14.1896 “president” 0.666667 -14.1896 51.2% 0.666667 -14.1896 lambek’s 0.666667 -14.1896 web-train 0.666667 -14.1896 (giles 0.666667 -14.1896 comparable: 0.666667 -14.1896 y〈ep,er〉 0.666667 -14.1896 outcome: 0.666667 -14.1896 probabilistically, 0.666667 -14.1896 ilk 0.5 -14.1902 durango 0.5 -14.1902 “turn 0.4 -14.1906 query cluster 0.666667 -14.1909 75.4% 0.666667 -14.1909 srcc: 0.666667 -14.1909 17.73 0.666667 -14.1909 part-of, 0.666667 -14.1909 processor, 0.666667 -14.1909 87.74 0.666667 -14.1909 (r(5393) 0.666667 -14.1909 ρ1 0.666667 -14.1909 determine, 0.666667 -14.1909 certainly, 0.666667 -14.1909 ℓj 0.666667 -14.1909 unreasonably 0.666667 -14.1909 gap: 0.666667 -14.1909 stevens 0.666667 -14.1909 unavoidable, 0.666667 -14.1909 1547 0.666667 -14.1909 53.11 0.666667 -14.1909 date), 0.666667 -14.1909 wasp−1++ 0.666667 -14.1909 southeast 0.666667 -14.1909 gilbert 0.666667 -14.1909 2.82 0.666667 -14.1909 1202 0.666667 -14.1909 member, 0.666667 -14.1909 (2004)), 0.5 -14.1911 journalists 0.4 -14.1914 cells. 0.225806 -14.1915 chunks. 0.666667 -14.1921 (iterative 0.666667 -14.1921 (godbole 0.666667 -14.1921 pseudodocument 0.666667 -14.1921 two-day 0.666667 -14.1921 lwdp 0.666667 -14.1921 fk(nullt∗null) 0.666667 -14.1921 (answer 0.666667 -14.1921 (tsur 0.666667 -14.1921 ja, 0.666667 -14.1921 parsing,which 0.666667 -14.1921 93.31 0.666667 -14.1921 1078 0.666667 -14.1921 acnegd 0.666667 -14.1921 xiaowen 0.666667 -14.1921 lying, 0.666667 -14.1921 ddi, 0.666667 -14.1921 siel 0.666667 -14.1921 56.5% 0.666667 -14.1921 53.0% 0.666667 -14.1921 muhammad 0.666667 -14.1921 3.94 0.666667 -14.1921 βν 0.666667 -14.1921 autism, 0.666667 -14.1921 attestations 0.666667 -14.1921 rear 0.0456182 -14.1924 amount 0.110294 -14.1927 this suggests 0.5 -14.193 but fail 0.666667 -14.1934 post-processed 0.666667 -14.1934 64.15 0.666667 -14.1934 downward-branching 0.666667 -14.1934 code; 0.666667 -14.1934 method1 0.666667 -14.1934 united, 0.666667 -14.1934 workforce, 0.666667 -14.1934 alec 0.666667 -14.1934 (underlined 0.666667 -14.1934 vagueness 0.666667 -14.1934 (real 0.666667 -14.1934 {initialize 0.666667 -14.1934 sidner 0.666667 -14.1934 (picard, 0.666667 -14.1934 1290 0.666667 -14.1934 θmatch 0.666667 -14.1934 quantifiers: 0.666667 -14.1934 ψ(z)i 0.666667 -14.1934 (sagi 0.666667 -14.1934 0.22, 0.666667 -14.1934 cross-slot 0.666667 -14.1934 chaudhuri 0.666667 -14.1934 satanjeev 0.666667 -14.1934 holmqvist, 0.666667 -14.1934 63.90 0.666667 -14.1934 hauptmann, 0.666667 -14.1934 (kuhn 0.26087 -14.1937 tx 0.0519685 -14.1942 the resulting 0.666667 -14.1947 compose-decompose 0.666667 -14.1947 tfs 0.666667 -14.1947 negative; 0.666667 -14.1947 22.16 0.666667 -14.1947 learnedby 0.666667 -14.1947 unlocking 0.666667 -14.1947 attends 0.666667 -14.1947 boch) 0.666667 -14.1947 up) 0.666667 -14.1947 paint 0.666667 -14.1947 lexico-grammatical 0.666667 -14.1947 learningfrom 0.666667 -14.1947 terrain 0.666667 -14.1947 (kikui 0.666667 -14.1947 consolidating 0.666667 -14.1947 28.27 0.666667 -14.1947 -fb} 0.666667 -14.1947 40.82 0.666667 -14.1947 1,...,k 0.666667 -14.1947 ilist 0.666667 -14.1947 2.53 0.666667 -14.1947 3.40 0.666667 -14.1947 thousands) 0.666667 -14.1947 (pipeline) 0.666667 -14.1947 causative/incoative 0.666667 -14.1947 (reported 0.666667 -14.1947 2nlp 0.666667 -14.1947 grec, 0.666667 -14.1947 drinks 0.666667 -14.1947 singapore, 0.666667 -14.1947 euclidean, 0.666667 -14.1947 (bahl 0.5 -14.1949 0.1982 0.5 -14.1959 beihan 0.666667 -14.196 wfsts 0.666667 -14.196 equaly 0.666667 -14.196 wconf 0.666667 -14.196 samar 0.666667 -14.196 0.419 0.666667 -14.196 setups: 0.666667 -14.196 stakes 0.666667 -14.196 'the 0.666667 -14.196 problems) 0.666667 -14.196 components), 0.666667 -14.196 defend 0.666667 -14.196 optimistic, 0.666667 -14.196 slm, 0.666667 -14.196 43.94 0.666667 -14.196 dec-1st 0.666667 -14.196 let‟s 0.666667 -14.196 α’s 0.666667 -14.196 yvon, 0.666667 -14.196 +0.70 0.666667 -14.196 (80%), 0.666667 -14.196 substantialy 0.666667 -14.196 bottom-most 0.666667 -14.196 semi-ring 0.666667 -14.196 2.28 0.294118 -14.1968 emacc 0.666667 -14.1972 optimises 0.666667 -14.1972 0.443 0.666667 -14.1972 moses 0.5987 0.666667 -14.1972 artificiality 0.666667 -14.1972 (neal 0.666667 -14.1972 s∗2 0.666667 -14.1972 -0.81 0.666667 -14.1972 (hlt-naacl 0.666667 -14.1972 fade 0.666667 -14.1972 skip-4 0.666667 -14.1972 punctuation), 0.666667 -14.1972 shieber’s 0.666667 -14.1972 scope-bearing 0.666667 -14.1972 hana 0.666667 -14.1972 a-patterns 0.666667 -14.1972 13,000 0.666667 -14.1972 (result) 0.666667 -14.1972 g(µ) 0.666667 -14.1972 55.4% 0.666667 -14.1972 unassignment 0.666667 -14.1972 ch(t) 0.666667 -14.1972 conservatively 0.666667 -14.1972 call duration 0.666667 -14.1972 model1, 0.666667 -14.1972 f12 0.3125 -14.1976 ku-ka 0.132653 -14.1977 al., 2011) 0.5 -14.1978 46.63 0.363636 -14.198 51.8 0.4 -14.1982 avenue, 0.4 -14.1982 sweet 0.666667 -14.1985 (sus) 0.666667 -14.1985 271-300 0.666667 -14.1985 (-ts) 0.666667 -14.1985 users; 0.666667 -14.1985 md-ii 0.666667 -14.1985 discharged 0.666667 -14.1985 krovetz 0.666667 -14.1985 interruption, 0.666667 -14.1985 28.38 0.666667 -14.1985 0%, 0.666667 -14.1985 stress, 0.666667 -14.1985 39.52 0.666667 -14.1985 seven-point 0.666667 -14.1985 arl, 0.666667 -14.1985 exhausted, 0.666667 -14.1985 waikato 0.666667 -14.1985 iic 0.666667 -14.1985 25.45 0.666667 -14.1985 refitting 0.666667 -14.1985 conveniently, 0.666667 -14.1985 788 0.666667 -14.1985 antal 0.666667 -14.1985 4,295 0.666667 -14.1985 (one-tailed 0.666667 -14.1985 else) 0.0425532 -14.1986 2009). 0.666667 -14.1998 =∇k+1 0.666667 -14.1998 nonspecific 0.666667 -14.1998 demanding, 0.666667 -14.1998 atlas.ti 0.666667 -14.1998 (simplified) 0.666667 -14.1998 (norris 0.666667 -14.1998 “y” 0.666667 -14.1998 pb(f) 0.666667 -14.1998 non-entity 0.666667 -14.1998 (udupa 0.666667 -14.1998 karin, 0.666667 -14.1998 19–27, 0.666667 -14.1998 (below 0.666667 -14.1998 inexpensively 0.666667 -14.1998 86.21 0.666667 -14.1998 815 0.666667 -14.1998 right-context 0.666667 -14.1998 æ» 0.666667 -14.1998 epstein, 0.666667 -14.1998 chunliang 0.666667 -14.1998 0.205 0.666667 -14.1998 acposd 0.666667 -14.1998 (beesley 0.666667 -14.1998 (schank 0.666667 -14.1998 21) 0.666667 -14.1998 speculatively 0.666667 -14.1998 2005b), 0.666667 -14.1998 specialized, 0.666667 -14.1998 (null)) 0.666667 -14.1998 takers, 0.666667 -14.1998 quasi 0.666667 -14.1998 36.49 0.666667 -14.1998 builtin 0.666667 -14.1998 (+ts) 0.666667 -14.1998 interrupts 0.225806 -14.2005 19-24 june 0.5 -14.2007 (m¨uller 0.5 -14.2007 92.64 0.363636 -14.2008 station 0.666667 -14.2011 (tap 0.666667 -14.2011 filtered 0.666667 -14.2011 vps, 0.666667 -14.2011 except, 0.666667 -14.2011  0.666667 -14.2011 freedman 0.666667 -14.2011 own) 0.666667 -14.2011 porter’s 0.666667 -14.2011 ieee/wic/acm 0.666667 -14.2011 (3.9) 0.666667 -14.2011 situation: 0.666667 -14.2011 blanco 0.666667 -14.2011 chang. 0.666667 -14.2011 treebanks: 0.666667 -14.2011 blum, 0.666667 -14.2011 85.31 0.666667 -14.2011 90.13 0.666667 -14.2011 finis 0.666667 -14.2011 message: 0.666667 -14.2011 unigram-based 0.666667 -14.2011 0.647 0.666667 -14.2011 takashi 0.666667 -14.2011 [2010] 0.666667 -14.2011 945 0.666667 -14.2011 228k 0.666667 -14.2011 monotonic, 0.666667 -14.2011 l(y,yprime) 0.666667 -14.2011 (ko) 0.666667 -14.2011 “$” 0.666667 -14.2011 (potentially) 0.666667 -14.2011 biparsing 0.666667 -14.2011 ecologically 0.666667 -14.2011 cheese 0.5 -14.2016 βα 0.5 -14.2016 qu 0.5 -14.2016 91.01 0.233333 -14.2021 two simple 0.666667 -14.2023 2006/7 0.666667 -14.2023 (wightman 0.666667 -14.2023 initiatives 0.666667 -14.2023 tol 0.666667 -14.2023 markup-induced 0.666667 -14.2023 ding, 0.666667 -14.2023 45.06 0.666667 -14.2023 worker, 0.666667 -14.2023 resolutions 0.666667 -14.2023 footnotes, 0.666667 -14.2023 dismal 0.666667 -14.2023 ratios: 0.666667 -14.2023 hadoop-based 0.666667 -14.2023 (medt) 0.666667 -14.2023 2.2m 0.666667 -14.2023 11.1% 0.666667 -14.2023 penalties: 0.666667 -14.2023 baseline” 0.666667 -14.2023 morphosyntacticcategory 0.666667 -14.2023 [al]: 0.666667 -14.2023 implementationdetails 0.666667 -14.2023 nieuws 0.666667 -14.2023 kouylekov 0.666667 -14.2023 2008] 0.666667 -14.2023 (burnard, 0.666667 -14.2023 gpe, 0.666667 -14.2023 woeparse’s 0.666667 -14.2023 2.3), 0.666667 -14.2023 bob’s 0.666667 -14.2023 discounts, 0.666667 -14.2023 wrong: 0.666667 -14.2023 slda, 0.666667 -14.2023 hazard, 0.666667 -14.2023 development; 0.666667 -14.2023 priming, 0.5 -14.2026 nina 0.5 -14.2026 birmingham, 0.5 -14.2026 babies 0.5 -14.2026 (nigam 0.216216 -14.2028 chain. 0.4 -14.2028 nato 0.216216 -14.2032 experiments we used 0.5 -14.2035 context dependent 0.666667 -14.2036 notationally 0.666667 -14.2036 cora 0.666667 -14.2036 splicing 0.666667 -14.2036 38.21 0.666667 -14.2036 hasan 0.666667 -14.2036 reassignment, 0.666667 -14.2036 0.420 0.666667 -14.2036 satisfaction” 0.666667 -14.2036 139, 0.666667 -14.2036 weighed 0.666667 -14.2036 87.99 0.666667 -14.2036 6+ 0.666667 -14.2036 shortcuts 0.666667 -14.2036 related: 0.666667 -14.2036 (dw) 0.666667 -14.2036 {0,1}, 0.666667 -14.2036 sustainable 0.666667 -14.2036 [see 0.666667 -14.2036 shift: 0.666667 -14.2036 ancestor, 0.666667 -14.2036 1.34 0.666667 -14.2036 openness 0.0285987 -14.2042 which is 0.5 -14.2045 (buyko 0.5 -14.2045 3-fold 0.666667 -14.2049 response-generation 0.666667 -14.2049 (nlu), 0.666667 -14.2049 ova 0.666667 -14.2049 fischer 0.666667 -14.2049 26.87 0.666667 -14.2049 cost(˚ι) 0.666667 -14.2049 ndcg@k 0.666667 -14.2049 iobj 0.666667 -14.2049 state’s 0.666667 -14.2049 cross-chain 0.666667 -14.2049 outright 0.666667 -14.2049 log-odds 0.666667 -14.2049 melamed, 0.666667 -14.2049 1.93 0.666667 -14.2049 6.1% 0.666667 -14.2049 fourier 0.666667 -14.2049 blown 0.666667 -14.2049 garnered 0.666667 -14.2049 (therefore 0.666667 -14.2049 dirt, 0.666667 -14.2049 amber: 0.666667 -14.2049 "i 0.666667 -14.2049 place-holder 0.666667 -14.2049 31.3% 0.666667 -14.2049 filters: 0.666667 -14.2049 fell, 0.666667 -14.2049 smr 0.666667 -14.2049 (events 0.666667 -14.2049 extracted; 0.666667 -14.2049 becausethey 0.666667 -14.2049 ufo 0.5 -14.2055 xin 0.4 -14.2059 text type 0.210526 -14.206 kneser-ney smoothing 0.666667 -14.2062 apollo 0.666667 -14.2062 4.38 0.666667 -14.2062 descendants) 0.666667 -14.2062 quotations, 0.666667 -14.2062 li,j 0.666667 -14.2062 goat 0.666667 -14.2062 nachfeld 0.666667 -14.2062 fragment: 0.666667 -14.2062 telecom 0.666667 -14.2062 blinding 0.666667 -14.2062 amantis, 0.666667 -14.2062 2000m 0.666667 -14.2062 “friends” 0.666667 -14.2062 same: 0.666667 -14.2062 discriminatively-trained 0.666667 -14.2062 ramanathan 0.666667 -14.2062 sub-cases 0.666667 -14.2062 document-labeling 0.666667 -14.2062 sync 0.666667 -14.2062 touches 0.666667 -14.2062 re-implementing 0.666667 -14.2062 heuristic-based 0.666667 -14.2062 1989 0.666667 -14.2062 so-tp 0.666667 -14.2062 (epsilon1,δ) 0.666667 -14.2062 post- 0.666667 -14.2062 latentvariable 0.666667 -14.2062 goodness(s) 0.666667 -14.2062 source-channel 0.666667 -14.2062 qian, 0.666667 -14.2062 mnr−1 0.5 -14.2064 prt 0.5 -14.2064 kenya, 0.152778 -14.2067 al., 2004), 0.0371567 -14.2067 the original 0.5 -14.2074 mann, 0.666667 -14.2075 decimals) 0.666667 -14.2075 bracket, 0.666667 -14.2075 feedbacktype, 0.666667 -14.2075 13.0% 0.666667 -14.2075 hpdictionary 0.666667 -14.2075 0.361 0.666667 -14.2075 subframe 0.666667 -14.2075 canceling 0.666667 -14.2075 imdb, 0.666667 -14.2075 52.5% 0.666667 -14.2075 (omitting 0.666667 -14.2075 information-delivering 0.666667 -14.2075 incoherence 0.666667 -14.2075 illformed, 0.666667 -14.2075 s(wn1) 0.666667 -14.2075 jette 0.666667 -14.2075 §5, 0.666667 -14.2075 0.233 0.666667 -14.2075 text-independent 0.666667 -14.2075 sridhar 0.666667 -14.2075 (hofmann 0.666667 -14.2075 menstrual 0.666667 -14.2075 explicating 0.666667 -14.2075 θunq 0.148649 -14.2077 knight, 0.666667 -14.2088 nh 0.666667 -14.2088 62.81 0.666667 -14.2088 loonybin 0.666667 -14.2088 charity 0.666667 -14.2088 messy 0.666667 -14.2088 3.) 0.666667 -14.2088 46.2% 0.666667 -14.2088 0.771 0.666667 -14.2088 transmit 0.666667 -14.2088 slot2], 0.666667 -14.2088 (national 0.666667 -14.2088 (free 0.666667 -14.2088 atama 0.666667 -14.2088 pennconverter 0.666667 -14.2088 behavior: 0.666667 -14.2088 4.32 0.666667 -14.2088 configure 0.666667 -14.2088 2010c), 0.666667 -14.2088 88.95 0.666667 -14.2088 “yet” 0.666667 -14.2088 m22 0.666667 -14.2088 sumo, 0.666667 -14.2088 c(xi) 0.666667 -14.2088 scope-3 0.4 -14.209 thank you 0.110294 -14.2097 – – 0.666667 -14.21 the”, 0.666667 -14.21 500000 0.666667 -14.21 extra-domain 0.666667 -14.21 ando 0.666667 -14.21 lvc-specific 0.666667 -14.21 cross-referencing 0.666667 -14.21 53.77 0.666667 -14.21 antisymmetry 0.666667 -14.21 80.6% 0.666667 -14.21 “very” 0.666667 -14.21 constructions) 0.666667 -14.21 carenini, 0.666667 -14.21 (4.a) 0.666667 -14.21 90.98 0.666667 -14.21 manufacturer 0.666667 -14.21 (beigman 0.666667 -14.21 1447 0.666667 -14.21 baseline+lexical 0.666667 -14.21 unigrams; 0.666667 -14.21 (room) 0.666667 -14.21 0.89, 0.666667 -14.21 (may, 0.666667 -14.21 cascata 0.666667 -14.21 undue 0.666667 -14.21 upshot 0.666667 -14.21 containing.r 0.666667 -14.21 etc.); 0.666667 -14.21 substances, 0.666667 -14.21 “arrested” 0.666667 -14.21 medline, 0.666667 -14.21 secrets 0.666667 -14.21 pennell 0.233333 -14.2102 can serve 0.363636 -14.2106 mrc 0.666667 -14.2113 (german, 0.666667 -14.2113 rectangular 0.666667 -14.2113 relationcardinality 0.666667 -14.2113 39.76 0.666667 -14.2113 so(w) 0.666667 -14.2113 =∇1 0.666667 -14.2113 stanzas, 0.666667 -14.2113 postings, 0.666667 -14.2113 (qiu 0.666667 -14.2113 cubic-time 0.666667 -14.2113 allergy 0.666667 -14.2113 effected 0.666667 -14.2113 46.22 0.666667 -14.2113 disabled, 0.666667 -14.2113 '(*/) 0.666667 -14.2113 mending 0.666667 -14.2113 33.52 0.666667 -14.2113 purities 0.666667 -14.2113 all-fields 0.666667 -14.2113 summary) 0.666667 -14.2113 1,377 0.666667 -14.2113 g1, 0.666667 -14.2113 bionlpst 0.666667 -14.2113 (incorrect 0.666667 -14.2113 moderate, 0.666667 -14.2113 0.950 0.666667 -14.2113 herrmann 0.5 -14.2113 3700 0.5 -14.2113 stein 0.3125 -14.2115 pscfgs 0.5 -14.2122 universidade 0.5 -14.2122 delete, 0.163934 -14.2125 sentence s 0.666667 -14.2126 kinda 0.666667 -14.2126 (geo-political 0.666667 -14.2126 qwerty 0.666667 -14.2126 4.42 0.666667 -14.2126 studiesdisease() 0.666667 -14.2126 (l1) 0.666667 -14.2126 crl(wprime1,wprime2) 0.666667 -14.2126 ofer 0.666667 -14.2126 (real-world 0.666667 -14.2126 quran 0.666667 -14.2126 1,412 0.666667 -14.2126 (ballesteros 0.666667 -14.2126 newssyscombtune2011 0.666667 -14.2126 auto-alignment 0.666667 -14.2126 mt-05 0.666667 -14.2126 (phonetic 0.666667 -14.2126 submodule 0.666667 -14.2126 20), 0.666667 -14.2126 decision: 0.666667 -14.2126 circled 0.666667 -14.2126 architecture: 0.666667 -14.2126 spoil 0.666667 -14.2126 jing, 0.666667 -14.2126 cross-experimental 0.666667 -14.2126 5.0% 0.666667 -14.2126 wild-type 0.666667 -14.2126 synthesizing 0.666667 -14.2126 decision) 0.666667 -14.2126 corrupt 0.666667 -14.2126 substance, 0.666667 -14.2126 batter 0.666667 -14.2126 glove 0.109489 -14.2129 we created 0.5 -14.2132 cn, 0.5 -14.2132 adut 0.666667 -14.2139 two-year 0.666667 -14.2139 nullijk 0.666667 -14.2139 hatwidefxi 0.666667 -14.2139 67.0% 0.666667 -14.2139 92.63 0.666667 -14.2139 next) 0.666667 -14.2139 self-defense 0.666667 -14.2139 clause’s 0.666667 -14.2139 “experts” 0.666667 -14.2139 out-degree 0.666667 -14.2139 brake 0.666667 -14.2139 grouping, 0.666667 -14.2139 titles; 0.666667 -14.2139 constitution, 0.666667 -14.2139 tag-specific 0.666667 -14.2139 method), 0.666667 -14.2139 computed: 0.666667 -14.2139 (نم 0.666667 -14.2139 l-lda, 0.666667 -14.2139 non-person 0.666667 -14.2139 extensions, 0.666667 -14.2139 left-arcl 0.666667 -14.2139 3.9% 0.666667 -14.2139 credit, 0.15873 -14.214 average. 0.363636 -14.2141 (grishman 0.101911 -14.2142 (och 0.5 -14.2142 0.883 0.666667 -14.2152 (pmi 0.666667 -14.2152 87.45 0.666667 -14.2152 2000); 0.666667 -14.2152 z-test 0.666667 -14.2152 ˜ φ 0.666667 -14.2152 α2, 0.666667 -14.2152 event) 0.666667 -14.2152 -0.94 0.666667 -14.2152 visually, 0.666667 -14.2152 “complement” 0.666667 -14.2152 (%)' 0.666667 -14.2152 f1score 0.666667 -14.2152 25.4% 0.666667 -14.2152 ‘but’, 0.666667 -14.2152 substantiate 0.666667 -14.2152 cheapest 0.666667 -14.2152 micro-blog 0.666667 -14.2152 bengio 0.666667 -14.2152 started, 0.666667 -14.2152 β’s 0.666667 -14.2152 incrementality, 0.666667 -14.2152 non-ontology 0.666667 -14.2152 dr-specific 0.666667 -14.2152 akıcı,and 0.666667 -14.2152 robot, 0.666667 -14.2152 0.95, 0.666667 -14.2152 rgb, 0.666667 -14.2152 s(hk) 0.428571 -14.2155 impr. 0.428571 -14.2155 grow-diag-final 0.294118 -14.2156 one should 0.1875 -14.2161 as a consequence, 0.363636 -14.2162 funds 0.3125 -14.2163 4.4. 0.428571 -14.2163 z z 0.666667 -14.2165 dbaseline 0.666667 -14.2165 norway 0.666667 -14.2165 kohli, 0.666667 -14.2165 systematize 0.666667 -14.2165 japanese-toenglish 0.666667 -14.2165 cooperates 0.666667 -14.2165 68.46 0.666667 -14.2165 positive-negative 0.666667 -14.2165 hasegawa, 0.666667 -14.2165 locus, 0.666667 -14.2165 false-negative 0.666667 -14.2165 50.10 0.666667 -14.2165 staying 0.666667 -14.2165 ‘materials 0.666667 -14.2165 lexicon-grammar 0.666667 -14.2165 coaching 0.666667 -14.2165 endow 0.666667 -14.2165 conection 0.666667 -14.2165 confusion: 0.666667 -14.2165 (fowler 0.666667 -14.2165 newstest2010, 0.666667 -14.2165 (piwek 0.666667 -14.2165 27.69 0.666667 -14.2165 training-test 0.666667 -14.2165 lives, 0.666667 -14.2165 inverted) 0.666667 -14.2165 tree’s 0.666667 -14.2165 zi,j,ww 0.666667 -14.2165 lumped 0.363636 -14.217 texttiling 0.666667 -14.2178 rouge2 0.666667 -14.2178 (event%1:03:00 0.666667 -14.2178 15.6% 0.666667 -14.2178 1131 0.666667 -14.2178 beigman 0.666667 -14.2178 (spelling 0.666667 -14.2178 0.410 0.666667 -14.2178 wehave 0.666667 -14.2178 usp, 0.666667 -14.2178 “ground-truth” 0.666667 -14.2178 fonts 0.666667 -14.2178 1788 0.666667 -14.2178 52.36 0.666667 -14.2178 (schlangen 0.666667 -14.2178 multi-perspective 0.666667 -14.2178 1974), 0.666667 -14.2178 remediating 0.666667 -14.2178 ill-formedness 0.666667 -14.2178 tradeoff, 0.666667 -14.2178 thumb 0.666667 -14.2178 scripted 0.666667 -14.2178 4(c) 0.666667 -14.2178 repeatlast 0.666667 -14.2178 24.74 0.666667 -14.2178 keselj 0.666667 -14.2178 homonym 0.666667 -14.2178 egyptian, 0.666667 -14.2178 mi,j, 0.666667 -14.2178 2.76 0.666667 -14.2178 84.32 0.666667 -14.2178 0.666667 -14.2309 (ts) 0.666667 -14.2309 (tokens 0.666667 -14.2309 30.63 0.666667 -14.2309 (final 0.666667 -14.2309 humanauthored 0.666667 -14.2309 over-generation 0.666667 -14.2309 y/n 0.666667 -14.2309 regularity, 0.666667 -14.2309 0.243 0.666667 -14.2309 27.54 0.4 -14.2309 britain 0.333333 -14.2315 pang, 0.5 -14.2318 0.696 0.5 -14.2318 pi1 0.5 -14.2318 rotaru, 0.5 -14.2318 se, 0.25 -14.2319 assistive 0.666667 -14.2322 centre, 0.666667 -14.2322 29.50 0.666667 -14.2322 nj: 0.666667 -14.2322 deemphasized 0.666667 -14.2322 (rouge) 0.666667 -14.2322 models), 0.666667 -14.2322 duffy 0.666667 -14.2322 psk 0.666667 -14.2322 colocational 0.666667 -14.2322 cnls 0.666667 -14.2322 city) 0.666667 -14.2322 2011): 0.666667 -14.2322 hmm-type-r 0.666667 -14.2322 setyprime 0.666667 -14.2322 (lacoste-julien 0.666667 -14.2322 undertake 0.666667 -14.2322 sub-topics 0.666667 -14.2322 id), 0.666667 -14.2322 ontonotes: 0.666667 -14.2322 burns 0.666667 -14.2322 [si] 0.666667 -14.2322 wurde 0.666667 -14.2322 4.5m 0.666667 -14.2322 54.8% 0.666667 -14.2322 over-fitted 0.26087 -14.2322 tetreault 0.428571 -14.2323 northeastern 0.3125 -14.2329 databases. 0.363636 -14.2334 notation for 0.666667 -14.2335 incentive 0.666667 -14.2335 hitherto 0.666667 -14.2335 consistent: 0.666667 -14.2335 signature: 0.666667 -14.2335 leheˇcka 0.666667 -14.2335 “features” 0.666667 -14.2335 26.46 0.666667 -14.2335 82.5% 0.666667 -14.2335 filter-based 0.666667 -14.2335 +0.56 0.666667 -14.2335 mentionentity 0.666667 -14.2335 83.23 0.666667 -14.2335 biochemical 0.666667 -14.2335 respectively); 0.666667 -14.2335 día 0.666667 -14.2335 foremost, 0.666667 -14.2335 svcs 0.666667 -14.2335 koppel, 0.666667 -14.2335 postprocessed 0.666667 -14.2335 (36%) 0.666667 -14.2335 vq 0.225806 -14.2335 the 15th 0.5 -14.2338 nacn 0.363636 -14.2341 cao 0.5 -14.2348 inequalities 0.5 -14.2348 537 0.666667 -14.2349 slang, 0.666667 -14.2349 0.2500 0.666667 -14.2349 back-end 0.666667 -14.2349 footballer 0.666667 -14.2349 cross-system 0.666667 -14.2349 opinions; 0.666667 -14.2349 grades, 0.666667 -14.2349 (deterministic) 0.666667 -14.2349 5as 0.666667 -14.2349 combinator 0.666667 -14.2349 84.54 0.666667 -14.2349 55m 0.666667 -14.2349 67.90 0.666667 -14.2349 validity; 0.666667 -14.2349 (title 0.666667 -14.2349 pstrict 0.666667 -14.2349 irrelevant) 0.666667 -14.2349 on-demand 0.666667 -14.2349 ¯nt 0.666667 -14.2349 nagy 0.666667 -14.2349 patient), 0.333333 -14.2355 also compute 0.5 -14.2358 (banea 0.666667 -14.2362 intra- 0.666667 -14.2362 waveform 0.666667 -14.2362 bossy 0.666667 -14.2362 reify 0.666667 -14.2362 graph–ptb 0.666667 -14.2362 beach 0.666667 -14.2362 λ(ω,vectorv).[ω 0.666667 -14.2362 (sun, 0.666667 -14.2362 conclude: 0.666667 -14.2362 sense), 0.666667 -14.2362 a12 0.666667 -14.2362 lowe 0.666667 -14.2362 task-oriented, 0.666667 -14.2362 column; 0.666667 -14.2362 (washtell 0.666667 -14.2362 1059 0.666667 -14.2362 palavras 0.666667 -14.2362 destructor 0.666667 -14.2362 handcurated 0.294118 -14.2364 papers are 0.195122 -14.2364 (note 0.428571 -14.2365 schein 0.5 -14.2368 authorities 0.112 -14.2371 conditioned 0.428571 -14.2374 (vsm) 0.666667 -14.2375 f(a 0.666667 -14.2375 disambiguation) 0.666667 -14.2375 round-robin 0.666667 -14.2375 type-logic 0.666667 -14.2375 timeout, 0.666667 -14.2375 operations) 0.666667 -14.2375 wb, 0.666667 -14.2375 ash 0.666667 -14.2375 2.46 0.666667 -14.2375 (preposition 0.666667 -14.2375 seventy 0.666667 -14.2375 920 0.666667 -14.2375 toolkit.2 0.666667 -14.2375 selected) 0.205128 -14.2377 this research is 0.163934 -14.2382 has only 0.428571 -14.2382 θa 0.666667 -14.2388 cameras, 0.666667 -14.2388 βa 0.666667 -14.2388 48.59 0.666667 -14.2388 re-train 0.666667 -14.2388 3.87 0.666667 -14.2388 “effect” 0.666667 -14.2388 79.43 0.666667 -14.2388 decoder) 0.666667 -14.2388 3-level 0.666667 -14.2388 roles) 0.666667 -14.2388 (frantzi 0.666667 -14.2388 speaking) 0.666667 -14.2388 (disjoint 0.666667 -14.2388 samaas 0.666667 -14.2388 11.14 0.666667 -14.2388 438) 0.666667 -14.2388 ceaf, 0.666667 -14.2388 agree; 0.428571 -14.2391 sweden, 0.428571 -14.24 knight (2009) 0.666667 -14.2402 “ed” 0.666667 -14.2402 protecting 0.666667 -14.2402 rate), 0.666667 -14.2402 (mixed 0.666667 -14.2402 79.3% 0.666667 -14.2402 prompts, 0.666667 -14.2402 (pseudo) 0.666667 -14.2402 refit 0.666667 -14.2402 2000)), 0.666667 -14.2402 you’ve 0.666667 -14.2402 gregor 0.666667 -14.2402 endeavor, 0.666667 -14.2402 traveled 0.666667 -14.2402 marker) 0.666667 -14.2402 another) 0.666667 -14.2402 84.3% 0.666667 -14.2402 o(∗) 0.666667 -14.2402 (domain 0.666667 -14.2402 (boneva 0.666667 -14.2402 dels 0.666667 -14.2402 self: 0.666667 -14.2402 particular) 0.666667 -14.2402 investments 0.666667 -14.2402 match(hs\it 0.225806 -14.2402 conjecture 0.4 -14.2404 40k 0.233333 -14.2413 as previously 0.363636 -14.2413 6×104 0.666667 -14.2415 interaction; 0.666667 -14.2415 bloody 0.666667 -14.2415 vancouver, 0.666667 -14.2415 (ravindra 0.666667 -14.2415 best-suited 0.666667 -14.2415 articulate 0.666667 -14.2415 33.50 0.666667 -14.2415 manual) 0.666667 -14.2415 1school 0.666667 -14.2415 board; 0.666667 -14.2415 1973), 0.666667 -14.2415 difficult; 0.666667 -14.2415 snapshot, 0.666667 -14.2415 22.9% 0.666667 -14.2415 (50%) 0.5 -14.2418 ωr 0.5 -14.2418 (roc) 0.5 -14.2418 “informative” 0.363636 -14.2421 null5 0.3125 -14.2422 statement. 0.118644 -14.2426 wish 0.666667 -14.2428 conditions) 0.666667 -14.2428 segw 0.666667 -14.2428 regularizers, 0.666667 -14.2428 ]s1 0.666667 -14.2428 749 0.666667 -14.2428 pattern; 0.666667 -14.2428 insures 0.666667 -14.2428 reportedly 0.666667 -14.2428 26.31 0.666667 -14.2428 (wtm) 0.666667 -14.2428 0.306 0.666667 -14.2428 viewpoints, 0.666667 -14.2428 rumorous 0.666667 -14.2428 31.82 0.333333 -14.2428 motorbike 0.5 -14.2428 23.81 0.173077 -14.2434 mooney 0.5 -14.2438 triples: 0.666667 -14.2442 summationtextni=1zi 0.666667 -14.2442 nonsymmetric 0.666667 -14.2442 4.02 0.666667 -14.2442 51.80 0.666667 -14.2442 twofold 0.666667 -14.2442 unrelated, 0.666667 -14.2442 multi-language 0.666667 -14.2442 relational-clustering 0.666667 -14.2442 dr, 0.666667 -14.2442 xxx, 0.666667 -14.2442 score.t;ct/ 0.666667 -14.2442 37.85 0.666667 -14.2442 desperately 0.666667 -14.2442 (observed 0.666667 -14.2442 (rousu 0.666667 -14.2442 25.64 0.666667 -14.2442 martinez, 0.666667 -14.2442 discoverable 0.666667 -14.2442 sampling; 0.428571 -14.2442 mentions per 0.5 -14.2448 martinez 0.666667 -14.2455 48.77 0.666667 -14.2455 asmira 0.666667 -14.2455 205; 0.666667 -14.2455 81.32 0.666667 -14.2455 ecti 0.666667 -14.2455 pre-selection 0.666667 -14.2455 (users 0.666667 -14.2455 ptm, 0.666667 -14.2455 (1989); 0.666667 -14.2455 manifold 0.666667 -14.2455 3some 0.666667 -14.2455 novelties 0.666667 -14.2455 78.25 0.666667 -14.2455 unified, 0.666667 -14.2455 bootstraping 0.666667 -14.2455 hofman 0.666667 -14.2455 hinrichs, 0.666667 -14.2455 wikipedias 0.666667 -14.2455 suggestion: 0.666667 -14.2455 fractions, 0.666667 -14.2455 poured 0.666667 -14.2455 homogenous 0.666667 -14.2455 r∈d 0.333333 -14.2455 p(d) 0.277778 -14.2455 (2007) use 0.5 -14.2458 citations, 0.5 -14.2458 a1h 0.428571 -14.246 3t 0.428571 -14.246 gerundive 0.666667 -14.2468 tha 0.666667 -14.2468 662 0.666667 -14.2468 dip 0.666667 -14.2468 top-line 0.666667 -14.2468 4.25 0.666667 -14.2468 “except 0.666667 -14.2468 oneword 0.666667 -14.2468 jochem 0.666667 -14.2468 disambiguated, 0.666667 -14.2468 non-fringe 0.666667 -14.2468 (leech, 0.666667 -14.2468 cater 0.666667 -14.2468 02-21), 0.666667 -14.2468 :affected 0.666667 -14.2468 (nominal) 0.666667 -14.2468 bertsekas 0.666667 -14.2468 sensors, 0.666667 -14.2468 18.89 0.666667 -14.2468 (illustrated 0.666667 -14.2468 η0 0.666667 -14.2468 proficient 0.666667 -14.2468 ms) 0.666667 -14.2468 mundane 0.666667 -14.2468 ain 0.666667 -14.2468 prepositions), 0.666667 -14.2468 blocs 0.666667 -14.2468 (per-corpus 0.666667 -14.2468 conflicts, 0.666667 -14.2468 (joint 0.666667 -14.2468 analytically, 0.666667 -14.2468 her, 0.666667 -14.2468 mishra 0.5 -14.2468 early-stopping 0.5 -14.2468 company’s 0.183673 -14.2473 two versions 0.4 -14.2476 expletive 0.666667 -14.2482 higher-scoring 0.666667 -14.2482 foc-det 0.666667 -14.2482 recapitulate 0.666667 -14.2482 interrogative, 0.666667 -14.2482 2.54 0.666667 -14.2482 voting: 0.666667 -14.2482 phagocytes, 0.666667 -14.2482 fortunately 0.666667 -14.2482 “several” 0.666667 -14.2482 ccb 0.666667 -14.2482 0.069 0.666667 -14.2482 lerner 0.666667 -14.2482 84.10 0.666667 -14.2482 ∀k 0.666667 -14.2482 tokenize, 0.666667 -14.2482 soft, 0.666667 -14.2482 situation; 0.666667 -14.2482 singleton, 0.666667 -14.2482 stest 0.666667 -14.2482 controversies 0.166667 -14.2485 learning from 0.666667 -14.2495 p13 0.666667 -14.2495 concordance, 0.666667 -14.2495 merlo 0.666667 -14.2495 ravens 0.666667 -14.2495 summer, 0.666667 -14.2495 multinational 0.666667 -14.2495 halliday, 0.666667 -14.2495 wsi, 0.666667 -14.2495 adj) 0.666667 -14.2495 approximations, 0.666667 -14.2495 average-expert 0.666667 -14.2495 o* 0.666667 -14.2495 wickets 0.666667 -14.2495 (nullnullnullnullnull) 0.666667 -14.2495 links/images 0.666667 -14.2495 81.10 0.333333 -14.2495 eθg-rl∗,d(cη 0.5 -14.2498 p17�s 0.428571 -14.2503 l(ψ) 0.146667 -14.2505 have proposed 0.666667 -14.2509 wβ) 0.666667 -14.2509 piled 0.666667 -14.2509 ill-formed, 0.666667 -14.2509 annotated: 0.666667 -14.2509 “base” 0.666667 -14.2509 buckets, 0.666667 -14.2509 “exact 0.666667 -14.2509 79.24 0.666667 -14.2509 21.11 0.666667 -14.2509 kok, 0.666667 -14.2509 “views” 0.666667 -14.2509 symmetric; 0.666667 -14.2509 1216 0.666667 -14.2509 vascular 0.666667 -14.2509 bythe 0.666667 -14.2509 stokoe 0.666667 -14.2509 hogan, 0.666667 -14.2509 j.» 0.666667 -14.2509 appeals 0.666667 -14.2509 double-blind 0.26087 -14.251 humanities, 0.428571 -14.252 avi 0.666667 -14.2522 classic, 0.666667 -14.2522 hours), 0.666667 -14.2522 bader, 0.666667 -14.2522 marytts 0.666667 -14.2522 azdc 0.666667 -14.2522 fapesp 0.666667 -14.2522 sullenberger 0.666667 -14.2522 47.16 0.666667 -14.2522 rerunning 0.666667 -14.2522 guitar 0.666667 -14.2522 (fraction 0.666667 -14.2522 (x,z) 0.666667 -14.2522 [bellare 0.666667 -14.2522 suband 0.428571 -14.2529 x0 0.666667 -14.2536 ptsgs 0.666667 -14.2536 gnd 0.666667 -14.2536 relabeled 0.666667 -14.2536 0.817 0.666667 -14.2536 ual 0.666667 -14.2536 1997); 0.666667 -14.2536  0.666667 -14.2536 aggregate statistics 0.666667 -14.2536 (1d) 0.666667 -14.2536 (leading 0.666667 -14.2536 two) 0.666667 -14.2536 0.230 0.666667 -14.2536 89.21 0.666667 -14.2536 fragments; 0.666667 -14.2536 dc0;:::;cm 0.666667 -14.2536 (proposed) 0.666667 -14.2536 87.12 0.666667 -14.2536 specia, 0.666667 -14.2536 confirmed: 0.666667 -14.2536 pittsburgh) 0.666667 -14.2536 emptyelement 0.0353303 -14.2538 2: 0.5 -14.2538 0.547 0.21875 -14.2541 lagrange 0.5 -14.2548 texter 2: 0.666667 -14.2549 haifeng 0.666667 -14.2549 two; 0.666667 -14.2549 jebara 0.666667 -14.2549 max-norm 0.666667 -14.2549 methodologies, 0.666667 -14.2549 (bollegala 0.666667 -14.2549 [per 0.666667 -14.2549 32.87 0.666667 -14.2549 topicspecific 0.666667 -14.2549 t2j 0.666667 -14.2549 48.87 0.666667 -14.2549 uncausative 0.666667 -14.2549 rpos 0.666667 -14.2549 31.81 0.666667 -14.2549 dash, 0.666667 -14.2549 collaborating 0.666667 -14.2549 gi ven 0.666667 -14.2549 (j), 0.666667 -14.2549 gestures: 0.666667 -14.2549 overgenerates 0.666667 -14.2549 wordsin 0.666667 -14.2549 3.95 0.666667 -14.2549 grayed 0.666667 -14.2549 processing; 0.139241 -14.2552 we estimate 0.0949721 -14.2552 and evaluation 0.272727 -14.2556 (witten 0.4 -14.2557 arts, 0.5 -14.2558 (isozaki et al., 0.666667 -14.2563 presentations, 0.666667 -14.2563 searching, 0.666667 -14.2563 goldsyn 0.666667 -14.2563 nullnull2 0.666667 -14.2563 33-attr 0.666667 -14.2563 genia,id,epi 0.666667 -14.2563 θ) 0.666667 -14.2563 65.49 0.666667 -14.2563 learner) 0.666667 -14.2563 holliman, 0.666667 -14.2563 broadens 0.666667 -14.2563 falls, 0.666667 -14.2563 gabbard 0.666667 -14.2563 amounted 0.666667 -14.2563 c(b 0.666667 -14.2563 hand-authored 0.666667 -14.2563 (noun) 0.666667 -14.2563 inanimate 0.428571 -14.2563 is-a, 0.428571 -14.2563 λ1, 0.428571 -14.2572 werden 0.428571 -14.2572 head-complement 0.21875 -14.2572 dc 0.666667 -14.2576 vk(x) 0.666667 -14.2576 -0.43 0.666667 -14.2576 proactively 0.666667 -14.2576 766 0.666667 -14.2576 tobi, 0.666667 -14.2576 ands2 0.666667 -14.2576 38.10 0.666667 -14.2576 6.3.2 0.666667 -14.2576 chips 0.666667 -14.2576 emu 0.666667 -14.2576 1368 0.666667 -14.2576 prominent, 0.666667 -14.2576 apple” 0.666667 -14.2576 explicitation, 0.666667 -14.2576 (ostendorf 0.666667 -14.2576 zi,ja 0.666667 -14.2576 rte-like 0.666667 -14.2576 population-center 0.666667 -14.2576 17.02 0.363636 -14.2589 feature classes 0.5 -14.2589 deubiquitination 0.666667 -14.259 92.82 0.666667 -14.259 fire, 0.666667 -14.259 anatomical 0.666667 -14.259 explainable 0.666667 -14.259 venkatagiri 0.666667 -14.259 segments) 0.666667 -14.259 am, 0.666667 -14.259 s3vms 0.666667 -14.259 combined) 0.666667 -14.259 bollegala 0.666667 -14.259 justifiable 0.666667 -14.259 1137 0.666667 -14.259 0.186 0.666667 -14.259 consentcanvas, 0.666667 -14.259 agreeable 0.3125 -14.2592 2006b; 0.15873 -14.2594 systems were 0.0582121 -14.2598 as shown 0.428571 -14.2598 belief, 0.5 -14.2599 glyph2 0.186047 -14.2603 up. 0.666667 -14.2603 barrels 0.666667 -14.2603 atachment 0.666667 -14.2603 natural-sounding 0.666667 -14.2603 performing, 0.666667 -14.2603 maxent) 0.666667 -14.2603 是 0.666667 -14.2603 six-lingual 0.666667 -14.2603 equivalent; 0.666667 -14.2603 framesets 0.666667 -14.2603 letter-to-sound 0.666667 -14.2603 maltparser’s 0.666667 -14.2603 ffr 0.666667 -14.2603 ii(+iii) 0.666667 -14.2603 ration 0.666667 -14.2603 scenes: 0.666667 -14.2603 millersville 0.666667 -14.2603 amarnag 0.666667 -14.2603 mittelfeld 0.666667 -14.2603 test-bed 0.666667 -14.2603 nasraoui, 0.666667 -14.2603 lowdensity 0.666667 -14.2603 12.57 0.0152357 -14.2607 number 0.130435 -14.2611 utterance. 0.666667 -14.2617 prepositions: 0.666667 -14.2617 klenner 0.666667 -14.2617 prank 0.666667 -14.2617 aloud 0.666667 -14.2617 88.4% 0.666667 -14.2617 sawaf 0.666667 -14.2617 muler 0.666667 -14.2617 (viterbi 0.666667 -14.2617 expressly 0.666667 -14.2617 x] 0.666667 -14.2617 confused, 0.666667 -14.2617 t-units, 0.666667 -14.2617 ambadar, 0.666667 -14.2617 “new” 0.666667 -14.2617 belgian 0.666667 -14.2617 15.44 0.666667 -14.2617 (selected 0.666667 -14.2617 △ 0.0818966 -14.262 conclude 0.666667 -14.263 jiao 0.666667 -14.263 “meaning” 0.666667 -14.263 87.06 0.666667 -14.263 np/z 0.666667 -14.263 moses-oracle 0.666667 -14.263 recall/precision 0.666667 -14.263 clas; 0.666667 -14.263 p14 0.666667 -14.263 perword 0.666667 -14.263 ni, 0.666667 -14.263 mareˇcek, 0.666667 -14.263 23.99 0.666667 -14.263 (u), 0.666667 -14.263 orlando, 0.666667 -14.263 luther 0.666667 -14.263 a150a151a152a143a144a145a146 0.666667 -14.263 4.98 0.666667 -14.263 1184 0.666667 -14.263 initiation) 0.666667 -14.263 em-trained 0.666667 -14.263 (bach 0.666667 -14.263 augs 0.428571 -14.2633 m∑ 0.122642 -14.2634 however, it is 0.4 -14.2638 (mt). 0.5 -14.264 42.75 0.5 -14.264 ip0,8 0.428571 -14.2642 .71 0.3125 -14.2643 universitat 0.666667 -14.2644 (h, 0.666667 -14.2644 predefined, 0.666667 -14.2644 mac, 0.666667 -14.2644 27.31 0.666667 -14.2644 0.042 0.666667 -14.2644 (atomic) 0.666667 -14.2644 codify 0.666667 -14.2644 mcallester 0.666667 -14.2644 ensembl, 0.666667 -14.2644 conquered 0.666667 -14.2644 6when 0.666667 -14.2644 increases; 0.666667 -14.2644 28.75 0.666667 -14.2644 ghastly 0.666667 -14.2644 (chawla 0.666667 -14.2644 τ∈ 0.666667 -14.2644 edict 0.666667 -14.2644 10.2 0.666667 -14.2644 (darpa/ipto) 0.4 -14.2646 rg 0.0519263 -14.2647 translation. 0.0606061 -14.265 as described 0.333333 -14.2657 by-product 0.666667 -14.2658 β-values 0.666667 -14.2658 swsd, 0.666667 -14.2658 slowest 0.666667 -14.2658 1,...,s 0.666667 -14.2658 cdt 0.666667 -14.2658 term(keyp) 0.666667 -14.2658 walk-through 0.666667 -14.2658 3,993 0.666667 -14.2658 695 0.666667 -14.2658 (it) 0.666667 -14.2658 93.27 0.666667 -14.2658 0.705 0.666667 -14.2658 operon 0.666667 -14.2658 head-out 0.666667 -14.2658 loanwords 0.666667 -14.2658 γl 0.666667 -14.2658 −−= 0.666667 -14.2658 21.17 0.666667 -14.2658 alright 0.666667 -14.2658 pakistan. 0.666667 -14.2658 100th 0.666667 -14.2658 sender-receiver 0.5 -14.266 chilean 0.134831 -14.2661 is too 0.4 -14.2663 geo: 0.4 -14.2663 throwing 0.363636 -14.2663 learninga 0.428571 -14.2668 gin 0.666667 -14.2671 cancedda, 0.666667 -14.2671 null(null3) 0.666667 -14.2671 treaty, 0.666667 -14.2671 effective 0.666667 -14.2671 rights, 0.666667 -14.2671 diag-nlp 0.666667 -14.2671 example,the 0.666667 -14.2671 85.56 0.666667 -14.2671 participial 0.666667 -14.2671 (nlu) 0.666667 -14.2671 (→ 0.666667 -14.2671 17.52 0.666667 -14.2671 (onishi 0.666667 -14.2671 1-6 0.666667 -14.2671 (mono 0.666667 -14.2671 wl2 0.666667 -14.2671 favor, 0.666667 -14.2671 populists 0.666667 -14.2671 amd 0.666667 -14.2685 20.8% 0.666667 -14.2685 88.5% 0.666667 -14.2685 mappings) 0.666667 -14.2685 parenrightbiggparenleftbigg 0.666667 -14.2685 chemistry, 0.666667 -14.2685 (10.a) 0.666667 -14.2685 negationslash∈ 0.666667 -14.2685 subtle, 0.666667 -14.2685 satisfied: 0.666667 -14.2685 84.87 0.666667 -14.2685 1058 0.666667 -14.2685 u(c) 0.666667 -14.2685 outcomes: 0.666667 -14.2685 ≥3 0.666667 -14.2685 {nl,iq 0.666667 -14.2685 15.00 0.666667 -14.2685 (besides 0.666667 -14.2685 corpora), 0.666667 -14.2685 φpop 0.666667 -14.2685 0.90, 0.666667 -14.2685 outputs; 0.666667 -14.2685 growth, 0.666667 -14.2685 “perfect” 0.0543071 -14.2685 supported 0.363636 -14.2686 rank> 0.16129 -14.2686 be quite 0.135802 -14.2689 be explained 0.5 -14.2691 nancy 0.24 -14.2692 k-nn graphs 0.666667 -14.2699 1stbl 0.666667 -14.2699 60.45 0.666667 -14.2699 42.57 0.666667 -14.2699 approach), 0.666667 -14.2699 emoticon-based 0.666667 -14.2699 33.05 0.666667 -14.2699 interlinearization 0.666667 -14.2699 (craswell 0.666667 -14.2699 ovat 0.666667 -14.2699 91.29 0.666667 -14.2699 (pra) 0.666667 -14.2699 67.27 0.666667 -14.2699 eulerian 0.666667 -14.2699 (interactive 0.666667 -14.2699 cross-lingually 0.666667 -14.2699 co-hyponyms, 0.666667 -14.2699 enew 0.666667 -14.2699 readymades, 0.666667 -14.2699 xslt 0.666667 -14.2712 xrt 0.666667 -14.2712 historians, 0.666667 -14.2712 lpst 0.666667 -14.2712 1.86 0.666667 -14.2712 collocates, 0.666667 -14.2712 θw 0.666667 -14.2712 (irrelevant 0.666667 -14.2712 precious 0.666667 -14.2712 622 0.666667 -14.2712 raters’ 0.666667 -14.2712 74.7% 0.666667 -14.2712 p16�p17 0.666667 -14.2712 grandmother 0.666667 -14.2712 triangle, 0.666667 -14.2712 problem-specific 0.666667 -14.2712 didactics 0.666667 -14.2712 feed, 0.666667 -14.2712 crowds, 0.666667 -14.2712 33.85 0.666667 -14.2712 51.69 0.4 -14.2712 (agarwal 0.25 -14.2712 node v 0.333333 -14.2712 t-units 0.5 -14.2722 f-structures 0.5 -14.2722 rithm 0.5 -14.2722 contained, 0.05386 -14.2723 hand, 0.666667 -14.2726 syn+hyper, 0.666667 -14.2726 t28 0.666667 -14.2726 svmperf 0.666667 -14.2726 68.57 0.666667 -14.2726 (sugiyama 0.666667 -14.2726 17.80 0.666667 -14.2726 (reckman 0.666667 -14.2726 dylan 0.666667 -14.2726 (gutiérrez 0.666667 -14.2726 pgprefix([v1,v2]) 0.666667 -14.2726 cleansed 0.666667 -14.2726 hexagonal 0.666667 -14.2726 ndcg@5 0.666667 -14.2726 host-part 0.666667 -14.2726 off-topic 0.666667 -14.2726 23.63 0.666667 -14.2726 ‘topic’ 0.666667 -14.2726 ubiquity 0.666667 -14.2726 94.73 0.666667 -14.2726 “lower 0.666667 -14.2726 clear: 0.666667 -14.2726 left-branching 0.181818 -14.2727 figure 1: an 0.5 -14.2732 0.5 -14.2732 theorems 0.0296215 -14.2737 we also 0.3125 -14.274 hereafter, 0.666667 -14.274 s∗, 0.666667 -14.274 rhmeant 0.666667 -14.274 61.90 0.666667 -14.274 multi-paragraph 0.666667 -14.274 e; 0.666667 -14.274 back-propagation 0.666667 -14.274 [w1, 0.666667 -14.274 2.72 0.666667 -14.274 non-related 0.666667 -14.274 (morin 0.666667 -14.274 pairings, 0.666667 -14.274 lipids 0.666667 -14.274 ‘computer’ 0.666667 -14.274 sites: 0.666667 -14.274 rumor, 0.666667 -14.274 /v/ 0.666667 -14.274 〈e,f〉 0.666667 -14.274 9,000 0.666667 -14.274 crime, 0.666667 -14.274 130m 0.666667 -14.274 0.267 0.666667 -14.274 4.56 0.666667 -14.274 evaluators, 0.666667 -14.274 0.089 0.666667 -14.274 kern 0.666667 -14.274 yeltsin 0.666667 -14.274 corenlp 0.666667 -14.274 (nell) 0.12037 -14.2741 hypothesize that 0.5 -14.2742 walton’s 0.333333 -14.2746 (moore 0.5 -14.2753 r1) 0.666667 -14.2754 blink 0.666667 -14.2754 reuses 0.666667 -14.2754 56.25 0.666667 -14.2754 40.33 0.666667 -14.2754 six-way 0.666667 -14.2754 lsp 0.666667 -14.2754 comprehensibility 0.666667 -14.2754 archaeological 0.666667 -14.2754 enough: 0.666667 -14.2754 2.02 0.666667 -14.2754 alif/ya 0.666667 -14.2754 (for), 0.666667 -14.2754 ban 0.666667 -14.2754 mini-document 0.666667 -14.2754 good/bad 0.666667 -14.2754 ‹a 0.666667 -14.2754 (breen, 0.666667 -14.2754 0.120 0.666667 -14.2754 homeomerous 0.428571 -14.2756 aria 0.277778 -14.2761 alternates 0.5 -14.2763 prop 0.428571 -14.2765 tongue 0.666667 -14.2767 highestscoring 0.666667 -14.2767 schapire 0.666667 -14.2767 vinf 0.666667 -14.2767 lbfgs-b 0.666667 -14.2767 39.43 0.666667 -14.2767 pre-verb 0.666667 -14.2767 falsity 0.666667 -14.2767 entities), 0.666667 -14.2767 vn-lvcs 0.666667 -14.2767 2– 0.666667 -14.2767 0.142 0.666667 -14.2767 58.90 0.666667 -14.2767 27.30 0.666667 -14.2767 1156 0.666667 -14.2767 autophosphorylation 0.666667 -14.2767 hand-crafted, 0.666667 -14.2767 457 0.666667 -14.2767 927 0.666667 -14.2767 attempt-to-persuade 0.666667 -14.2767 sampling-based 0.666667 -14.2767 initiated, 0.666667 -14.2767 non-prolific 0.666667 -14.2767 (1967) 0.666667 -14.2767 populates 0.666667 -14.2767 structure-preserving 0.12037 -14.2771 influenced 0.5 -14.2773 mihaylova 0.666667 -14.2781 α-values 0.666667 -14.2781 a108a108a44a44 0.666667 -14.2781 egg 0.666667 -14.2781 practices, 0.666667 -14.2781 syntactically-constrained 0.666667 -14.2781 class-wise 0.666667 -14.2781 static, 0.666667 -14.2781 cdp’s 0.666667 -14.2781 95.40 0.666667 -14.2781 subject/predicate 0.666667 -14.2781 tensors 0.666667 -14.2781 40.38 0.428571 -14.2783 (minnen 0.363636 -14.2783 salim 0.5 -14.2784 performs, 0.24 -14.2786 unsupervised semantic 0.363636 -14.2791 verbocean 0.5 -14.2794 plitg 0.5 -14.2794 nullnull1nullnull 0.5 -14.2794 only), 0.666667 -14.2795 word-correction 0.666667 -14.2795 0.123 0.666667 -14.2795 cross-linguistically 0.666667 -14.2795 (nsfc) 0.666667 -14.2795 0.200 0.666667 -14.2795 33.30 0.666667 -14.2795 experiment) 0.666667 -14.2795 2,5 0.666667 -14.2795 (kondrak 0.666667 -14.2795 item; 0.666667 -14.2795 0.666667 -14.2795 parametrization 0.666667 -14.2795 0.283 0.666667 -14.2795 (main 0.666667 -14.2795 opennlp2 0.666667 -14.2795 interleave 0.666667 -14.2795 elaborative 0.666667 -14.2795 (problem 0.666667 -14.2795 madonna 0.205882 -14.28  0.25 -14.28 remove all 0.5 -14.2805 steelers 0.666667 -14.2809 4500 0.666667 -14.2809 43.40 0.666667 -14.2809 dietrich 0.666667 -14.2809 war, 0.666667 -14.2809 sustain 0.666667 -14.2809 gaussian, 0.666667 -14.2809 bureaucratic 0.666667 -14.2809 87.55 0.666667 -14.2809 53.59 0.666667 -14.2809 qanus 0.666667 -14.2809 ng+lex 0.666667 -14.2809 ijcai 0.666667 -14.2809 knullda 0.666667 -14.2809 greater-than-the-mean 0.666667 -14.2809 product-of-experts 0.666667 -14.2809 bulgarianand 0.666667 -14.2809 e1,...,en 0.666667 -14.2809 1,824 0.5 -14.2815 87.20 0.666667 -14.2823 ”wrap” 0.666667 -14.2823 george, 0.666667 -14.2823 (kingsbury 0.666667 -14.2823 louisiana, 0.666667 -14.2823 gui, 0.666667 -14.2823 27.27 0.666667 -14.2823 singhal 0.666667 -14.2823 z’s 0.666667 -14.2823 events”, 0.666667 -14.2823 dmv-em 0.666667 -14.2823 1.35 0.666667 -14.2823 film, 0.666667 -14.2823 consequents 0.666667 -14.2823 81.69 0.666667 -14.2823 preferences: 0.666667 -14.2823 (partee, 0.666667 -14.2823 sug 0.666667 -14.2823 pragmatic-condition 0.666667 -14.2823 indicator, 0.666667 -14.2823 93.65 0.666667 -14.2823 russia, 0.666667 -14.2823 relieve 0.666667 -14.2823 42k 0.666667 -14.2823 56.2% 0.666667 -14.2823 post; 0.666667 -14.2823 66.43 0.0787402 -14.2823 parser. 0.428571 -14.2827 aditya 0.3125 -14.283 ´o 0.4 -14.2836 remark 0.5 -14.2836 subsenses 0.666667 -14.2837 1actually, 0.666667 -14.2837 docs), 0.666667 -14.2837 (manually) 0.666667 -14.2837 self-selection 0.666667 -14.2837 “tweets” 0.666667 -14.2837 magnitudes, 0.666667 -14.2837 mentally 0.666667 -14.2837 (ninomiya 0.666667 -14.2837 schools, 0.666667 -14.2837 55.06 0.666667 -14.2837 schedule, 0.666667 -14.2837 open-ended, 0.666667 -14.2837 minimized, 0.666667 -14.2837 post-test, 0.666667 -14.2837 14.99 0.666667 -14.2837 d(s) 0.666667 -14.2837 836 0.666667 -14.2837 k09 0.666667 -14.2837 69.57 0.666667 -14.2837 [+/–] 0.666667 -14.2837 [9] 0.428571 -14.2845 holmes, 0.5 -14.2846 factory 0.5 -14.2846 cure 0.666667 -14.285 p(wordk) 0.666667 -14.285 child) 0.666667 -14.285 (pls) 0.666667 -14.285 65.68 0.666667 -14.285 46.59 0.666667 -14.285 acl/dci 0.666667 -14.285 (ttm) 0.666667 -14.285 matwin, 0.666667 -14.285 format) 0.666667 -14.285 33.33% 0.666667 -14.285 s0h, 0.666667 -14.285 (wellner 0.666667 -14.285 93.81 0.666667 -14.285 92.21 0.666667 -14.285 businessman 0.666667 -14.285 λ(d) 0.666667 -14.285 blue) 0.666667 -14.285 bj(vj) 0.666667 -14.285 rossberg 0.333333 -14.285 10 100 0.190476 -14.2856 excerpt 0.12037 -14.2858 adjectives and 0.363636 -14.2859 mentions were 0.666667 -14.2864 wh-words 0.666667 -14.2864 modale 0.666667 -14.2864 indri 0.666667 -14.2864 thirteen 0.666667 -14.2864 rolesets, 0.666667 -14.2864 cancer) 0.666667 -14.2864 after) 0.666667 -14.2864 65.97 0.666667 -14.2864 yahoo, 0.666667 -14.2864 baa 0.666667 -14.2864 (setting: 0.666667 -14.2864 charlotte 0.666667 -14.2864 h4 0.666667 -14.2864 81.1% 0.666667 -14.2864 1a, 0.666667 -14.2864 83.39 0.666667 -14.2864 no-rel 0.666667 -14.2864 supersense 0.666667 -14.2864 string-similarity 0.666667 -14.2864 distancebased 0.666667 -14.2864 grossly 0.666667 -14.2878 %prob 0.666667 -14.2878 prb 0.666667 -14.2878 variation: 0.666667 -14.2878 np#1 0.666667 -14.2878 entrance 0.666667 -14.2878 charge, 0.666667 -14.2878 ‘true’ 0.666667 -14.2878 60, 0.666667 -14.2878 overview, 0.666667 -14.2878 (wrong) 0.666667 -14.2878 66.2% 0.666667 -14.2878 “simple” 0.666667 -14.2878 times), 0.225806 -14.2882 preferable 0.428571 -14.289 xiaojun 0.666667 -14.2892 39.97 0.666667 -14.2892 movingwindow 0.666667 -14.2892 2σ2 0.666667 -14.2892 heightened 0.666667 -14.2892 hersh, 0.666667 -14.2892 0.509 0.666667 -14.2892 a,b 0.666667 -14.2892 “lists” 0.666667 -14.2892 (xi 0.666667 -14.2892 parser) 0.666667 -14.2892 identifiers, 0.666667 -14.2892 (ng) 0.666667 -14.2892 c(m,n) 0.666667 -14.2892 darpa’s 0.666667 -14.2892 50.59 0.666667 -14.2892 86.22 0.666667 -14.2892 closes 0.666667 -14.2892 standardized, 0.110236 -14.2894 all these 0.0625 -14.2895 basis 0.428571 -14.2899 experimentalresults 0.428571 -14.2899 ies 0.666667 -14.2906 macroand 0.666667 -14.2906 accords 0.666667 -14.2906 ym 0.666667 -14.2906 crf: 0.666667 -14.2906 (wordnet 0.666667 -14.2906 870 0.666667 -14.2906 +prior+cov+sib 0.666667 -14.2906 j∈s 0.666667 -14.2906 (13 0.666667 -14.2906 qiaozhu 0.666667 -14.2906 rda 0.666667 -14.2906 4,500 0.666667 -14.2906 tur 0.666667 -14.2906 45.65 0.666667 -14.2906 paraphraser, 0.3125 -14.2909 consumers’ 0.428571 -14.2917 z∈z 0.666667 -14.292 mois 0.666667 -14.292 directs 0.666667 -14.292 52.69 0.666667 -14.292 decoder; 0.666667 -14.292 captured, 0.666667 -14.292 1arabic 0.666667 -14.292 (ppht 0.666667 -14.292 (craig 0.666667 -14.292 chine 0.666667 -14.292 hobbs, 0.666667 -14.292 (0,1) 0.666667 -14.292 forms); 0.666667 -14.292 0.01; 0.666667 -14.292 ulc 0.666667 -14.292 alerted 0.666667 -14.292 uzh 0.666667 -14.292 +0.22 0.666667 -14.292 7.04 0.666667 -14.292 (likely 0.666667 -14.292 knesser-ney 0.666667 -14.292 token; 0.666667 -14.292 93.23 0.666667 -14.292 titled 0.205882 -14.2925 show, 0.5 -14.293 laughing 0.666667 -14.2934 667 0.666667 -14.2934 0.134 0.666667 -14.2934 @15 0.666667 -14.2934 np-hard, 0.666667 -14.2934 sidestep 0.666667 -14.2934 highly-connected 0.666667 -14.2934 203) 0.666667 -14.2934 ftib, 0.666667 -14.2934 prot 0.666667 -14.2934 tho 0.666667 -14.2934 h4: 0.666667 -14.2934 1273 0.666667 -14.2934 paul, 0.666667 -14.2934 “moves 0.666667 -14.2934 (now 0.666667 -14.2934 perceived, 0.666667 -14.2934 ’i’ 0.5 -14.294 57.22 0.666667 -14.2948 hdt, 0.666667 -14.2948 shortest, 0.666667 -14.2948 nlp-centric 0.666667 -14.2948 bridged 0.666667 -14.2948 blended 0.666667 -14.2948 88.1% 0.666667 -14.2948 subzone” 0.666667 -14.2948 jackets 0.666667 -14.2948 (very) 0.666667 -14.2948 (whitelaw 0.666667 -14.2948 extent) 0.666667 -14.2948 existing, 0.363636 -14.2951 carlson et 0.107692 -14.2956 existence 0.666667 -14.2962 args 0.666667 -14.2962 rayson, 0.666667 -14.2962 en-cz 0.666667 -14.2962 29.35 0.666667 -14.2962 (parsing) 0.666667 -14.2962 64bit 0.666667 -14.2962 86.50 0.666667 -14.2962 0.351 0.666667 -14.2962 stalling 0.666667 -14.2962 ipc 0.666667 -14.2962 costarray 0.428571 -14.2962 my, 0.5 -14.2962 injury, 0.0655738 -14.2971 methods for 0.428571 -14.2971 (content 0.428571 -14.2971 diego 0.666667 -14.2976 31.50 0.666667 -14.2976 nlp-based 0.666667 -14.2976 53.6% 0.666667 -14.2976 isss 0.666667 -14.2976 deciphered 0.666667 -14.2976 tal 0.666667 -14.2976 ibm-3 0.666667 -14.2976 89.30 0.666667 -14.2976 (4.9) 0.666667 -14.2976 short-distance 0.666667 -14.2976 u; 0.666667 -14.2976 collection: 0.666667 -14.2976 steady-state 0.666667 -14.2976 met: 0.666667 -14.2976 tice, 0.666667 -14.2976 28.41 0.666667 -14.2976 4.48 0.666667 -14.2976 mt-baseline 0.666667 -14.2976 n(x) 0.666667 -14.2976 bi-phrases 0.666667 -14.2976 10.26 0.5 -14.2983 selma 0.116071 -14.2985 we selected 0.666667 -14.2991 fullimpl, 0.666667 -14.2991 periphery 0.666667 -14.2991 ofn-grams 0.666667 -14.2991 ££part-of 0.666667 -14.2991 bedrooms 0.666667 -14.2991 scherer, 0.666667 -14.2991 literacy, 0.666667 -14.2991 anaphor, 0.666667 -14.2991 simplest, 0.666667 -14.2991 platt’s 0.666667 -14.2991 reactive 0.666667 -14.2991 hereditary 0.666667 -14.2991 52.50 0.666667 -14.2991 phrenology 0.666667 -14.2991 (speech 0.666667 -14.2991 gold’s 0.666667 -14.2991 subdivisions 0.666667 -14.2991 〈t,t〉-lexicographic 0.104895 -14.2993 convert 0.5 -14.2993 1.84 0.135802 -14.3001 eisner, 0.666667 -14.3005 50.90 0.666667 -14.3005 62.26 0.666667 -14.3005 sequence” 0.666667 -14.3005 synsets) 0.666667 -14.3005 7.28 0.666667 -14.3005 totally, 0.666667 -14.3005 word-pairs, 0.666667 -14.3005 seed-stage 0.666667 -14.3005 compounds: 0.666667 -14.3005 5.1%, 0.666667 -14.3005 (men 0.666667 -14.3005 93.5%, 0.666667 -14.3005 cao, 0.666667 -14.3005 paula 0.666667 -14.3005 0.175 0.666667 -14.3005 expresions, 0.666667 -14.3005 ones), 0.666667 -14.3005 nullkknull 0.666667 -14.3005 mushroom 0.666667 -14.3005 t(38) 0.666667 -14.3005 p2) 0.666667 -14.3005 leadership 0.333333 -14.3011 uppsala, 0.5 -14.3014 times; 0.666667 -14.3019 nakagawa, 0.666667 -14.3019 (1964) 0.666667 -14.3019 basic-level 0.666667 -14.3019 (atb) 0.666667 -14.3019 time-sliced 0.666667 -14.3019 (savova 0.666667 -14.3019 block” 0.666667 -14.3019 0.527 0.666667 -14.3019 template-like 0.666667 -14.3019 bytespan 0.666667 -14.3019 bin-2-lr 0.666667 -14.3019 syntax–semantics 0.666667 -14.3019 english-french, 0.666667 -14.3019 (akiba 0.666667 -14.3019 19k 0.666667 -14.3019 bijection 0.666667 -14.3019 l2crf 0.666667 -14.3019 planner, 0.428571 -14.3026 definition-only 0.428571 -14.3026 u) 0.3125 -14.3027 extensions. 0.363636 -14.3027 iran 0.333333 -14.3032 prominently 0.666667 -14.3033 patry 0.666667 -14.3033 d-dimensional 0.666667 -14.3033 a(h,r) 0.666667 -14.3033 pt(z) 0.666667 -14.3033 27.22 0.666667 -14.3033 2.5.2 0.666667 -14.3033 22.41 0.666667 -14.3033 children) 0.666667 -14.3033 lattice; 0.666667 -14.3033 “activities 0.666667 -14.3033 1010 0.666667 -14.3033 sail 0.666667 -14.3033 =(v 0.666667 -14.3033 10.0% 0.666667 -14.3033 liver 0.666667 -14.3033 died, 0.666667 -14.3033 sociolinguists 0.666667 -14.3033 walmart 0.666667 -14.3033 unstructured, 0.666667 -14.3033 vague, 0.0333333 -14.3035 linguistics 0.5 -14.3036 audimus 0.277778 -14.3041 talked 0.666667 -14.3047 perfect: 0.666667 -14.3047 plug-in 0.666667 -14.3047 1157 0.666667 -14.3047 rdbms 0.666667 -14.3047 -ly 0.666667 -14.3047 endeavored 0.666667 -14.3047 remarks, 0.666667 -14.3047 end-toend 0.666667 -14.3047 (okazaki 0.666667 -14.3047 meta-constraints 0.666667 -14.3047 arise: 0.666667 -14.3047 declaration 0.666667 -14.3047 kuo 0.0444727 -14.3056 respect 0.666667 -14.3061 42.10 0.666667 -14.3061 below; 0.666667 -14.3061 32.58 0.666667 -14.3061 connor 0.666667 -14.3061 ref, 0.666667 -14.3061 development/test 0.666667 -14.3061 one-hour 0.666667 -14.3061 idds 0.666667 -14.3061 taging, 0.666667 -14.3061 s5: 0.666667 -14.3061 stretch 0.666667 -14.3061 wikipedia-derived 0.666667 -14.3061 lvs 0.666667 -14.3061 (harmonic 0.666667 -14.3061 occasionally, 0.666667 -14.3061 changes: 0.666667 -14.3061 ‘break’ 0.666667 -14.3061 lscript2-regularized 0.666667 -14.3061 udo 0.666667 -14.3061 semn∗ 0.666667 -14.3076 flight, 0.666667 -14.3076 grawltcq, 0.666667 -14.3076 48.14 0.666667 -14.3076 padding 0.666667 -14.3076 granted, 0.666667 -14.3076 hatwidey 0.666667 -14.3076 (tsarfaty, 0.666667 -14.3076 p@100 0.666667 -14.3076 corections 0.666667 -14.3076 jacobs, 0.666667 -14.3076 subproblems: 0.5 -14.3078 medtronic 0.0307314 -14.3082 whether 0.666667 -14.309 74.63 0.666667 -14.309 0.347 0.666667 -14.309 (entry 0.666667 -14.309 positive-labeled 0.666667 -14.309 interpolated, 0.666667 -14.309 713 0.666667 -14.309 colon, 0.666667 -14.309 〈i,j,k〉 0.666667 -14.309 0.586 0.666667 -14.309 fertility, 0.666667 -14.309 1234 0.666667 -14.309 constituency-to-string 0.666667 -14.309 unigram-level 0.666667 -14.309 1.62 0.333333 -14.3096 relative, 0.666667 -14.3104 undoubtedly, 0.666667 -14.3104 confinements 0.666667 -14.3104 “unidirectional 0.666667 -14.3104 39.64 0.666667 -14.3104 nlm 0.666667 -14.3104 polysynthetic 0.666667 -14.3104 lower-ranked 0.666667 -14.3104 26.8% 0.666667 -14.3104 72.6% 0.666667 -14.3104 precision/recall/f-score 0.666667 -14.3104 perturbs 0.666667 -14.3104 (44 0.666667 -14.3104 90.82% 0.666667 -14.3104 (iordanskaja 0.666667 -14.3104 1271 0.666667 -14.3104 capping 0.666667 -14.3104 balance: 0.666667 -14.3104 lm2 0.5 -14.311 grosz 0.5 -14.311 “eparl+nc” 0.130952 -14.3111 (klein 0.363636 -14.3113 personout 0.666667 -14.3118 “original” 0.666667 -14.3118 gnd, 0.666667 -14.3118 bination 0.666667 -14.3118 [event] 0.666667 -14.3118 freshman 0.666667 -14.3118 unary-reduce 0.666667 -14.3118 68.2% 0.666667 -14.3118 dependency-parsing 0.666667 -14.3118 1993); 0.666667 -14.3118 minimising 0.666667 -14.3118 orthogonalization 0.666667 -14.3118 subtopics 0.666667 -14.3118 [raux 0.666667 -14.3118 54.12 0.666667 -14.3118 0.451 0.666667 -14.3118 correspondency 0.666667 -14.3118 lpcm 0.666667 -14.3118 details) 0.428571 -14.3126 eugenio, 0.666667 -14.3133 1359 0.666667 -14.3133 rajkumar’s 0.666667 -14.3133 “ordinary” 0.666667 -14.3133 week), 0.666667 -14.3133 token-roles 0.666667 -14.3133 vos 0.666667 -14.3133 0.806 0.666667 -14.3133 œ (se) 0.666667 -14.3133 nu-lex, 0.666667 -14.3133 standard: 0.666667 -14.3133 echoed 0.666667 -14.3133 34.42 0.666667 -14.3133 situations: 0.666667 -14.3133 17.10 0.666667 -14.3133 +hj) 0.5 -14.3143 ctb2tp 0.5 -14.3143 69.23 0.666667 -14.3147 frag 0.666667 -14.3147 +biaslm 0.666667 -14.3147 blno−rl 0.666667 -14.3147 preliminarily 0.666667 -14.3147 http://nlp.stanford.edu/software/lex-parser.shtml 0.666667 -14.3147 (newman, 0.666667 -14.3147 autoand 0.666667 -14.3147 0.255 0.666667 -14.3147 fraud 0.666667 -14.3147 implicitely 0.666667 -14.3147 reach, 0.666667 -14.3147 online-behavior, 0.666667 -14.3147 21.84 0.666667 -14.3147 jointn-gram 0.666667 -14.3147 34.19 0.666667 -14.3147 domain-transfer 0.666667 -14.3147 “dell 0.666667 -14.3147 tweet-pair 0.24 -14.3151 receives a 0.0815451 -14.3153 aims 0.428571 -14.3154 “my 0.263158 -14.3154 patrick 0.0516351 -14.3156 2006; 0.0431034 -14.316 summationdisplay 0.666667 -14.3161 za, 0.666667 -14.3161 in  0.666667 -14.3161 1226 0.666667 -14.3161 mcenery, 0.666667 -14.3161 14.0% 0.666667 -14.3161 openness, 0.666667 -14.3161 positive-definite 0.666667 -14.3161 si: 0.666667 -14.3161 generation-based 0.666667 -14.3161 709 0.666667 -14.3161 on-line, 0.666667 -14.3161 4.0% 0.0528233 -14.3164 word. 0.263158 -14.3165 users were 0.666667 -14.3176 (show 0.666667 -14.3176 supervised) 0.666667 -14.3176 1147 0.666667 -14.3176 (pdtb, 0.666667 -14.3176 parenleftbign ·j 0.666667 -14.3176 silva 0.666667 -14.3176 940 0.666667 -14.3176 category-specific 0.666667 -14.3176 the〈t,t〉weights 0.666667 -14.3176 noticeable, 0.666667 -14.3176 140k 0.666667 -14.3176 phonetics, 0.666667 -14.3176 demoting 0.666667 -14.3176 epithets 0.666667 -14.3176 34.47 0.666667 -14.3176 cancels 0.666667 -14.3176 1,140 0.666667 -14.3176 politician, 0.666667 -14.3176 verbosity 0.666667 -14.3176 thrax, 0.666667 -14.3176 transitivity, 0.666667 -14.3176 s-correct 0.666667 -14.3176 integrated, 0.666667 -14.3176 optnull1 0.666667 -14.3176 cognizant 0.130952 -14.3181 noting 0.109375 -14.3186 stands 0.263158 -14.3188 (blum and 0.666667 -14.319 (saur´ı 0.666667 -14.319 ratio: 0.666667 -14.319 cytoplasmic 0.666667 -14.319 s+ 0.666667 -14.319 0.602 0.666667 -14.319 i> 0.666667 -14.319 act-r 0.666667 -14.319 78.33 0.666667 -14.319 person”, 0.666667 -14.319 ordering: 0.666667 -14.319 51.26 0.666667 -14.319 sw, 0.666667 -14.319 oov-handle 0.666667 -14.319 facilities, 0.666667 -14.319 smoking, 0.666667 -14.319 advcl, 0.666667 -14.319 spss 0.666667 -14.319 neutralizes 0.666667 -14.319 ebay, 0.666667 -14.319 pp-node 0.666667 -14.319 -0.23 0.428571 -14.3191 13.5% 0.12 -14.3193 have different 0.5 -14.3196 1,1 0.5 -14.3196 arc-standard, 0.263158 -14.3199 brandeis 0.666667 -14.3204 end-position 0.666667 -14.3204 forest-string 0.666667 -14.3204 listing, 0.666667 -14.3204 renewal 0.666667 -14.3204 49.50 0.666667 -14.3204 sie, 0.666667 -14.3204 pi-chuan 0.666667 -14.3204 layla 0.666667 -14.3204 (approximated 0.666667 -14.3204 (ppi) 0.666667 -14.3204 populous 0.666667 -14.3204 near-class 0.666667 -14.3204 string) 0.666667 -14.3204 1(c) 0.666667 -14.3204 mark-execute 0.666667 -14.3204 p), 0.666667 -14.3204 semiautomatic 0.666667 -14.3204 sb(w) 0.666667 -14.3204 chiarcos 0.666667 -14.3204 3.14 0.666667 -14.3204 metonymies, 0.5 -14.3207 psd 0.428571 -14.3209 pine 0.194444 -14.3218 questions that 0.666667 -14.3219 cpl, 0.666667 -14.3219 8.3% 0.666667 -14.3219 p6 0.666667 -14.3219 (showing 0.666667 -14.3219 dire 0.666667 -14.3219 ipu-internal 0.666667 -14.3219 kullback–leibler 0.666667 -14.3219 queryfeature 0.666667 -14.3219 dumps 0.666667 -14.3219 regain 0.666667 -14.3219 (voutilainen, 0.666667 -14.3219 208,832 0.666667 -14.3219 0.207 0.666667 -14.3219 known: 0.666667 -14.3219 49.21 0.666667 -14.3219 (elhadad 0.666667 -14.3233 blocksin 0.666667 -14.3233 shall, 0.666667 -14.3233 “bag-of-words” 0.666667 -14.3233 precλ 0.666667 -14.3233 gerrish 0.666667 -14.3233 butnariu 0.666667 -14.3233 computation: 0.666667 -14.3233 venkataraman’s 0.666667 -14.3233 “active” 0.666667 -14.3233 snorm(wl) 0.666667 -14.3233 bcpos 0.666667 -14.3233 survived 0.666667 -14.3233 c(t) 0.666667 -14.3233 (better 0.666667 -14.3233 mira, 0.666667 -14.3233 87.60 0.666667 -14.3233 (m1) 0.666667 -14.3233 92k 0.666667 -14.3233 verbocean, 0.666667 -14.3233 suf 0.666667 -14.3233 63.97 0.666667 -14.3233 cq(q) 0.666667 -14.3233 30.77 0.666667 -14.3233 41.77 0.666667 -14.3233 needed; 0.666667 -14.3233 (≤ 0.263158 -14.3233 this stage, 0.428571 -14.3237 several advantages 0.0821918 -14.324 into two 0.5 -14.324 s(i) 0.116071 -14.3244 table 2, 0.333333 -14.3247 y∈y(x) 0.666667 -14.3248 metrica 0.666667 -14.3248 20.0% 0.666667 -14.3248 aj−1 0.666667 -14.3248 ourselves, 0.666667 -14.3248 38.20 0.666667 -14.3248 w·x−b 0.666667 -14.3248 herself 0.666667 -14.3248 me-rec 0.666667 -14.3248 93.98 0.666667 -14.3248 paradoxically, 0.666667 -14.3248 0.493 0.666667 -14.3248 billing 0.666667 -14.3248 deepqa 0.666667 -14.3248 bodies, 0.666667 -14.3248 (20k) 0.666667 -14.3248 elwell 0.294118 -14.3256 glance, 0.666667 -14.3262 16.88 0.666667 -14.3262 q.text 0.666667 -14.3262 sfactors 0.666667 -14.3262 south city’s 0.666667 -14.3262 1762 0.666667 -14.3262 ireland-based 0.666667 -14.3262 dewac 0.666667 -14.3262 trigraphs 0.666667 -14.3262 feats 0.666667 -14.3262 0.301 0.666667 -14.3262 web” 0.666667 -14.3262 30.19 0.666667 -14.3262 (done 0.666667 -14.3262 1+answer-answer 0.666667 -14.3262 log(1/f) 0.263158 -14.3262 structure> 0.428571 -14.3265 pi∗ 0.333333 -14.3269 hirst 0.5 -14.3272 l(θ) 0.5 -14.3272 preterminalized 0.294118 -14.3275 *** 0.666667 -14.3277 unrehearsed 0.666667 -14.3277 ofs 0.666667 -14.3277 “following” 0.666667 -14.3277 predisambiguation 0.666667 -14.3277 plied 0.666667 -14.3277 clearparser 0.666667 -14.3277 kjv 0.666667 -14.3277 16k 0.666667 -14.3277 bold-font 0.666667 -14.3277 sigdial, 0.666667 -14.3277 78.30 0.666667 -14.3277 32.18 0.666667 -14.3277 018 waltham, 0.666667 -14.3277 muc-7 0.666667 -14.3277 42.70 0.666667 -14.3277 needed: 0.666667 -14.3277 analogies 0.428571 -14.3284 wk+1 0.428571 -14.3284 watkinson 0.169811 -14.3288 not statistically 0.666667 -14.3291 6.9m 0.666667 -14.3291 (development 0.666667 -14.3291 whi 0.666667 -14.3291 1014 0.666667 -14.3291 87.52 0.666667 -14.3291 676 0.666667 -14.3291 retention 0.666667 -14.3291 cheerful 0.666667 -14.3291 removal) 0.333333 -14.3291 target) 0.230769 -14.3303 be rewritten 0.5 -14.3305 i;j 0.5 -14.3305 allocation, 0.666667 -14.3306 a10a9 0.666667 -14.3306 1019 0.666667 -14.3306 worldw 0.666667 -14.3306 components; 0.666667 -14.3306 -*-, 0.666667 -14.3306 markmwe 0.666667 -14.3306 “tree” 0.666667 -14.3306 (causal) 0.666667 -14.3306 miltsakaki, 0.666667 -14.3306 sub-systems 0.666667 -14.3306 geo: nyc 0.666667 -14.3306 pml 0.666667 -14.3306 expression-level 0.060241 -14.3312 size of the 0.5 -14.3316 (kumaran 0.5 -14.3316 solo 0.5 -14.3316 0::b 0.666667 -14.332 fastr 0.666667 -14.332 207), 0.666667 -14.332  0.666667 -14.332 pouliquen, 0.666667 -14.332 shared-tasks 0.666667 -14.332 non-deverbal 0.666667 -14.332 fader, 0.666667 -14.332 slowdown 0.666667 -14.332 disambiguator 0.666667 -14.332 starred 0.666667 -14.332 fidaplus 0.666667 -14.332 paraphrase-enriched 0.666667 -14.332 tn) 0.666667 -14.332 conll’05 0.666667 -14.332 combinedpattern 0.333333 -14.332 clc-hdp-lg 0.5 -14.3327 (slu) 0.333333 -14.3334 relative) 0.333333 -14.3334 (cl) 0.666667 -14.3335 chebi 0.666667 -14.3335 87.42 0.666667 -14.3335 5.4% 0.666667 -14.3335 headmodifier 0.666667 -14.3335 org.adm 0.666667 -14.3335 +” 0.666667 -14.3335 assumeq(x) 0.666667 -14.3335 (quinlan 0.666667 -14.3335 establish-credibility 0.666667 -14.3335 (unordered) 0.666667 -14.3335 jokes 0.666667 -14.3335 5k+100k 0.666667 -14.3335 “recep 0.666667 -14.3335 misplaced 0.666667 -14.3335 x(c) 0.666667 -14.3335 15.9% 0.666667 -14.3335 folows 0.666667 -14.3335 (golding 0.5 -14.3338 β(v) 0.5 -14.3338 singh, 0.186047 -14.3341 first, it 0.294118 -14.3346 a71a71a71 0.212121 -14.3348 been considered 0.428571 -14.3349 º 0.5 -14.3349   0.666667 -14.335 one-point 0.666667 -14.335 2.5.1 0.666667 -14.335 date, most 0.666667 -14.335 29.15 0.666667 -14.335 id) 0.666667 -14.335 32.44 0.666667 -14.335 76.28 0.666667 -14.335 (hirschman 0.666667 -14.335 (1988), 0.666667 -14.335 :explicit 0.666667 -14.335 1088 0.666667 -14.335 “sure” 0.666667 -14.335 workings 0.666667 -14.335 6.3.1 0.666667 -14.335 ‘there 0.666667 -14.335 btg-based 0.666667 -14.335 47.41 0.666667 -14.335 +γ 0.666667 -14.335 statmt 0.666667 -14.335 (2009a), 0.666667 -14.335 ib1 0.666667 -14.335 (hearst 0.666667 -14.335 biased, 0.666667 -14.335 α=0.4 0.666667 -14.3364 23.00 0.666667 -14.3364 40.30 0.666667 -14.3364 non-empty, 0.666667 -14.3364 x→y 0.666667 -14.3364 onr, 0.666667 -14.3364 “alternative” 0.666667 -14.3364 29.10 0.666667 -14.3364 1142 0.666667 -14.3364 与 0.666667 -14.3364 1972; 0.666667 -14.3364 multinomial-dirichlet 0.666667 -14.3364 27.40 0.666667 -14.3364 glow, 0.666667 -14.3364 weighting) 0.666667 -14.3364 ‘de’ 0.666667 -14.3364 seriously, 0.666667 -14.3364 overlap: 0.666667 -14.3364 quote-based 0.666667 -14.3364 darker 0.117647 -14.337 experiments were 0.333333 -14.3378 (foster 0.666667 -14.3379 create, 0.666667 -14.3379 85.59 0.666667 -14.3379 (larger 0.666667 -14.3379 §4.1 0.666667 -14.3379 ’not 0.666667 -14.3379 english-slovene 0.666667 -14.3379 co-svm 0.666667 -14.3379 d(w) 0.666667 -14.3379 section2 0.666667 -14.3379 acknowledging 0.666667 -14.3379 amini 0.666667 -14.3379 phonetic, 0.666667 -14.3379 nation’s 0.666667 -14.3379 bner 0.0135432 -14.338 table 0.5 -14.3382 ella 0.307692 -14.3383 also asked 0.428571 -14.3387 rm-a 0.149254 -14.3392 determine which 0.666667 -14.3393 jubilee 0.666667 -14.3393 unk-fst 0.666667 -14.3393 signalling 0.666667 -14.3393 10.05 0.666667 -14.3393 1,827 0.666667 -14.3393 743 0.666667 -14.3393 context-rich 0.666667 -14.3393 saarbruecken, 0.666667 -14.3393 (magdy 0.666667 -14.3393 broadest 0.666667 -14.3393 inheriting 0.666667 -14.3393 11.6% 0.666667 -14.3393 7.2% 0.666667 -14.3393 180k 0.666667 -14.3393 pythia 0.666667 -14.3393 knearest 0.666667 -14.3393 notations: 0.666667 -14.3393 56.57 0.333333 -14.3393 (las) 0.5 -14.3393  0.5 -14.3393 tagging: 0.186047 -14.3395 al., 2000). 0.0993377 -14.34 this allows 0.5 -14.3404 treebank data. 0.666667 -14.3408 count[a,b] 0.666667 -14.3408 (published 0.666667 -14.3408 91.47 0.666667 -14.3408 ψ(t) 0.666667 -14.3408 (empty 0.666667 -14.3408 1241 0.666667 -14.3408 category; 0.666667 -14.3408 i-hmm, 0.666667 -14.3408 question-type 0.666667 -14.3408 tpl+-formula 0.666667 -14.3408 pronoun; 0.666667 -14.3408 '(*-) 0.666667 -14.3408 93.60 0.666667 -14.3408 0.0143 0.666667 -14.3408 (fleiss, 1971) 0.666667 -14.3408 minu 0.428571 -14.3415 task1 0.428571 -14.3415 segmentation) 0.5 -14.3415 celex 0.194444 -14.3422 arithmetic 0.666667 -14.3423 moroccan 0.666667 -14.3423 30.3% 0.666667 -14.3423 françois 0.666667 -14.3423 drc 0.666667 -14.3423 langley, 0.666667 -14.3423 gabrilovich, 0.666667 -14.3423 sequence) 0.666667 -14.3423 lexicon-seed) 0.666667 -14.3423 -0.3 0.666667 -14.3423 85.51 0.666667 -14.3423 underperform 0.666667 -14.3423 programming; 0.666667 -14.3423 here’s 0.666667 -14.3423 karen 0.170213 -14.3428 of emnlp 0.428571 -14.3434 (3b) 0.5 -14.3437 thousands. 0.5 -14.3437 cathedral 0.5 -14.3437 braceleftbig 0.5 -14.3437 24.09 0.307692 -14.3438 sagae, 0.666667 -14.3438 uio 0.666667 -14.3438 entropic 0.666667 -14.3438 formance 0.666667 -14.3438 automata, 0.666667 -14.3438 (esol) 0.666667 -14.3438 widening 0.666667 -14.3438 selfand 0.135802 -14.3442 other two 0.222222 -14.3442 fillers. 0.5 -14.3448 hanneman 0.666667 -14.3452 acurately 0.666667 -14.3452 f+m 0.666667 -14.3452 515k 0.666667 -14.3452 alert 0.666667 -14.3452 “neither 0.666667 -14.3452 tern 0.666667 -14.3452 21.85 0.666667 -14.3452 1.51 0.666667 -14.3452 s(np) 0.428571 -14.3453 tokens, but 0.307692 -14.3458 scratch. 0.0425273 -14.3458 results in 0.5 -14.3459 (epsilon1 0.5 -14.3459 631 0.0955414 -14.3463 group of 0.666667 -14.3467 ‘c’ 0.666667 -14.3467 ˆθk 0.666667 -14.3467 proceed, 0.666667 -14.3467 syllable-based 0.666667 -14.3467 (sleator 0.666667 -14.3467 rate2 0.666667 -14.3467 “what?” 0.666667 -14.3467 publishers, 0.666667 -14.3467 nc,e(z) 0.666667 -14.3467 interviewers 0.666667 -14.3467 homosexuals 0.666667 -14.3467 wrd 0.666667 -14.3467 multilingual, 0.666667 -14.3467 0.101 0.5 -14.347 −α 0.5 -14.347 procrustes 0.277778 -14.3481 ++ 0.5 -14.3481 shit 0.5 -14.3481 s32s100s80s85s80s65 0.5 -14.3481 (10-fold 0.5 -14.3481 funding) 0.5 -14.3481 cl. 0.5 -14.3481 holling 0.5 -14.3481 lasdiff 0.5 -14.3481 0.19, 0.5 -14.3481 nai-lung 0.5 -14.3481 tsao 0.5 -14.3481 phrased-based 0.5 -14.3481 siamese 0.5 -14.3481 publication: 0.5 -14.3481 dots) 0.5 -14.3481 affect-enriched 0.5 -14.3481 2to 0.5 -14.3481 [%] 0.5 -14.3481 transporting 0.5 -14.3481 bkg 0.5 -14.3481 links: 0.666667 -14.3482 f11 0.666667 -14.3482 balık 0.666667 -14.3482 (global 0.666667 -14.3482 tanaka-ishii 0.666667 -14.3482 builders 0.666667 -14.3482 30.50 0.666667 -14.3482 drdas, 0.307692 -14.3485 4.3, 0.114035 -14.3489 consistent with the 0.375 -14.349 digital library 0.375 -14.349 kurohashi, 0.307692 -14.3492 alphabet. 0.307692 -14.3492 wordnet similarity 0.5 -14.3492 token-seg 0.5 -14.3492 critical, 0.5 -14.3492 toppathsd(s) 0.5 -14.3492 pseudodocuments 0.5 -14.3492 86.82 0.5 -14.3492 performed: 0.5 -14.3492 a59 0.5 -14.3492 65k 0.5 -14.3492 1985 0.5 -14.3492 24.01 0.5 -14.3492 tuesday 0.5 -14.3492 r/d/c 0.5 -14.3492 haschild 0.5 -14.3492 deducing 0.5 -14.3492 7for 0.363636 -14.3494 promising, 0.363636 -14.3494 arr 0.214286 -14.3495 file. 0.666667 -14.3497 structureof 0.666667 -14.3497 0.103 0.666667 -14.3497 “windows 0.666667 -14.3497 utts 0.666667 -14.3497 24.37 0.666667 -14.3497 ewe 0.666667 -14.3497 (strassel 0.666667 -14.3497 5.4, 0.666667 -14.3497 differences; 0.666667 -14.3497 0.513 0.666667 -14.3497 96.21 0.666667 -14.3497 ambient 0.666667 -14.3497 l1-paraphrases 0.666667 -14.3497 non-co¨occurring 0.666667 -14.3497 569 0.666667 -14.3497 deposited 0.666667 -14.3497 40.16 0.666667 -14.3497 schubert 0.666667 -14.3497 1281 0.666667 -14.3497 sánchez, 0.666667 -14.3497 dir, 0.666667 -14.3497 1090 0.666667 -14.3497 snomed, 0.0143587 -14.3498 than 0.5 -14.3503 (bl), 0.5 -14.3503 4.50 0.5 -14.3503 acoustic, 0.5 -14.3503 730 0.5 -14.3503 2( 0.5 -14.3503 733 0.5 -14.3503 mande 0.333333 -14.3511 (carroll 0.666667 -14.3512 63.60 0.666667 -14.3512 training/development/testing 0.666667 -14.3512 reg-total 0.666667 -14.3512 alt 0.666667 -14.3512 34.65 0.666667 -14.3512 (monotone) 0.666667 -14.3512 inapplicability 0.666667 -14.3512 87.40 0.666667 -14.3512 geo:area(y 0.666667 -14.3512 invariable 0.666667 -14.3512 abu 0.666667 -14.3512 1097 0.666667 -14.3512 (yielding 0.5 -14.3515 fcc 0.5 -14.3515 n-gram–based 0.5 -14.3515 10/1 0.363636 -14.3518 (2011) for 0.363636 -14.3518 cept 0.428571 -14.352 tot 0.375 -14.3523 f(a) 0.5 -14.3526 5.7% 0.5 -14.3526 entity clusters. 0.5 -14.3526 1284 0.5 -14.3526 asumption 0.5 -14.3526 oft 0.5 -14.3526 pattern,” 0.5 -14.3526 baum-welch 0.5 -14.3526 input-unfaithful 0.666667 -14.3527 31.8% 0.666667 -14.3527 framenet-based 0.666667 -14.3527 ´a 0.666667 -14.3527 omissions, 0.666667 -14.3527 repercussions 0.666667 -14.3527 consensus, 0.666667 -14.3527 32.5% 0.666667 -14.3527 t∈t, 0.666667 -14.3527 sho 0.666667 -14.3527 (abbreviated 0.666667 -14.3527 ncbo 0.666667 -14.3527 93.68 0.375 -14.3531 weights: 0.230769 -14.3533  0.5 -14.3537 rx 0.5 -14.3537 v-ing”) 0.5 -14.3537 direction), 0.5 -14.3537 medical, 0.666667 -14.3541 25.76 0.666667 -14.3541 1more 0.666667 -14.3541 power: 0.666667 -14.3541 37.22 0.666667 -14.3541 n−k 0.666667 -14.3541 luís 0.666667 -14.3541 jungen menschen, 0.666667 -14.3541 morphologicallyrich 0.666667 -14.3541 eclipse 0.666667 -14.3541 axiomatized 0.666667 -14.3541 sentinel 0.263158 -14.3546 that returns 0.5 -14.3548 playback 0.5 -14.3548 paraphr≥100 0.5 -14.3548 indifference 0.5 -14.3548 chapter, 0.5 -14.3548 “similar 0.5 -14.3548 interface: 0.5 -14.3548 21.10 0.666667 -14.3556 ‘but 0.666667 -14.3556 “out 0.666667 -14.3556 433 0.666667 -14.3556 austrian 0.666667 -14.3556 supreme 0.666667 -14.3556 mahwah, 0.666667 -14.3556 4.26 0.666667 -14.3556 22.96 0.666667 -14.3556 http://www.nist.gov/speech/tests/ace/ 0.666667 -14.3556 inter-page 0.666667 -14.3556 (distributional 0.333333 -14.3556 sie:you 0.333333 -14.3556 = {q 0.5 -14.3559 improperly 0.5 -14.3559 (bp), 0.5 -14.3559 dying 0.5 -14.3559 (2010c) 0.5 -14.3559 12.0% 0.0535714 -14.3561 is defined 0.307692 -14.3568 berkeley, 0.428571 -14.3568 “machine 0.5 -14.357 pulse 0.5 -14.357 21.06 0.5 -14.357 k-fold 0.5 -14.357 queries; 0.5 -14.357 93.69 0.666667 -14.3571 90.23% 0.666667 -14.3571 underlined) 0.666667 -14.3571 senticorpus-pt 0.666667 -14.3571 mixture+em 0.666667 -14.3571 coupons 0.666667 -14.3571 casablanca 0.666667 -14.3571 88.18 0.666667 -14.3571 cofer: 0.666667 -14.3571 “if”, 0.666667 -14.3571 marking, 0.666667 -14.3571 conv, 0.666667 -14.3571 984 0.666667 -14.3571 21.22 0.666667 -14.3571 andrei popescu-belis 0.666667 -14.3571 p11 0.666667 -14.3571 f(·) 0.277778 -14.3574 seattle, 0.428571 -14.3577 rtk 0.428571 -14.3577 ilavach 0.222222 -14.3581 asociated 0.307692 -14.3582 table 2: number 0.5 -14.3582 8.6% 0.5 -14.3582 hits-based 0.5 -14.3582 46.39 0.5 -14.3582 excell 0.5 -14.3582 αi, 0.5 -14.3582 informations 0.5 -14.3582 countably 0.5 -14.3582 1.95 0.666667 -14.3586 6if 0.666667 -14.3586 79.17 0.666667 -14.3586 licence 0.666667 -14.3586 elisabeth 0.666667 -14.3586 xinyan 0.666667 -14.3586 (1c), 0.666667 -14.3586 70.28 0.666667 -14.3586 alle 0.666667 -14.3586 revisionapi 0.666667 -14.3586 “w1 0.666667 -14.3586 (100), 0.666667 -14.3586 judith 0.666667 -14.3586 cut-based 0.173913 -14.3588 approach outperforms 0.375 -14.359 ξi 0.5 -14.3593 snull 0.5 -14.3593 (“i 0.5 -14.3593 genderspecific 0.666667 -14.3601 nano 0.666667 -14.3601 “change 0.666667 -14.3601 laser 0.666667 -14.3601 (hamon 0.666667 -14.3601 concerns: 0.666667 -14.3601 “intermediate” 0.666667 -14.3601 0.217 0.666667 -14.3601 hl 0.666667 -14.3601 makhoul, 0.666667 -14.3601 24.81 0.666667 -14.3601 48.31 0.666667 -14.3601 +0.46 0.666667 -14.3601 non-clicks 0.666667 -14.3601 the] 0.307692 -14.3602 updated. 0.5 -14.3604 per-instance 0.5 -14.3604 z(h) 0.5 -14.3604 binarised 0.5 -14.3604 port-media 0.294118 -14.3607 analyzer. 0.5 -14.3615 bad) 0.5 -14.3615 sentenceor 0.5 -14.3615 ning 0.5 -14.3615 br-comp 0.5 -14.3615 manually-annotated 0.666667 -14.3616 (100) 0.666667 -14.3616 scenario: 0.666667 -14.3616 23.23 0.666667 -14.3616 (annotated) 0.666667 -14.3616 0.821 0.666667 -14.3616 (block-v3’ 0.666667 -14.3616 rw 0.666667 -14.3616 lim, 0.666667 -14.3616 90.36 0.666667 -14.3616 skolem 0.666667 -14.3616 wprime∈v 0.666667 -14.3616 sall 0.666667 -14.3616 timing: 0.666667 -14.3616 by” 0.428571 -14.3616 schemas, 0.428571 -14.3616 omar 0.363636 -14.3624 en-cs 0.428571 -14.3626 29.90 0.5 -14.3627 minimizations 0.5 -14.3627 ={(i,j) 0.5 -14.3627 intentionality 0.5 -14.3627 apears 0.5 -14.3627 query-independent 0.5 -14.3627 3.03 0.5 -14.3627 nce 0.666667 -14.3631 deletion: 0.666667 -14.3631 84.79 0.666667 -14.3631 31.09 0.666667 -14.3631 1-2% 0.666667 -14.3631 connectingthe 0.666667 -14.3631 alignmenta, 0.666667 -14.3631 -2, 0.666667 -14.3631 70.59 0.666667 -14.3631 marble 0.666667 -14.3631 7.34 0.666667 -14.3631 “false” 0.666667 -14.3631 20-newsgroups 0.666667 -14.3631 postgraduate 0.136986 -14.3635 large corpus 0.428571 -14.3635 dipanjan 0.5 -14.3638 tests: 0.5 -14.3638 keyphrase-based 0.5 -14.3638 springt 0.5 -14.3638 “add 0.5 -14.3638 windowed 0.5 -14.3638 a.2 0.5 -14.3638 percentages) 0.5 -14.3638 registers 0.5 -14.3638 prioritization 0.307692 -14.3644 minnesota 0.666667 -14.3646 10; 0.666667 -14.3646 students), 0.666667 -14.3646 nomination 0.666667 -14.3646 si) 0.666667 -14.3646 gegeben 0.666667 -14.3646 powell’s 0.666667 -14.3646 (bolinger, 0.666667 -14.3646 30.36 0.666667 -14.3646 0.5 -14.3751 pre-terminals 0.5 -14.3751 (non-empty) 0.666667 -14.3752 0.965 0.666667 -14.3752 ravichandran’s 0.666667 -14.3752 (rb) 0.666667 -14.3752 ofc1 0.666667 -14.3752 wmt10e 0.666667 -14.3752 933 0.666667 -14.3752 biomolecules 0.666667 -14.3752 constituent; 0.666667 -14.3752 93.15 0.666667 -14.3752 496 0.666667 -14.3752 42.29 0.666667 -14.3752 l−1 0.666667 -14.3752 (inter-annotator 0.333333 -14.3752 publically 0.181818 -14.376 crucial to 0.25 -14.3763 intractable. 0.5 -14.3763 horrible 0.5 -14.3763 expandable 0.5 -14.3763 subsegment 0.666667 -14.3767 explain, 0.666667 -14.3767 copper 0.666667 -14.3767 852 0.666667 -14.3767 heaps, 0.666667 -14.3767 0.065 0.666667 -14.3767 (herrera 0.666667 -14.3767 1/k 0.666667 -14.3767 99, 0.666667 -14.3767 19.10 0.428571 -14.3772 officials 0.363636 -14.3773 2.4. 0.5 -14.3774 mercer 0.5 -14.3774 token-pos 0.375 -14.3777 … ∪ 0.666667 -14.3783 ‘y 0.666667 -14.3783 illness 0.666667 -14.3783 over) 0.666667 -14.3783 λs 0.666667 -14.3783 istv´an 0.666667 -14.3783 932 0.666667 -14.3783 14.71 0.666667 -14.3783 (hence, 0.666667 -14.3783 73.06 0.5 -14.3786 889 0.5 -14.3786 thoracic viscus> 0.5 -14.3786 ={ 0.5 -14.3786 arrows) 0.5 -14.3786 documentary 0.163265 -14.3789 zettlemoyer 0.363636 -14.3789 english/arabic 0.428571 -14.3791 11.85 0.5 -14.3797 sensitive, 0.5 -14.3797 excellence, 0.5 -14.3797 6.1) 0.5 -14.3797 upm 0.5 -14.3797 arrays, 0.5 -14.3797 exponentiated 0.5 -14.3797 2.1.4 0.666667 -14.3798 record’s 0.666667 -14.3798 parametersθ 0.666667 -14.3798 bn, 0.666667 -14.3798 models’, 0.666667 -14.3798 68.7% 0.666667 -14.3798 noncollaborative 0.666667 -14.3798 assume, 0.666667 -14.3798 74.39 0.25 -14.3809 (q 0.5 -14.3809 thursday 0.5 -14.3809 p2: 0.5 -14.3809 19.7 0.5 -14.3809 xbleu 0.5 -14.3809 amends 0.5 -14.3809 +0.33 0.5 -14.3809 41.42 0.5 -14.3809 3.4.2 0.5 -14.3809 21.77 0.375 -14.3812 bell, 0.666667 -14.3813 fully-learnt, 0.666667 -14.3813 maria, 0.666667 -14.3813 84.31 0.666667 -14.3813 vv, 0.666667 -14.3813 quad 0.666667 -14.3813 noun-adjective 0.666667 -14.3813 1593 0.666667 -14.3813 pacts 0.666667 -14.3813 nursing 0.666667 -14.3813 nullwnullu 0.5 -14.382 translocation 0.5 -14.382 link-da 0.5 -14.382 1911 0.5 -14.382 728 0.0699301 -14.3823 experiments, we 0.666667 -14.3829 418 0.666667 -14.3829 0.268 0.666667 -14.3829 1.2x 0.666667 -14.3829 cloak 0.666667 -14.3829 crying 0.666667 -14.3829 dann 0.666667 -14.3829 ,1 0.666667 -14.3829 endowing 0.666667 -14.3829 1254 0.666667 -14.3829 digit, 0.666667 -14.3829 intransitive, 0.666667 -14.3829 recursively, 0.5 -14.3832 p(“x 0.5 -14.3832 and  0.0727969 -14.3833 description of 0.5 -14.3843 obviates 0.5 -14.3843 inconsistently 0.5 -14.3843 spur 0.5 -14.3843 w1w2w3∊ 0.5 -14.3843 ge’09 0.666667 -14.3844 formulation: 0.666667 -14.3844 undetermined 0.666667 -14.3844 (hayes 0.666667 -14.3844 prices, 0.666667 -14.3844 66.71 0.666667 -14.3844 andp2 0.666667 -14.3844 gere 0.666667 -14.3844 ωj 0.666667 -14.3844 9.71 0.666667 -14.3844 (della 0.666667 -14.3844 c(w1w2w3) 0.666667 -14.3844 klsh, 0.666667 -14.3844 41.39 0.375 -14.3846  0.157895 -14.3849 give an 0.375 -14.3855 fromthe 0.5 -14.3855 miriam 0.5 -14.3855 wsj; 0.5 -14.3855 protein interaction 0.5 -14.3855 morning, 0.5 -14.3855 word-lattice 0.666667 -14.3859 identification) 0.666667 -14.3859 2as 0.666667 -14.3859 front, 0.666667 -14.3859 villavicencio, 0.666667 -14.3859 biological proof-based 0.666667 -14.3859 1152 0.666667 -14.3859 proof, 0.666667 -14.3859 auto-feedback 0.666667 -14.3859 twa, 0.666667 -14.3859 88.38 0.428571 -14.386 kay, 0.263158 -14.3863 〉 0.375 -14.3864 ict 0.5 -14.3866 entity” 0.5 -14.3866 256, 0.666667 -14.3875 piano, 0.666667 -14.3875 g(β,at(β,λ)) 0.666667 -14.3875 webster’s 0.666667 -14.3875 whilst, 0.666667 -14.3875 99.58% 0.666667 -14.3875 7.92 0.666667 -14.3875 eating, 0.666667 -14.3875 effects: 0.666667 -14.3875 bp’s 0.177778 -14.3877 we review 0.5 -14.3878 percentageof 0.428571 -14.388 savg 0.375 -14.3881 teamfinalscore(g, 0.5 -14.3889 hier, 0.5 -14.3889 813 0.5 -14.3889 french), 0.666667 -14.389 (“xfer-orig”) 0.666667 -14.389 commission’s 0.666667 -14.389 1948; 0.666667 -14.389 duchampian 0.666667 -14.389 gd 0.666667 -14.389 29.37 0.666667 -14.389 cross-annotation 0.666667 -14.389 owned 0.666667 -14.389 summationtextkj=1 0.666667 -14.389 [n 0.0548523 -14.3894 believe 0.5 -14.3901 0.242 0.5 -14.3901 -0.50 0.5 -14.3901 act) 0.666667 -14.3906 41.32 0.666667 -14.3906 anaphors: 0.666667 -14.3906 middle-left 0.666667 -14.3906 mmr-md 0.666667 -14.3906 imap 0.666667 -14.3906 es(pi 0.666667 -14.3906 30.87 0.666667 -14.3906 junction 0.666667 -14.3906 penalty: 0.666667 -14.3906 firth, 0.666667 -14.3906 promoters 0.666667 -14.3906 hlds 0.666667 -14.3906 ready-to-go 0.666667 -14.3906 +0.90 0.666667 -14.3906 significantly: 0.666667 -14.3906 alonso 0.666667 -14.3906 hand; 0.666667 -14.3906 24.26 0.5 -14.3913 (202) 0.5 -14.3913 814 0.5 -14.3913 all+ 0.5 -14.3913 model,which 0.263158 -14.3918 r ∈ 0.428571 -14.392 (statistically 0.666667 -14.3921 [choi 0.666667 -14.3921 (provide 0.666667 -14.3921 unplugged’) 0.666667 -14.3921 palate 0.666667 -14.3921 systemof 0.666667 -14.3921 particular: 0.375 -14.3924 no overlap 0.5 -14.3924 34.55 0.5 -14.3924 (a3) 0.5 -14.3924 ecnu 0.5 -14.3936 68.5% 0.5 -14.3936 (a1, 0.666667 -14.3937 fairness, 0.666667 -14.3937 full-featured 0.666667 -14.3937 nervous 0.666667 -14.3937 nouns,the 0.666667 -14.3937 gimpel, 0.666667 -14.3937 babies, 0.666667 -14.3937 took, 0.666667 -14.3937 18.09 0.666667 -14.3937 challenging: 0.666667 -14.3937 nya, 0.666667 -14.3937 (whnp 0.666667 -14.3937 (2003): 0.666667 -14.3937 ecoc 0.666667 -14.3937 effects), 0.666667 -14.3937 sub-entity 0.666667 -14.3937 cummins 0.666667 -14.3937 handed 0.5 -14.3947 bionlp-st’11 0.5 -14.3947 oslo, 0.5 -14.3947 4.57 0.375 -14.395 j1 0.375 -14.395 ministre 0.666667 -14.3952 aha 0.666667 -14.3952 87.35 0.666667 -14.3952 lyons 0.666667 -14.3952 84.55 0.666667 -14.3952 said) 0.666667 -14.3952 schwarzschild 0.666667 -14.3952 snr, 0.666667 -14.3952 validly 0.666667 -14.3952 905 0.666667 -14.3952 adposition 0.666667 -14.3952 ρ= 0.666667 -14.3952 qald-1 0.666667 -14.3952 1.59 0.666667 -14.3952 edge) 0.666667 -14.3952 88.54 0.5 -14.3959 (combining 0.5 -14.3959 42.22 0.5 -14.3959 70.30 0.5 -14.3959 approximation: 0.666667 -14.3968 alien 0.666667 -14.3968 erage 0.666667 -14.3968 gaandhii’ 0.666667 -14.3968 predicates), 0.666667 -14.3968 independent: 0.666667 -14.3968 1243 0.666667 -14.3968 genre: 0.666667 -14.3968 22.50 0.666667 -14.3968 hair, 0.666667 -14.3968 −0.1 0.666667 -14.3968 ¯null 0.666667 -14.3968 core12, 0.666667 -14.3968 future improvements. 0.5 -14.3971 18.14 0.5 -14.3971 γ(w;(x,y)) 0.5 -14.3971 w(p) 0.5 -14.3971 0.043 0.5 -14.3971 vs(v) 0.5 -14.3971 try, 0.5 -14.3971 818 0.5 -14.3971 symbolize 0.375 -14.3977 21, 0.375 -14.3977 abbreviations: 0.5 -14.3982 44.41 0.5 -14.3982 non-compounds 0.5 -14.3982 extremely sensitive 0.5 -14.3982 covered-norm 0.666667 -14.3983 det+noun 0.666667 -14.3983 fjk 0.666667 -14.3983 adv) 0.666667 -14.3983 histones 0.666667 -14.3983 singulars 0.666667 -14.3983 cole, 0.666667 -14.3983 twoway 0.666667 -14.3983 analysis), 0.666667 -14.3983 mislead 0.666667 -14.3983 words.5 0.666667 -14.3983 ops/ng/n 0.666667 -14.3983 78.45 0.0540541 -14.3986 2005; 0.285714 -14.399 has fewer 0.307692 -14.3991 ngs-u 0.375 -14.3994 found, we 0.5 -14.3994 decomposed-ilp 0.5 -14.3994 16.85 0.333333 -14.3998 2010 workshop 0.666667 -14.3999 obama” 0.666667 -14.3999 koch (1977) 0.666667 -14.3999 przybocki, 0.666667 -14.3999 anotation, 0.666667 -14.3999 97.51 0.666667 -14.3999 1995): 0.666667 -14.3999 0.028 0.666667 -14.3999 2.99 0.333333 -14.4006 extractors. 0.5 -14.4006 [hay 0.5 -14.4006 [ronaldo]figure 0.5 -14.4006 generaly 0.5 -14.4006 suites, 0.123596 -14.4008 1 . 0.428571 -14.401 sigkdd 0.263158 -14.401 (jj 0.152542 -14.401 interface. 0.375 -14.4012 (open 0.666667 -14.4014 lp-mert, 0.666667 -14.4014 catalan, 0.666667 -14.4014 p/pprime 0.666667 -14.4014 l2-regularization 0.666667 -14.4014 mid-utterance 0.666667 -14.4014 domain-dependent, 0.666667 -14.4014 rodrigo 0.666667 -14.4014 altlex, 0.666667 -14.4014 summarizesthe 0.666667 -14.4014 prosody, 0.666667 -14.4014 forthe 0.5 -14.4017 s(a,b) 0.5 -14.4017 dean 0.5 -14.4017 overlapscore 0.5 -14.4017 50.18 0.104 -14.4021 introduced in 0.5 -14.4029 continent 0.5 -14.4029 2] 0.5 -14.4029 network; 0.5 -14.4029 glucose 0.666667 -14.403 vehicles, 0.666667 -14.403 68.59 0.666667 -14.403 huang2009 0.666667 -14.403 30.48 0.666667 -14.403 lehmann 0.666667 -14.403 (p/r)f1 0.666667 -14.403 airlines, 0.666667 -14.403 welser 0.666667 -14.403 38.8% 0.666667 -14.403 english-arabic: 0.666667 -14.403 (distance 0.666667 -14.403 zhifei 0.666667 -14.403 ecels 0.666667 -14.403 2.69 0.263158 -14.4035 and lavie 0.0842105 -14.4041 function, 0.5 -14.4041 sub-task 0.5 -14.4041 10.3% 0.5 -14.4041 fuzzy, 0.5 -14.4041 speaker‟s 0.666667 -14.4046 schmid, 0.666667 -14.4046 szeged 0.666667 -14.4046 mode) 0.666667 -14.4046 df, 0.666667 -14.4046 twosided 0.666667 -14.4046 62.00 0.666667 -14.4046 77.55 0.666667 -14.4046 mit csail 0.666667 -14.4046 (locally 0.285714 -14.405 qmp 0.5 -14.4053 mello 0.5 -14.4053 employed: 0.5 -14.4053 idea: 0.5 -14.4053 mothers 0.5 -14.4053 foremost 0.285714 -14.4057 robust. 0.666667 -14.4061 ‘syntax’ 0.666667 -14.4061 95.11 0.666667 -14.4061 table), 0.666667 -14.4061 0.759 0.666667 -14.4061 baas 0.666667 -14.4061 korea) 0.5 -14.4065 ∼bt 0.5 -14.4065 modification: 0.5 -14.4065 personalities 0.5 -14.4065 _িব 0.307692 -14.407 hw, 0.375 -14.4074 (sharifi 0.375 -14.4074 acoustically 0.5 -14.4076 edmund 0.5 -14.4076 to] 0.5 -14.4076 nn-prop srbya 0.5 -14.4076 bl-3 0.5 -14.4076 competitor 0.5 -14.4076 12.1% 0.5 -14.4076 (highlighted 0.666667 -14.4077 28.87 0.666667 -14.4077 16.46 0.666667 -14.4077 24.28 0.666667 -14.4077 landmark) 0.666667 -14.4077 wordsense 0.666667 -14.4077 wolf, 0.666667 -14.4077 (lr), 0.666667 -14.4077 57.87 0.666667 -14.4077 cau 0.666667 -14.4077 nn”, 0.666667 -14.4077 volcanoes 0.666667 -14.4077 g} 0.666667 -14.4077 cmu’s 0.214286 -14.408 characterization 0.111111 -14.4082 bionlp shared 0.375 -14.4082 53.33 0.5 -14.4088 iid 0.666667 -14.4093 josep 0.666667 -14.4093 56.70 0.666667 -14.4093 grad 0.666667 -14.4093 85.82 0.666667 -14.4093 “gorgeous” 0.666667 -14.4093 media: 0.666667 -14.4093 94.71 0.666667 -14.4093 (4.8) 0.666667 -14.4093 affär 0.666667 -14.4093 cpt 0.375 -14.41 damsl 0.5 -14.41 78.85 0.5 -14.41 scheme1 0.5 -14.41 (x), 0.5 -14.41 9this 0.5 -14.41 search-space 0.5 -14.41 icwsm 0.238095 -14.4104 .001 0.230769 -14.4105 bo 0.0572792 -14.4108 assign 0.666667 -14.4109 prot(x) 0.666667 -14.4109 linguistically-rich 0.666667 -14.4109 35.56 0.666667 -14.4109 kurt 0.666667 -14.4109 artist’s 0.428571 -14.4111 [i 0.428571 -14.4111 p<.01), 0.5 -14.4112 0.433 0.5 -14.4112 immune 0.5 -14.4112 ibm’s 0.156863 -14.4118 3.1. 0.375 -14.4118 θv,w 0.666667 -14.4124 2-by-2 0.666667 -14.4124 present) 0.666667 -14.4124 dialogue; 0.666667 -14.4124 55.33 0.666667 -14.4124 notice, 0.666667 -14.4124 99.5% 0.666667 -14.4124 unlabeled, 0.666667 -14.4124 54.07 0.666667 -14.4124 lengths: 0.666667 -14.4124 “good,” 0.666667 -14.4124 θb 0.666667 -14.4124 no: 0.666667 -14.4124 ϕ(b 0.666667 -14.4124 30.46 0.666667 -14.4124 17.04 0.666667 -14.4124 941 0.666667 -14.4124 lipid 0.5 -14.4124 platforms, 0.0472313 -14.4133 languages. 0.0825243 -14.4134 belong 0.5 -14.4135 elming 0.5 -14.4135 p¯c 0.375 -14.4136 tedeval 0.666667 -14.414 manaal 0.666667 -14.414 low-order 0.666667 -14.414 footprint, 0.666667 -14.414 78.03 0.666667 -14.414 deficiencies, 0.666667 -14.414 (dasgupta 0.666667 -14.414 emission probabilities, 0.5 -14.4147 memms 0.5 -14.4147 cypriot 0.5 -14.4147 eacl, 0.5 -14.4147 defeat 0.5 -14.4147 nuance 0.5 -14.4147 valdev 0.666667 -14.4156 54.64 0.666667 -14.4156 aj1 0.666667 -14.4156 1.27e-03 0.666667 -14.4156 0.775 0.666667 -14.4156 brigham 0.666667 -14.4156 tags’ 0.666667 -14.4156 segment(c) 0.666667 -14.4156 20.96 0.666667 -14.4156   0.666667 -14.4156 hearst’s 0.666667 -14.4156 93.63 0.285714 -14.4158 logs, 0.106557 -14.4158 this kind 0.5 -14.4159 renyi 0.5 -14.4159 entailment-based 0.375 -14.4162 badminton 0.0979021 -14.4167 only if 0.5 -14.4171 87k 0.5 -14.4171 spelling: 0.5 -14.4171 hallmarks 0.5 -14.4171 tear 0.666667 -14.4172 1,788 0.666667 -14.4172 30.58 0.666667 -14.4172 51.92 0.666667 -14.4172 3.43 0.666667 -14.4172 environment: 0.666667 -14.4172 −0.11 0.666667 -14.4172 user; 0.666667 -14.4172 88.49 0.666667 -14.4172 33.13 0.428571 -14.4172 laurent 0.0740741 -14.4179 experimented 0.375 -14.418 z∗(j)q 0.5 -14.4183 78.80 0.5 -14.4183 32.94 0.5 -14.4183 edible 0.5 -14.4183 head; 0.166667 -14.4188 marneffe 0.666667 -14.4188 larry 0.666667 -14.4188 detection) 0.666667 -14.4188 1117 0.666667 -14.4188 ts(f 0.666667 -14.4188 admittedly, 0.666667 -14.4188 1336 0.666667 -14.4188 endings, 0.666667 -14.4188 eruption, 0.110092 -14.419 (kim et 0.5 -14.4195 jac: 0.666667 -14.4204 (turn) 0.666667 -14.4204 scheible 0.666667 -14.4204 german-toenglish 0.666667 -14.4204 90.71 0.666667 -14.4204 “grammy 0.666667 -14.4204 reuters-10 0.666667 -14.4204 “voice 0.666667 -14.4204 table3 0.666667 -14.4204 +ci−11 0.666667 -14.4204 (city) 0.666667 -14.4204 scheme: 0.666667 -14.4204 revolves 0.666667 -14.4204 51.17 0.5 -14.4207 compute: 0.5 -14.4207 44.43 0.5 -14.4207 sequence), 0.5 -14.4207 ˆpunif 0.277778 -14.4211 phrases were 0.214286 -14.4217 concluded 0.5 -14.4219 nmd 0.5 -14.4219 low-rank 0.5 -14.4219 dency 0.5 -14.4219 87.43 0.666667 -14.422 (relation 0.666667 -14.422 storm 0.666667 -14.422 (ekman, 0.666667 -14.422 (tsp) 0.666667 -14.422 di,j 0.666667 -14.422 ac: 0.666667 -14.422 syn-st 0.666667 -14.422 accuracies: 0.666667 -14.422 snubbed 0.666667 -14.422 extraordinary 0.0675676 -14.422 we present a 0.0576923 -14.4224 context. 0.142857 -14.4226 summing 0.5 -14.4231 biographies 0.5 -14.4231 wearing 0.5 -14.4231 support-verb 0.5 -14.4231 chunk–pair–distance 0.666667 -14.4236 p@5, 0.666667 -14.4236 animations 0.666667 -14.4236 (two-tailed 0.666667 -14.4236 21.94 0.666667 -14.4236 qe, 0.666667 -14.4236 edited, 0.666667 -14.4236 shame 0.666667 -14.4236 (γ)] 0.666667 -14.4236 599 0.375 -14.4243 (n ag 0.5 -14.4243 30m 0.5 -14.4243 topic/sense 0.5 -14.4243 dowty’s 0.5 -14.4243 pleonastic 0.5 -14.4243 knowitall 0.5 -14.4243 grafting, 0.5 -14.4243 qij 0.5 -14.4243 617 0.104 -14.4248 are obtained 0.333333 -14.4251 triangleright 0.666667 -14.4252 1004 0.666667 -14.4252 synonym-based 0.666667 -14.4252 schatz 0.666667 -14.4252 (constituting 0.666667 -14.4252 bloom73 0.666667 -14.4252 25.5% 0.666667 -14.4252 swat-s 0.666667 -14.4252 (past 0.666667 -14.4252 2.24 0.666667 -14.4252 81.75 0.666667 -14.4252 takeover 0.666667 -14.4252 (group 0.666667 -14.4252 (listed 0.666667 -14.4252 84.48 0.5 -14.4255 93.59 0.5 -14.4255 la, 0.5 -14.4255 mr, 0.375 -14.4261 bhattacharyya, 0.5 -14.4267 p(michael 0.5 -14.4267 50.54 0.5 -14.4267 ˘ 0.5 -14.4267 trainingdata 0.5 -14.4267 comparee 0.5 -14.4267 einer 0.5 -14.4267 you) 0.5 -14.4267 aberrant 0.5 -14.4267 ξ, 0.5 -14.4267 automated, 0.666667 -14.4268 97.01 0.666667 -14.4268 sstk 0.666667 -14.4268 normtweets 0.666667 -14.4268 4.21 0.666667 -14.4268 actualy 0.666667 -14.4268 bhagat, 0.375 -14.427 env(q 0.25 -14.4273 popped 0.163265 -14.4279 into four 0.5 -14.4279 +hyper 0.5 -14.4279 87.41 0.5 -14.4279 wmt09n 0.5 -14.4279 saturday 0.5 -14.4279 developmentdata 0.5 -14.4279 60.60 0.5 -14.4279 add multiple 0.666667 -14.4284 nyberg, 0.666667 -14.4284 nonoverlapped 0.666667 -14.4284 mani, 0.666667 -14.4284 usvt 0.666667 -14.4284 74.88 0.666667 -14.4284 (slightly 0.666667 -14.4284 terman 0.666667 -14.4284 +0.10 0.666667 -14.4284 silk 0.666667 -14.4284 ingenious 0.666667 -14.4284 tongue, 0.666667 -14.4284 confidence) 0.666667 -14.4284 lands 0.052 -14.4287 we use a 0.0719697 -14.429 not. 0.5 -14.4291 stranger” 0.5 -14.4291 lmss 0.5 -14.4291 c1: 0.5 -14.4291 loop: 0.5 -14.4291 ahn, 0.189189 -14.4294 f = 0.375 -14.4297 transverse 0.666667 -14.43 54.29 0.666667 -14.43 marjorie 0.666667 -14.43 impressive: 0.666667 -14.43 states” 0.666667 -14.43 dual+cm 0.666667 -14.43 (geometric 0.666667 -14.43 vectortd(a,b) 0.666667 -14.43 (original 0.666667 -14.43 logics, 0.666667 -14.43 plaza 0.285714 -14.4302 dataskdelay 0.5 -14.4303 appositions 0.5 -14.4303 j} 0.5 -14.4303 σ(t) 0.5 -14.4303 triangular 0.5 -14.4303 1996); 0.375 -14.4306 32.28 0.375 -14.4306 progresses, 0.5 -14.4315 chouinard 0.5 -14.4315 )φ 0.666667 -14.4316 discriminability 0.666667 -14.4316 parent: 0.666667 -14.4316 marie-catherine 0.666667 -14.4316 genefamily 0.666667 -14.4316 90.65 0.666667 -14.4316 prevention relations 0.666667 -14.4316 @c 0.666667 -14.4316 emphasised 0.666667 -14.4316 low-pass 0.333333 -14.4323 o-dep 0.428571 -14.4326 +negative 0.5 -14.4327 partitioning, 0.5 -14.4327 (generic 0.5 -14.4327 alistair 0.5 -14.4327 k(xi,xj) 0.666667 -14.4332 fats 0.666667 -14.4332 (2009a; 0.666667 -14.4332 2.2% 0.666667 -14.4332 yanjun 0.666667 -14.4332 ρxr 0.666667 -14.4332 machinetranslation 0.666667 -14.4332 overlooks 0.666667 -14.4332 kirk 0.666667 -14.4332 44.06 0.666667 -14.4332 humanoid 0.666667 -14.4332 dropout periods 0.666667 -14.4332 strangeman, 0.666667 -14.4332 “nearly 0.666667 -14.4332 73.5% 0.375 -14.4333 candidate paraphrases. 0.307692 -14.4334 linear order 0.5 -14.4339 retreat 0.5 -14.4339 properly, 0.5 -14.4339 53.8% 0.5 -14.4339 463 0.5 -14.4339 comonly 0.5 -14.4339 derivable 0.666667 -14.4348 subfigure 0.666667 -14.4348 agendabased 0.666667 -14.4348 olympus/ravenclaw 0.666667 -14.4348 (raoult, 0.666667 -14.4348 1129 0.666667 -14.4348 subscribe 0.666667 -14.4348 ∈v 0.666667 -14.4348 expert-driven 0.5 -14.4351 conceivably, 0.5 -14.4363 “lettertran” 0.666667 -14.4364 youngjun 0.666667 -14.4364 rt: 0.666667 -14.4364 (0.4324 0.666667 -14.4364 sreqs(x,a) 0.666667 -14.4364 :γ 0.222222 -14.4369 gaz 0.375 -14.437 predicate sense 0.375 -14.437 tmp 0.5 -14.4376 username 0.5 -14.4376 stem: 0.5 -14.4376 encloses 0.5 -14.4376 88.2% 0.5 -14.4376 1d 0.428571 -14.4378 resultson 0.428571 -14.4378 ˆyt 0.307692 -14.4379 noticing 0.666667 -14.438 1388 0.666667 -14.438 55.26 0.666667 -14.438 64.82 0.666667 -14.438 71.35 0.666667 -14.438 i,k 0.666667 -14.438 lastly 0.666667 -14.438 vier 0.666667 -14.438 matching) 0.666667 -14.438 (j−i) 0.666667 -14.438 chile, 0.5 -14.4388 83.34 0.666667 -14.4397 slumped 0.666667 -14.4397 deaf, 0.666667 -14.4397 anymore, 0.666667 -14.4397 141, 0.666667 -14.4397 ml, 0.666667 -14.4397 14.64 0.666667 -14.4397 55.24 0.375 -14.4397 btng 0.375 -14.4397 20.6 0.5 -14.44 ts[pi] 0.5 -14.44 nullx 0.5 -14.44 6available 0.5 -14.44 was) 0.428571 -14.4409 n×n 0.5 -14.4412 svm-hmm 0.5 -14.4412 24.20 0.5 -14.4412 (specifically 0.5 -14.4412 wn-affect 0.666667 -14.4413 81.81 0.666667 -14.4413 ∆f1 0.666667 -14.4413 hung 0.666667 -14.4413 0.066 0.666667 -14.4413 0.930 0.666667 -14.4413 manhattan, 0.179487 -14.4419 follows: • 0.428571 -14.442 prating 0.170732 -14.4423 in the following, 0.5 -14.4424 13we 0.5 -14.4424 8,476 0.5 -14.4424 q-elab 0.5 -14.4424 0.693 0.5 -14.4424 election, 0.666667 -14.4429 0.174 0.666667 -14.4429 disorders, 0.666667 -14.4429 joão 0.666667 -14.4429 “request 0.666667 -14.4429 rdm, 0.666667 -14.4429 lebanon 0.666667 -14.4429 ys, 0.666667 -14.4429 bronze 0.666667 -14.4429 shuly wintner 0.666667 -14.4429 speech: 0.666667 -14.4429 fpe 0.666667 -14.4429 appropriated 0.25 -14.4431 right context 0.222222 -14.4434 unchanged. 0.16 -14.4434 al., 2000; 0.375 -14.4434 0.075 0.5 -14.4436 hk10, 0.5 -14.4436 22.42 0.5 -14.4436 tanier 0.5 -14.4436 polarityrank 0.5 -14.4436 producers 0.5 -14.4436 (german 0.5 -14.4436 ns(n) 0.666667 -14.4445 svr-pmi 0.666667 -14.4445 “overlap”, 0.666667 -14.4445 〈w,wprime〉 0.666667 -14.4445 structure” 0.666667 -14.4445 51.64 0.666667 -14.4445 37.29 0.666667 -14.4445 94.7% 0.666667 -14.4445 1154 0.666667 -14.4445 autosemantic 0.666667 -14.4445 better: 0.666667 -14.4445 powell 0.5 -14.4449 disrupt 0.5 -14.4449 medlda-based 0.5 -14.4449 ‘m’ 0.5 -14.4449 tree-kernel 0.136364 -14.4451 stage. 0.375 -14.4452 vossen, 0.109091 -14.4456 we start 0.5 -14.4461 ink 0.5 -14.4461 oposed 0.5 -14.4461 (can 0.5 -14.4461 30.13 0.666667 -14.4462 εt 0.666667 -14.4462 admire, 0.666667 -14.4462 1027 0.100775 -14.4468 cluster. 0.5 -14.4473 833 0.5 -14.4473 du, 0.5 -14.4473 κji 0.5 -14.4473 tq 0.078341 -14.4477 is represented 0.666667 -14.4478 shaft 0.666667 -14.4478 prep(in), 0.666667 -14.4478 nucleus-satellite 0.666667 -14.4478 tst 0.666667 -14.4478 desks 0.666667 -14.4478 15.97 0.666667 -14.4478 (§6), 0.666667 -14.4478 4(b), 0.2 -14.4481 300 400 0.5 -14.4485 families: 0.5 -14.4485 51.06 0.5 -14.4485 svmmulticlass 0.5 -14.4485 566 0.5 -14.4485 6,6the 0.238095 -14.4488 (2000). 0.666667 -14.4495 4.94 0.666667 -14.4495 (mishne 0.666667 -14.4495 88.50 0.666667 -14.4495 38.38 0.666667 -14.4495 ramifications 0.666667 -14.4495 “have” 0.666667 -14.4495 41.04 0.666667 -14.4495 queues 0.666667 -14.4495 (lsh) 0.666667 -14.4495 prob(e) 0.5 -14.4498 stipulate 0.5 -14.4498 642 0.11215 -14.45 participated 0.138462 -14.4508 tools for 0.5 -14.451 parameter-free 0.5 -14.451 91.36 0.5 -14.451 cooccurs 0.666667 -14.4511 inputoutput 0.666667 -14.4511 pb( 0.666667 -14.4511 f-score: 0.666667 -14.4511 ipod, 0.666667 -14.4511 (frank 0.666667 -14.4511 (dc) 0.666667 -14.4511 whichhas 0.666667 -14.4511 nigel 0.666667 -14.4511 46.23 0.666667 -14.4511 58.66 0.666667 -14.4511 splits: 0.666667 -14.4511 (matsoukas 0.238095 -14.4512 thesis, 0.333333 -14.4519 therapist 0.5 -14.4522 “enter”, 0.5 -14.4522 1997): 0.5 -14.4522 house) 0.5 -14.4522 c2: 0.666667 -14.4527 achieved: 0.666667 -14.4527 (leech 0.666667 -14.4527 meteor-rank 0.666667 -14.4527 rationality, 0.666667 -14.4527 57.37 0.666667 -14.4527 denied 0.666667 -14.4527 80.88 0.666667 -14.4527 +0.17 0.666667 -14.4527 49.65 0.123457 -14.4528 is estimated 0.08 -14.453 way. 0.428571 -14.4535 [w] 0.375 -14.4535 diab 0.5 -14.4535 allocates 0.5 -14.4535 donia 0.5 -14.4535 74.70 0.666667 -14.4544 24.00 0.666667 -14.4544 wsj: 0.666667 -14.4544 39.75 0.666667 -14.4544 80.42 0.666667 -14.4544 -1.38 0.666667 -14.4544 38.65 0.666667 -14.4544 preprocessing: 0.666667 -14.4544 <3 0.666667 -14.4544 40.94 0.666667 -14.4544 ££month 0.666667 -14.4544 jang, 0.666667 -14.4544 cascades 0.666667 -14.4544 nss 0.666667 -14.4544 (22), 0.666667 -14.4544 18.8% 0.666667 -14.4544 cross(i,j,x,y) 0.666667 -14.4544 best-first, 0.375 -14.4544 sightseeing 0.5 -14.4547 synchrony 0.5 -14.4547 +hyp2 0.5 -14.4547 ∇ 0.5 -14.4547 cranor, 0.156863 -14.455 3 0 0.238095 -14.4559 convenience, 0.5 -14.4559 image: 0.5 -14.4559 ep: 0.5 -14.4559 inhibit 0.5 -14.4559 catherine 0.5 -14.4559 periphrases, 0.5 -14.4559 morphology-sensitive 0.5 -14.4559 (2004, 0.666667 -14.456 headword-lexicalized 0.666667 -14.456 mat 0.666667 -14.456 φunq(rlscript) 0.666667 -14.456 l2f 0.666667 -14.456 window; 0.666667 -14.456 copyflags 0.666667 -14.456 cheryl 0.666667 -14.456 sydney 0.666667 -14.456 anywhere, 0.666667 -14.456 0.07% 0.0643087 -14.4564 2004). 0.2 -14.4569 chodorow, 0.156863 -14.457 response. 0.5 -14.4572 english-hindiparallel 0.5 -14.4572 70,000 0.5 -14.4572 enginering, 0.666667 -14.4577 routledge, 0.666667 -14.4577 confirm, 0.666667 -14.4577 cohorts 0.666667 -14.4577 adjusted, 0.666667 -14.4577 50.43 0.666667 -14.4577 [specific] 0.666667 -14.4577 k′1, 0.5 -14.4584 improvements: 0.5 -14.4584 bot, 0.5 -14.4584 73.39 0.375 -14.4591 >> 0.666667 -14.4593 classically 0.666667 -14.4593 in(v) 0.666667 -14.4593 granger 0.666667 -14.4593 life” 0.666667 -14.4593 0.1381 0.666667 -14.4593 1098 0.131579 -14.4593 structures are 0.5 -14.4597 n-gram) 0.5 -14.4597 pat 0.5 -14.4597 embraces 0.5 -14.4597 89.05 0.5 -14.4597 (starting 0.375 -14.46 company1 0.5 -14.4609 desideratum 0.5 -14.4609 f-s 0.5 -14.4609 anonymously 0.666667 -14.461 pleasure 0.666667 -14.461 preiss 0.666667 -14.461 judged, 0.666667 -14.461 ∧infected(x,malaria) 0.666667 -14.461 online-a 0.666667 -14.461 cartoon 0.666667 -14.461 95.52 0.666667 -14.461 “artificial 0.666667 -14.461 27.7% 0.666667 -14.461 4.82 0.666667 -14.461 browsers 0.666667 -14.461 hmeant’s 0.0743802 -14.4617 the value of 0.5 -14.4622 0.239 0.5 -14.4622 soft-margin 0.5 -14.4622 “of 0.5 -14.4622 inverse, 0.5 -14.4622 crc(p) 0.5 -14.4622 electric 0.666667 -14.4627 sugiyama 0.666667 -14.4627 assignments), 0.666667 -14.4627 0.456 0.666667 -14.4627 20.48 0.666667 -14.4627 xs 0.666667 -14.4627 expansion: 0.666667 -14.4627 children: 0.666667 -14.4627 42.24 0.666667 -14.4627 pattern-matching 0.214286 -14.463 experimental results are 0.5 -14.4634 table6 0.5 -14.4634 fluffy 0.5 -14.4634 nullnull(null) 0.5 -14.4634 67.80 0.666667 -14.4643 itf 0.666667 -14.4643 876 0.666667 -14.4643 camp 0.666667 -14.4643 hal, 0.666667 -14.4643 40.57 0.666667 -14.4643 black women 0.666667 -14.4643 1,2,3 0.666667 -14.4643 x? 0.666667 -14.4643 dyck 0.666667 -14.4643 levenshtein, 0.666667 -14.4643 bleul 0.175 -14.4645 for english and 0.5 -14.4646 false-positives 0.5 -14.4646 0.08, 0.0734694 -14.4648 2011). 0.079602 -14.4649 practice, 0.0837989 -14.4649 we therefore 0.333333 -14.4651 cj. 0.156863 -14.4656 (2007) and 0.5 -14.4659 tx. 0.5 -14.4659 segments: 0.666667 -14.466 s.p 0.666667 -14.466 αεi 0.666667 -14.466 reconstruction, 0.666667 -14.466 70-80% 0.666667 -14.466 filler, 0.666667 -14.466 a71random words−js 0.666667 -14.466 equates 0.666667 -14.466 [it] 0.666667 -14.466 -a 0.666667 -14.466 m. cafarella, 0.666667 -14.466 81.6 78.8 80.2 0.5 -14.4671 reformulating 0.5 -14.4671 clc-dpmm 0.5 -14.4671 pcs 0.5 -14.4671 “would 0.375 -14.4675 folded 0.25 -14.4678 different topics 0.428571 -14.4695 (dolan 0.5 -14.4697 icle 0.5 -14.4697 log(1 0.5 -14.4697 edikt 0.5 -14.4697 (broder 0.5 -14.4697 (ranging 0.5 -14.4697 22.00 0.333333 -14.4709 tri 0.5 -14.4709 5all 0.5 -14.4709 projecteddependency 0.5 -14.4709 neuter 0.5 -14.4709 1416 0.0545455 -14.4714 a very 0.163265 -14.4721 we replace 0.189189 -14.4722 analysis shows 0.5 -14.4722 myxedema,... 0.5 -14.4722 wordpredictor 0.5 -14.4722 johan 0.5 -14.4722 said: 0.5 -14.4722 mozambique 0.5 -14.4722 (nor 0.333333 -14.4726 (luong 0.5 -14.4734 halliday 0.5 -14.4734 unpublished 0.333333 -14.4743 rec prec 0.5 -14.4747 ti−1 0.5 -14.4747 webps 0.5 -14.4747 1bi 0.5 -14.4759 (call 0.5 -14.4759 651 0.375 -14.476 penn, 0.428571 -14.4771 for arabic. 0.5 -14.4772 54.19 0.5 -14.4772 clicked, 0.5 -14.4772 tols 0.5 -14.4772 seal 0.156863 -14.4774 threshold for 0.119565 -14.4775 they also 0.333333 -14.4777 apfel 0.227273 -14.4779 bergler, 0.428571 -14.4782 “small 0.5 -14.4785 jena 0.5 -14.4785 “un” 0.375 -14.4798 not–shared 0.5 -14.481 argument”, 0.5 -14.481 linked directly 0.5 -14.481 pretrained 0.5 -14.481 o1, 0.5 -14.481 md-i 0.5 -14.481 insult 0.5 -14.481 1993): 0.5 -14.481 1143 0.5 -14.481 (query 0.5 -14.4823 1395 0.5 -14.4823 summaryof 0.5 -14.4823 1,353 0.428571 -14.4825 corpus, so 0.375 -14.4826 (mert) (och, 0.266667 -14.4835 bonnie 0.5 -14.4836 872 0.5 -14.4836 $0.05 0.333333 -14.4844 (bar-hillel 0.375 -14.4845 opensource 0.184211 -14.4848 agrees 0.5 -14.4848 22.28 0.5 -14.4848 regexp 0.5 -14.4848 sys+sim 0.333333 -14.4853 (1999) and 0.041779 -14.4857 refer 0.428571 -14.4858 qualifiers 0.428571 -14.4858 drafts 0.5 -14.4861 yangarber, 0.5 -14.4861 document), 0.5 -14.4861 ring 0.428571 -14.4869 5-tuple 0.333333 -14.487 forbidding 0.375 -14.4874 cit 0.5 -14.4874 523 0.5 -14.4887 (em 0.5 -14.4887 imperfective 0.175 -14.4889 of acl 0.428571 -14.4891 nicola 0.5 -14.4899 (features 0.307692 -14.4909 tjong 0.307692 -14.4909 1–9, 0.5 -14.4912 (regular 0.5 -14.4912 stolfi 0.5 -14.4912 formulations, 0.428571 -14.4924 p(yi 0.5 -14.4925 luxury 0.217391 -14.493 1986). 0.238095 -14.4931 breast 0.5 -14.4938 fb3 0.5 -14.4938 624 0.5 -14.4938 ger 0.5 -14.4938 .; 0.375 -14.4941 baseball-hockey 0.375 -14.4941 massimo 0.428571 -14.4946 1443 0.5 -14.4964 isaiah 0.5 -14.4964 pr(y 0.5 -14.4964 [n] 0.5 -14.4964 s(n) 0.166667 -14.4969 tweet. 0.5 -14.4976 u0 0.5 -14.4976 (salakhutdinov 0.145455 -14.4981 also experimented 0.227273 -14.4989 suffices 0.5 -14.4989 beth 0.5 -14.4989 elevated 0.5 -14.4989 implicatives 0.428571 -14.499 nullij 0.428571 -14.499 pos-rules 0.375 -14.499 mayo 0.132353 -14.4993 7 0 0.307692 -14.4996 sentence si 0.333333 -14.4998 éplucher 0.375 -14.4999 trg 0.428571 -14.5001 38123 0.5 -14.5002 therefore: 0.5 -14.5002 (& 0.214286 -14.5006 be stored 0.0291777 -14.5014   0.5 -14.5015 (vi,vj) 0.5 -14.5015 concept-specific 0.5 -14.5015 n′ 0.5 -14.5015 kiran 0.375 -14.5019 24.16 0.428571 -14.5023 unseen. 0.428571 -14.5023 sift 0.428571 -14.5023 str 0.428571 -14.5023 φ(x,y). 0.428571 -14.5023 of [2/ 0.428571 -14.5023 och. 0.428571 -14.5023 19c 0.428571 -14.5023 darpa. 0.428571 -14.5023 sdev 0.428571 -14.5023 tents 0.1875 -14.5024 these four 0.5 -14.5028 (predicted 0.5 -14.5028 94.22 0.5 -14.5028 hw. 0.5 -14.5028 demonstrate, 0.5 -14.5028 primitive, 0.227273 -14.5036 also help 0.206897 -14.504 respectively) 0.5 -14.5041 0.927 0.5 -14.5041 second-topmost 0.5 -14.5041 ∀s,t 0.126582 -14.5044 constraints are 0.307692 -14.5052 (t, 0.5 -14.5054 point), 0.333333 -14.5059 0.879 0.5 -14.5067 privacy, 0.5 -14.5067 distinctive  0.5 -14.5067 2.15 0.150943 -14.5072 factor. 0.266667 -14.5075 (artstein 0.5 -14.508 featuresfor 0.5 -14.508 virginia 0.5 -14.5093 nulln 0.5 -14.5093 blogs: 0.5 -14.5093 ∈n 0.375 -14.5097 mona 0.5 -14.5106 “≥all 0.5 -14.5106 bikes). 0.5 -14.5106 ”is 0.285714 -14.5113 resnik, 2008; 0.227273 -14.5113 prev 0.125 -14.5113 k=1 0.375 -14.5116 (lavergne 0.5 -14.5119 p1) 0.5 -14.5119 1574 0.5 -14.5119 per-target 0.0605187 -14.5126 nature 0.333333 -14.5128 es→en 0.333333 -14.5128 null4 0.238095 -14.5129 uncover 0.5 -14.5133 469 0.5 -14.5133 location-based 0.5 -14.5133 5.1.4 0.170732 -14.5136 ksummationdisplay 0.119048 -14.5138 is measured 0.217391 -14.5156 been introduced 0.5 -14.5159 ‘gap 0.5 -14.5159 as) 0.25 -14.5166 and vanderwende 0.375 -14.5166 −1) 0.103448 -14.5171 this set 0.333333 -14.5172 miles 0.333333 -14.5172 search algorithms 0.5 -14.5172 26.26 0.5 -14.5185 “web 0.0625 -14.5192 grant 0.375 -14.5195 bama 0.5 -14.5198 agt 0.5 -14.5198 president’s 0.333333 -14.5199 viewpoint. 0.132353 -14.5204 attributed 0.266667 -14.5208 len 0.162791 -14.5208 for translating 0.5 -14.5211 (ks 0.307692 -14.5213 formats. 0.103448 -14.5217 remainder of 0.5 -14.5225 null0,1nullnullnullnull 0.5 -14.5225 manfred 0.5 -14.5225 o(l) 0.0248385 -14.5228 that are 0.159091 -14.5234 is. 0.375 -14.5235 agirre, 0.5 -14.5238 quadruplet 0.150943 -14.5239 workshop. 0.5 -14.5251 knife 0.5 -14.5251 rip 0.5 -14.5264 1(yi,j 0.5 -14.5264 morpho 0.5 -14.5264 (four 0.5 -14.5264 grifs 0.375 -14.5265 spencerk 0.111111 -14.5284 reflects the 0.5 -14.5291 (basic 0.5 -14.5291 [k] 0.333333 -14.5296 l(w) 0.5 -14.5304 29.97 0.5 -14.5304 r10 0.5 -14.5304 holds: 0.5 -14.5304 a80 0.5 -14.5318 2–21 0.5 -14.5318 amy 0.113636 -14.532 is added 0.097561 -14.5323 side, 0.5 -14.5331 approachis 0.5 -14.5331 doctor’s 0.175 -14.5337 ministry of 0.5 -14.5344 mv 0.5 -14.5344 devoicing 0.25 -14.5345 (hahn 0.25 -14.5358 conforms 0.5 -14.5358 currhypo 0.333333 -14.5367 ca: 0.238095 -14.5369 “look 0.266667 -14.5371 void 0.5 -14.5371 characteristic, 0.5 -14.5371 44.05 0.5 -14.5371 cues: 0.0283828 -14.5381 note 0.105769 -14.5383 the web. 0.375 -14.5385 octants 0.5 -14.5385 time/date 0.5 -14.5385 23.9 0.126761 -14.5398 fashion 0.137931 -14.54 is denoted 0.307692 -14.5402 making this 0.375 -14.5405 this is illustrated 0.333333 -14.5412 we cannot compare 0.5 -14.5412 anecdotally, 0.5 -14.5412 else: 0.5 -14.5412 correlationof 0.375 -14.5425 ies, 0.307692 -14.5427 (arun 0.166667 -14.5439 sciences, 0.333333 -14.5439 “less 0.5 -14.5439 φ1 0.5 -14.5439 (crfs), 0.333333 -14.5457 dell 0.266667 -14.5458 dar 0.0363036 -14.5461 to use 0.5 -14.5466 critics 0.5 -14.5466 clothes, 0.5 -14.5466 perseus 0.137931 -14.5471 position. 0.5 -14.5479 ’he 0.5 -14.5479 honestly 0.5 -14.5479 wait, 0.5 -14.5479 (probably 0.5 -14.5479 generalises 0.176471 -14.549 an example is 0.333333 -14.5493 zamparelli, 0.333333 -14.5493 org, 0.5 -14.5493 comme 0.5 -14.5506 p+t+r+c 0.5 -14.5506 systemon 0.266667 -14.5508 4.1 evaluation 0.5 -14.552 kelly 0.5 -14.552 splitg 0.375 -14.5527 (ttsa-a-te, 0.5 -14.5547 (water 0.5 -14.5547 gene_expression 0.5 -14.5547 “#” 0.266667 -14.5552 unknown, 0.1875 -14.5553 steps are 0.333333 -14.5557 {act 0.5 -14.5561 (calzolari 0.5 -14.5561 cqp 0.5 -14.5561 lexicographical 0.333333 -14.5566 youth 0.5 -14.5575 45.38 0.5 -14.5575 community; 0.5 -14.5575 “complete 0.5 -14.5575 specializations 0.0387016 -14.5581 system, 0.5 -14.5602 u2: 0.5 -14.5602 (movie 0.333333 -14.5612 rivaling 0.5 -14.5616 res, 0.5 -14.5616 southwest 0.5 -14.563 stanford-ubc 0.5 -14.563 isomorphicto 0.0802139 -14.5631 evaluate our 0.5 -14.5643 motwani 0.5 -14.5643 boost, 0.5 -14.5643 40000 0.333333 -14.5648 t.j. 0.096 -14.5649 nsf 0.155556 -14.565 was obtained 0.5 -14.5657 bracketleftbigtask 0.5 -14.5657 project) 0.5 -14.5657 701 0.0858896 -14.5664 it will 0.208333 -14.5667 lillian 0.5 -14.5671 graphg 0.5 -14.5671 dipper 0.1875 -14.5672 responses are 0.0757576 -14.5673 representation. 0.03223 -14.5677 be used 0.181818 -14.5677 its use 0.0654545 -14.5679 other. 0.11236 -14.5682 to. 0.5 -14.5685 plank 0.5 -14.5685 bp, 0.333333 -14.5694 summarizers, 0.5 -14.5699 (alshawi 0.5 -14.5699 adjective), 0.103774 -14.5704 would also 0.5 -14.5713 subconcepts 0.285714 -14.5721 construed 0.4 -14.5724 (chiticariu 0.4 -14.5724 sri-lm 0.4 -14.5724 albums 0.4 -14.5724 (representing 0.4 -14.5724 “la 0.4 -14.5724 pos-2grams 0.4 -14.5724 sp. 0.4 -14.5724 anotate 0.4 -14.5724 0.408 0.4 -14.5724 mtv 0.4 -14.5724 ¯s 0.4 -14.5724 acyg3(% 0.5 -14.5726 hopefully, 0.5 -14.5726 youtube 0.5 -14.5726 asvm 0.3 -14.573 (primeprime) 0.0454545 -14.5737 experiments, 0.5 -14.574 australopithecus 0.5 -14.574 cooling 0.5 -14.574 ps, 0.0723982 -14.5743 referring 0.4 -14.5747 (smolensky, 0.4 -14.5747 s-n 0.0853659 -14.5747 specifically, we 0.5 -14.5754  0.5 -14.5754 us: 0.4 -14.5769 yngve, 0.4 -14.5769 error return 0.0513393 -14.5775 correspond 0.5 -14.5782 sous 0.5 -14.5782 obtained: 0.5 -14.5782 gh 0.4 -14.5791 relative difference 0.0573066 -14.5796 presence 0.5 -14.5796 593 0.5 -14.5796 suggestions, 0.1 -14.5805 proportional 0.5 -14.581 p+t+r 0.5 -14.581 bush, 0.3 -14.5822 nouns, verbs 0.5 -14.5824 owl, 0.5 -14.5824 nowo-pcfg 0.4 -14.5825 0.835 0.4 -14.5825 whichwe 0.208333 -14.5835 r-1 0.4 -14.5836 0.02, 0.0647482 -14.5837 for any 0.5 -14.5838 daand 0.5 -14.5838 member-of 0.3 -14.5839 dredze, 0.4 -14.5847 “je 0.3 -14.5848 f5 0.0952381 -14.5849 remainder 0.5 -14.5852 discriminant-based 0.5 -14.5852 æk 0.5 -14.5852 “such 0.4 -14.5859 0.954 0.0677291 -14.5859 classifier. 0.0707965 -14.5861 in contrast 0.5 -14.5867 hell, 0.5 -14.5867 pa(x) 0.5 -14.5867 fas 0.193548 -14.587 2011 system 0.3 -14.5873 online-behavior 0.375 -14.5881 q(θ) 0.5 -14.5881 synset; 0.5 -14.5881 1003 0.5 -14.5881 between: 0.4 -14.5893 mbc news 0.5 -14.5895 αij(1) 0.5 -14.5895 headof 0.5 -14.5895 39.05 0.133333 -14.59 head. 0.375 -14.5902 (:* 0.333333 -14.5909 ∀i 0.208333 -14.5912 for details). 0.0916031 -14.5914 statistics of 0.227273 -14.5917 conditionally 0.5 -14.5923 paraphrase) 0.4 -14.5927 (pas¸ca, 0.5 -14.5937 (case 0.5 -14.5937 question), 0.5 -14.5937 definition 0.5 -14.5952 burger 0.5 -14.5952 parser: 0.5 -14.5952 90.43 0.25 -14.5959 2008a). 0.3 -14.5967 tlm 0.4 -14.5972 content-container 0.4 -14.5972 α,β 0.4 -14.5972 h&k 0.375 -14.5977 emmanuel 0.5 -14.598 sparsely 0.4 -14.5983 (ccg; 0.5 -14.5994 21.35 0.5 -14.5994 “np{answer} 0.1875 -14.6009 vds-7 0.5 -14.6009 mention-based 0.5 -14.6009 1.2m 0.0677291 -14.6021 occurring 0.5 -14.6023 enjoys 0.5 -14.6023 p+t+c 0.5 -14.6023 goldsmith 0.118421 -14.603 example, consider 0.5 -14.6037 notarrowright 0.5 -14.6037 adding, 0.4 -14.6041 multi-tasking 0.25 -14.6045 orig 0.4 -14.6052 currently uses 0.4 -14.6052 dallas, 0.333333 -14.6052 )⇒( 0.5 -14.6052 0.5 -14.6052 dmitry 0.333333 -14.6062 in parenthesis. 0.4 -14.6064 quarterly, 0.5 -14.6066 r∗ 0.3 -14.607 fellowship. 0.5 -14.608 b¨ogel 0.5 -14.608 (apart 0.333333 -14.6081 6.0 0.3 -14.6087 forty 0.137931 -14.6088 wanted 0.152174 -14.6091 item. 0.375 -14.6095 approx-stopwords 0.5 -14.6095 postech 0.0930233 -14.61 for further 0.375 -14.6106 = 4, 0.113636 -14.6116 et al., 2002; 0.285714 -14.6118 hartu 0.4 -14.6121 rachele 0.5 -14.6124 (ordered 0.5 -14.6124 ˆza 0.181818 -14.6125 synset. 0.3 -14.6139 joe 0.333333 -14.6139 negated, 0.148936 -14.614 this value 0.4 -14.6145 96.8 0.171429 -14.6146 significant 0.4 -14.6156 theme(core 0.4 -14.6156 jakob 0.5 -14.6167 35, 0.5 -14.6167 patel 0.5 -14.6167 ago, 0.5 -14.6167 crack 0.285714 -14.6168 opponents 0.4 -14.6168 85.7% 0.4 -14.6168 (a1) 0.0786517 -14.6169 phrase. 0.375 -14.6171 ∂ 0.129032 -14.6179 (fellbaum, 0.5 -14.6182 recursively: 0.166667 -14.6187 it is therefore 0.333333 -14.6187 kruengkrai 0.4 -14.6191 98.1 0.375 -14.6193 (norm) 0.5 -14.6196 irst 0.5 -14.6196 cognitive, 0.3 -14.62  0.285714 -14.6201 liu. 0.102804 -14.6201 we simply 0.4 -14.6203 98195, 0.4 -14.6203 r&d 0.25 -14.6211 semiring, 0.5 -14.6211 “iphone 0.171429 -14.6216 acm 0.118421 -14.6219 call this 0.217391 -14.6221 as negative. 0.5 -14.6226 tribute 0.5 -14.624 panic 0.5 -14.624 yesterday, 0.266667 -14.6254 algorithm 1, 0.5 -14.6255 parenleftbigsummationdisplay 0.5 -14.6255 bicknell.64 0.4 -14.6261 ejection 0.4 -14.6261 yellow, 0.5 -14.6269 wα, 0.5 -14.6269 msmsn 0.4 -14.6273 dephosphorylation 0.375 -14.6281 pscfgs. 0.375 -14.6281 m∏ 0.5 -14.6284 (diff 0.5 -14.6284 scientist, 0.5 -14.6284 table2 0.4 -14.6285 cheese, 0.4 -14.6285 argminw 0.4 -14.6285 21.3 0.131148 -14.6286 was funded 0.4 -14.6297 high-pass 0.5 -14.6299 (6-b,c) 0.375 -14.6303 1,i 0.333333 -14.6304 acm, 0.208333 -14.6309 squares 0.375 -14.6314 (es) 0.5 -14.6314 fh 0.0724638 -14.6317 relation. 0.5 -14.6328 funds, 0.4 -14.6332 (chi 0.4 -14.6332 turmo, 0.375 -14.6336 orientation. 0.4 -14.6344 sergio 0.375 -14.6347 wonderful 0.4 -14.6356 91.24 0.4 -14.6356 0.570 0.4 -14.6356 match(di,dj) 0.4 -14.6356 0.369 0.375 -14.6358 pre-order 0.375 -14.6358 programme. 0.375 -14.6358 costing 0.375 -14.6358 zens, 0.3 -14.6359 seattle 0.1 -14.6362 contract 0.125 -14.6369 abstractness 0.5 -14.6373 29.48 0.159091 -14.6375 preceded 0.4 -14.6379 regroup 0.3 -14.6385 percentile 0.5 -14.6388 bioevent 0.5 -14.6388 i’s 0.4 -14.6391 coref: 0.4 -14.6391 g(θ) 0.3 -14.6394 _? 0.0156078 -14.6398 there 0.4 -14.6403 sh, 0.4 -14.6403 (weischedel 0.5 -14.6403 30th 0.5 -14.6403 phys 0.5 -14.6417 morphemes: 0.5 -14.6417 “food” 0.118421 -14.6418 f: 0.3 -14.6421 event representation. 0.4 -14.6427 γprime 0.5 -14.6432 wl, 0.5 -14.6432 misspelling, 0.5 -14.6432 anullsemc 0.333333 -14.6433 = 5) 0.4 -14.6439 nounphrase 0.333333 -14.6443 rubber 0.5 -14.6447 edge’s 0.0543478 -14.645 in natural language 0.4 -14.6451 0.932 0.4 -14.6451 -0.42 0.0940171 -14.6453 york 0.333333 -14.6453 sparsifying 0.333333 -14.6453 19, 0.266667 -14.6459 be successful 0.5 -14.6462 θdkj 0.5 -14.6462 wp1 0.4 -14.6463 -­‐1 0.4 -14.6463 0.249 0.4 -14.6487 domestic 0.4 -14.6487 bones 0.208333 -14.6489 has, 0.217391 -14.6492 fahrr¨ader 0.333333 -14.6493 spots 0.0543478 -14.6498 models are 0.4 -14.6499 hire 0.4 -14.6499 s(x,y) 0.4 -14.6499 graphs: 0.1125 -14.6499 5 conclusions 0.0964912 -14.6507 work by 0.4 -14.6511 , 立场 0.4 -14.6511 switzerland, 0.148936 -14.6519 alignment model 0.5 -14.6522 bmt 0.208333 -14.6533 as noted in 0.4 -14.6535 disagreed with 0.5 -14.6537 4): 0.222222 -14.655 advp 0.5 -14.6552 generalist 0.5 -14.6552 (“the 0.333333 -14.6563 yon 0.4 -14.6571 criminals 0.333333 -14.6573 (1b) 0.266667 -14.658 (gap+) 0.4 -14.6583 omri 0.4 -14.6583 each) 0.4 -14.6583 2002): 0.333333 -14.6583 rieser, 0.5 -14.6583 signify 0.4 -14.6595 2.67 0.5 -14.6598 miguel 0.5 -14.6598 en-de, 0.5 -14.6598 ηt 0.4 -14.6608 focusedon 0.4 -14.6608 interlayer 0.4 -14.6608 ‘yes’ 0.235294 -14.6608 π( 0.142857 -14.6609 johns hopkins 0.3 -14.6611 lch 0.5 -14.6613 m.d. 0.5 -14.6613 ≻x 0.3 -14.662 -に 0.208333 -14.6627 contain all 0.5 -14.6628 lmt 0.5 -14.6628 can: 0.5 -14.6628 4available 0.0142375 -14.664 in this 0.5 -14.6643 ‘not 0.0584416 -14.6659 goal is 0.5 -14.6659 664 0.5 -14.6659 বnull 0.208333 -14.6665 p´olya 0.25 -14.6667 accur 0.4 -14.6668 americas, 0.4 -14.6668 historically, 0.4 -14.6668 precisely: 0.0839161 -14.6671 be easily 0.5 -14.6674 treatedas 0.272727 -14.6675 solver. 0.0598007 -14.668 interested 0.4 -14.6681 hirohata 0.5 -14.6689 (cohen’s 0.4 -14.6693 (4b) 0.4 -14.6693 dct 0.5 -14.6705 578 0.5 -14.6705 h-3 0.4 -14.6717 wn-1f 0.4 -14.6717 4.03 0.4 -14.6717 p(v 0.4 -14.6717 durch 0.0787879 -14.6717 distribution. 0.4 -14.673 0.563 0.4 -14.673 50000 0.4 -14.673 bhutto 0.333333 -14.6736 rol 0.4 -14.6742 fori 0.4 -14.6742 tw(g 0.4 -14.6742 10this 0.4 -14.6742 cilibrasi 0.4 -14.6742 1410 0.4 -14.6742 t,h summationdisplay 0.076087 -14.6758 using this 0.272727 -14.6758 plausible. 0.5 -14.6766 interesting: 0.5 -14.6766 subconcept 0.5 -14.6766 crouch 0.5 -14.6766 buchholz, 0.4 -14.6767 75.47 0.4 -14.6767 (e1) 0.4 -14.6767 ldc2003e07, 0.4 -14.6767 rhythmic 0.4 -14.6767 capture, 0.4 -14.6779 rui 0.0666667 -14.678 the results are 0.5 -14.6782 mathew 0.5 -14.6782 spanmap 0.4 -14.6791 bt45 0.4 -14.6791 kahn 0.4 -14.6791 spanish, french, german 0.134615 -14.6799 subject. 0.10101 -14.6799 items. 0.4 -14.6804 lil 0.4 -14.6816 0.814 0.4 -14.6816 rigau, 0.4 -14.6816 prov 0.4 -14.6816 mnb/pool 0.4 -14.6816 restriction: 0.5 -14.6828 cj: 0.4 -14.6829 heated 0.4 -14.6829 semiparametric 0.3 -14.6832 of monte 0.4 -14.6841 latbleu 0.4 -14.6841 swapna 0.5 -14.6844 si1 0.5 -14.6844 monte cristo 0.192308 -14.6849 yamada 0.3 -14.6851 p@5 0.4 -14.6854 that” 0.5 -14.6859 new←[] 0.0863309 -14.6863 value. 0.4 -14.6866 2.08 0.4 -14.6866 sbarq 0.4 -14.6866 richardson 0.272727 -14.6868 (kipper 0.5 -14.6875 τkl 0.5 -14.6875 ajar 0.4 -14.6878 yajuan 0.171429 -14.6889 word distribution 0.5 -14.689 n-grams: 0.4 -14.6891 1892 0.4 -14.6891 cold, 0.272727 -14.6902 10%, 0.4 -14.6916 arg2-sou 0.5 -14.6922 willen 0.4 -14.6929 rakuten 0.4 -14.6929 0.649 0.4 -14.6929 (e, 0.5 -14.6937 545 0.5 -14.6937 finalreward: 0.5 -14.6937 carolina, 0.166667 -14.6938 systems trained 0.4 -14.6941 ry 0.4 -14.6954 (mcnemar, 0.178571 -14.6964 nakov 0.4 -14.6966 per-token 0.4 -14.6966 gradients, 0.5 -14.6969 dipankar 0.272727 -14.6971 /> 0.056962 -14.6978 july 0.4 -14.6979 0.707 0.4 -14.6979 rom(x,bos) 0.5 -14.6985 1’s 0.5 -14.6985 lassen 0.272727 -14.6988 dragomir 0.5 -14.7 eugene 0.0961538 -14.7003 suited 0.4 -14.7004 xiaohua 0.5 -14.7016 potential arcs 0.4 -14.7029 notified 0.5 -14.7032 dis 0.109756 -14.7034 be added 0.222222 -14.7038 3.1 data 0.272727 -14.704 congressional 0.4 -14.7042 䕛 0.5 -14.7048 (none), 0.5 -14.7048 ‘semantically 0.5 -14.7048 hybrid: 0.5 -14.7048 yellowstone 0.5 -14.7048 (7): 0.5 -14.7064 rain, 0.5 -14.7064 stick-breaking 0.4 -14.7068 0.4 -14.708 diabetes 0.5 -14.708 even, 0.4 -14.7093 z: 0.5 -14.7096 “positive 0.4 -14.7106 (unused) 0.4 -14.7106 wiki200 0.0678733 -14.7111 which was 0.25 -14.7112 highest-scored 0.4 -14.7119 configurational 0.272727 -14.7127 tag; 0.5 -14.7128 ′ln 0.153846 -14.7129 es-en 0.4 -14.7131 percolation 0.4 -14.7131 p7 0.153846 -14.7134 modifies 0.4 -14.7144 can't 0.333333 -14.7144 0.0005 0.5 -14.7144 439 0.5 -14.7144 original: 0.3 -14.7154 (higashinaka 0.4 -14.7157 substantial, 0.4 -14.7157 interchangeability 0.4 -14.7157 onomatopoeic 0.4 -14.7157 λ) 0.0851064 -14.7158 give a 0.5 -14.716 thatthe 0.185185 -14.7161 negative sentiment 0.4 -14.717 mitre 0.192308 -14.7172 8 shows 0.5 -14.7176 1.81 0.3 -14.7183 __ 0.3 -14.7183 maxima 0.4 -14.7183 jean 0.0241379 -14.7186 use of 0.5 -14.7192 475 0.5 -14.7192 query-rand 0.115942 -14.7193 resource. 0.4 -14.7196 0.545 0.4 -14.7196 token: 0.0480549 -14.7196 a novel 0.4 -14.7208 servers 0.25 -14.7208 valletta, 0.0909091 -14.7208 bag 0.5 -14.7208 aur 0.5 -14.7208 scales, “happy 0.272727 -14.7214 constructive 0.038864 -14.7215 june 0.4 -14.7221 55.45 0.272727 -14.7223 ˆpnull 0.5 -14.7224 indyk 0.5 -14.724 “down” 0.333333 -14.7251 landis 0.5 -14.7256 lloyd, 0.4 -14.726 patchan 0.4 -14.726 (adjectives 0.4 -14.726 basket 0.4 -14.7273 ist, 0.4 -14.7273 74.56 0.4 -14.7273 cm, 0.5 -14.7273 ”, 1 -14.7536 crater 1 -14.7536 finally,wehaveshownthisimproves 1 -14.7536 ct: 1 -14.7536 ”asl“ 1 -14.7536 bsettles/abner/) 1 -14.7536 tlds 1 -14.7536 non-questions. 1 -14.7536 adequatelyreflectsimilarityinmeaning. 1 -14.7536 microaverage 1 -14.7536 succes-processing 1 -14.7536 wethenusedtheselectedfeaturesetandhyperparameters 1 -14.7536 hmm-lda, 1 -14.7536 backtracker’s 1 -14.7536 late? 1 -14.7536 decodingiswitharatherslowstochastic 1 -14.7536 singleton. 1 -14.7536 binning: 1 -14.7536 3http://jung.sourceforge.net/ 1 -14.7536 setycan 1 -14.7536 “projectable” 1 -14.7536 “query-focused 1 -14.7536 patt. 1 -14.7536 enju6 1 -14.7536 scopefinder, 1 -14.7536 2news 1 -14.7536 post-edition, 1 -14.7536 meantcanberunusinginexpensiveuntrainedmonolinguals 1 -14.7536 nist’06 1 -14.7536 mislead.v 1 -14.7536 do553 1 -14.7536 segmentationfor 1 -14.7536 =70.0, 1 -14.7536 iniseed 1 -14.7536 tool-kit 1 -14.7536 (ori.) 1 -14.7536 themostrelevant 1 -14.7536 •cu-2b 1 -14.7536 intranscriber 1 -14.7536 1.1.3 1 -14.7536 semdial 1 -14.7536 uating 1 -14.7536 2·mpur·accmpur 1 -14.7536 samreen 1 -14.7536 hitspos 1 -14.7536 predictor). 1 -14.7536 emdc, 1 -14.7536 (zaslavskiy 1 -14.7536 moreparsetreesmay 1 -14.7536 mereology, 1 -14.7536 tuplef 1 -14.7536 3,526 1 -14.7536 bcbs 1 -14.7536 familynameof; 1 -14.7536 skip/repeat 1 -14.7536 themodelissymmetric: 1 -14.7536 iula-universitat 1 -14.7536 baeza-yates, 1 -14.7536 θ(vinullk+1,...,vi) 1 -14.7536 thereisonlyonemorphologicalfeatureper 1 -14.7536 correc-939 1 -14.7536 1document 1 -14.7536 all-on/all-off 1 -14.7536 93.44%13. 1 -14.7536 32–37, 1 -14.7536 “@user” 1 -14.7536 utilitarian 1 -14.7536 (nullinullnullj) 1 -14.7536 m2-m4 1 -14.7536 typicallyaddsdependenciesamong 1 -14.7536 jj))))))) 1 -14.7536 58,000 1 -14.7536 1486–1495, 1 -14.7536 wegenerallyhandleseveraldifferenttypesoffeatures 1 -14.7536 33–37, 1 -14.7536 question-add 1 -14.7536 (gota 1 -14.7536 premloc: 1 -14.7536 emwith2-gramlm 1 -14.7536 49—57, 1 -14.7536 relevance: 1 -14.7536 610–619. 1 -14.7536 (leftmost) 1 -14.7536 nail, 1 -14.7536 ms,n 1 -14.7536 theseextrafeaturesassesstranslationqualitypast 1 -14.7536 troutman 1 -14.7536 integration) 1 -14.7536 wikipedia2), 1 -14.7536 (203), 1 -14.7536 chun-jen 1 -14.7536 {xi,yi,i 1 -14.7536 there1for 1 -14.7536 (0.122), 1 -14.7536 attemptat 1 -14.7536 peircian 1 -14.7536 171–176, 1 -14.7536 10again, 1 -14.7536 comicdialoguesystem. 1 -14.7536 (ate 1 -14.7536 stories/metaphors 1 -14.7536 generic) 1 -14.7536 acl/afnlp’09. 1 -14.7536 action-states, 1 -14.7536 330–336, 1 -14.7536 dtds 1 -14.7536 1436 1 -14.7536 icwsm’2007, 1 -14.7536 selectively; 1 -14.7536 4selectional 1 -14.7536 grefenstetteandsadrzadeh(2011)thatrelationalwordvec- 1 -14.7536 ric 1 -14.7536 sentence.”] 1 -14.7536 http:/uima.apache.org. 1 -14.7536 colt’98. 1 -14.7536 ak-best 1 -14.7536 nullrnull 1 -14.7536 trans5: 1 -14.7536 72–80, 1 -14.7536 medline® 1 -14.7536 null-calculus, 1 -14.7536 zs∈ys, 1 -14.7536 lp0883416 1 -14.7536 (lang´e, 1 -14.7536 ((a) 1 -14.7536 i∈{1,...,nd} 1 -14.7536 interestingly,heilman 1 -14.7536 transfering 1 -14.7536 word-type 1 -14.7536 tieindicates 1 -14.7536 opennlp1 1 -14.7536 46st 1 -14.7536 19(2):313–30. 1 -14.7536 jiri@us.ibm.com 1 -14.7536 13141516 1 -14.7536 hashimoto, 1 -14.7536 proximal-gradient 1 -14.7536 allresultspresentedhere,wereorderthetrainingand 1 -14.7536 graphbasedandtransition-basedsystemstendtomakedif- 1 -14.7536 {conj} 1 -14.7536 word-abbreviation 1 -14.7536 taku/software/cabocha/ 1 -14.7536 (xbleu) 1 -14.7536 grandparent-grandchild 1 -14.7536 costner 1 -14.7536 implications: 1 -14.7536 “states 1 -14.7536 μ 1 -14.7536 in§4, 1 -14.7536 97.40 1 -14.7536 93.1% 1 -14.7536 theoretically,mstparser 1 -14.7536 rescaled, 1 -14.7536 as human beings we use different ways to express  1 -14.7536 (2003a), 1 -14.7536 thecategoriesareusedatthesamelevel 1 -14.7536 def-cloth-pl-obj-acc 1 -14.7536 6.3.3 1 -14.7536 guage, 1 -14.7536 seventh, 1 -14.7536 maider 1 -14.7536 1032(177) 1 -14.7536 13malakasiotis 1 -14.7536 neitheroftheindividualwordpair(e.g., 1 -14.7536 248–258, 1 -14.7536 non� 1 -14.7536 brewed/*drank 1 -14.7536 demyristoylation) 1 -14.7536 “mnb/pool” 1 -14.7536 ‘neutral’ 1 -14.7536 extendnon-negativematrixtri-factorizationtoactive 1 -14.7536 thresholdk 1 -14.7536 word-constraint 1 -14.7536 80.32 1 -14.7536 “tweet,” 1 -14.7536 2.2insightsfromnon-verbaltaskswitching 1 -14.7536 (rebase) 1 -14.7536 lemmata, 1 -14.7536 apadapted 1 -14.7536 decoderisparameterized,andestimatingtheoptimal 1 -14.7536 vec.rae0.6 1 -14.7536 tesla-f3 1 -14.7536 perpetr 1 -14.7536 l⊆tσ 1 -14.7536 clue, 1 -14.7536 timeparsingispossibleforanyfixedlcfrsgrammar,butthedegreeofthepolynomial 1 -14.7536 “product” 1 -14.7536 viterbi-optimal 1 -14.7536 8.2.2 1 -14.7536 ‘pp’ 1 -14.7536 however,thescoresoftypebased 1 -14.7536 .10). 1 -14.7536 93–101, 1 -14.7536 node/concept 1 -14.7536 d59 1 -14.7536 spe1we 1 -14.7536 inaddition,thebest-performingfaust 1 -14.7536 pp.137-140. 1 -14.7536 3note, 1 -14.7536 miler. 1 -14.7536 qh 1 -14.7536 understudied, 1 -14.7536 setexpansion 1 -14.7536 weused 1 -14.7536 feats(i) 1 -14.7536 nl+nlp: 1 -14.7536 (fine-grained) 1 -14.7536 447–451, 1 -14.7536 “singapore”) 1 -14.7536 indexiis 1 -14.7536 1http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalog 1 -14.7536 eachdocument 1 -14.7536 austin: 1 -14.7536 (confirm:from(fl,atl)) 1 -14.7536 pn) 1 -14.7536 piwowar 1 -14.7536 ity, 1 -14.7536 91.1%. 1 -14.7536 (dias, 1 -14.7536 rsynth: 1 -14.7536 toachieveextensibility,weneedtorestricttheinformation 1 -14.7536 hand,it 1 -14.7536 6,we 1 -14.7536 near-paretooptimal 1 -14.7536 scf+pp) 1 -14.7536 81.71 1 -14.7536 x·,j 1 -14.7536 m.o.dzikovska 1 -14.7536 alignmentsmt(at) 1 -14.7536 stereotypically, 1 -14.7536 311-318. 1 -14.7536 human-likeness, 1 -14.7536 92–96, 1 -14.7536 him); 1 -14.7536 262–272, 1 -14.7536 key-list, 1 -14.7536 emnlp’03. 1 -14.7536 4.1.4 1 -14.7536 accuracies/f-scores 1 -14.7536 log-odds-ratio 1 -14.7536 two-set 1 -14.7536 thierry.bazillon@univ-avignon.fr 1 -14.7536 speeds-up 1 -14.7536 kylin 1 -14.7536 pseudo-word’s 1 -14.7536 letqsb 1 -14.7536 ¯xn)}are 1 -14.7536 1346–1355, 1 -14.7536 isavalidtreedecompositionofg 1 -14.7536 tag-consistency: 1 -14.7536 703–711, 1 -14.7536 hexateuch: 1 -14.7536 3not 1 -14.7536 706 1 -14.7536 gorithms 1 -14.7536 timemachine, 1 -14.7536 zhangetal.(2007a))canpotentiallyovercomesome 1 -14.7536 matchingskipsteadilyworsened 1 -14.7536 wordnet3 1 -14.7536 (nrc). 1 -14.7536 thetranslationand 1 -14.7536 macr, 1 -14.7536 table4.the 1 -14.7536 wordnet’smanuallydefinedcomprehensivetaxonomymotivatesmany 1 -14.7536 describable 1 -14.7536 6b, 1 -14.7536 berstein 1 -14.7536 29(1) 1 -14.7536 (*some/most/all) 1 -14.7536 1600–1610, 1 -14.7536 generation/typing 1 -14.7536 amts 1 -14.7536 machines: 1 -14.7536 sive 1 -14.7536 22(1). 1 -14.7536 community.4 1 -14.7536 #374: 1 -14.7536 homogeneitydirectly 1 -14.7536 57–64, 1 -14.7536 external: 1 -14.7536 mulation 1 -14.7536 “growdiag” 1 -14.7536 emiel 1 -14.7536 pcfg-rules 1 -14.7536 cleans 1 -14.7536 tine’s 1 -14.7536 88.88 1 -14.7536 propertyattribute 1 -14.7536 herbivores), 1 -14.7536 [experiment-ii: 1 -14.7536 mrf’s 1 -14.7536 hecht, 1 -14.7536 p-ac 1 -14.7536 query/summary 1 -14.7536 sense-word 1 -14.7536 sistently 1 -14.7536 obtained8,677 1 -14.7536 14implying 1 -14.7536 sentence.” 1 -14.7536 database-based 1 -14.7536 org/gpe 1 -14.7536 sherwood 1 -14.7536 voll, 1 -14.7536 picked. 1 -14.7536 s3hf, 1 -14.7536 datasets) 1 -14.7536 short: 1 -14.7536 (subset) 1 -14.7536 28-34, 1 -14.7536 serena’s 1 -14.7536 9one 1 -14.7536 runigrambaswdp 1 -14.7536 reva 1 -14.7536 markov-chain 1 -14.7536 61-69, 1 -14.7536 ∗−→ 1 -14.7536 efectiveness 1 -14.7536 itly 1 -14.7536 “recap” 1 -14.7536 7-8, 1 -14.7536 1023–1033, 1 -14.7536 6http://www.psy.uwa.edu.au/mrcdatabase/uwa 1 -14.7536 (17d), 1 -14.7536 thisislikelydueinpartto 1 -14.7536 (i>0) 1 -14.7536 (btg, 1 -14.7536 d(z=k)i 1 -14.7536 l[wi] 1 -14.7536 thispaperhaspresentedalanguage-specificmethod 1 -14.7536 ours; 1 -14.7536 t1∩t2 1 -14.7536 datapoints 1 -14.7536 karrowbothv 1 -14.7536 home/bacteria-biotopes 1 -14.7536 mwetoolkit1 1 -14.7536 yoshimi 1 -14.7536 571–582, 1 -14.7536 word-selection 1 -14.7536 ciation 1 -14.7536 ‘words 1 -14.7536 depen5we 1 -14.7536 monolingual-filtering 1 -14.7536 itistheorganizationofspatialrelationships-(verbandpreposition)andtheperspective 1 -14.7536 (grow-diag-final 1 -14.7536 approachindickinsonandmeurers(2003b)hascertain 1 -14.7536 φ(cqk,o(cqk)j 1 -14.7536 targeting. 1 -14.7536 non-combinator 1 -14.7536 portionof 1 -14.7536 tf: 1 -14.7536 baldewein 1 -14.7536 cur 1 -14.7536 36(3):303-339. 1 -14.7536 0.506 1 -14.7536 totol 1 -14.7536 relabel-arc 1 -14.7536 sub-topics, 1 -14.7536 unknownwords 1 -14.7536 4-post 1 -14.7536 there864 1 -14.7536 pre* 1 -14.7536 rea24 1 -14.7536 thegold-standarddatasetisevenlysplitinto 1 -14.7536 dung 1 -14.7536 thetaskofquantificationresolutioninvolvesfindingasemanticrepresentationthatgoesbeyondwhat 1 -14.7536 flow: 1 -14.7536 msdbidirectional-fe 1 -14.7536 lemmatizing, 1 -14.7536 ge-tasks, 1 -14.7536 provoking 1 -14.7536 biadsy 1 -14.7536 anotheradvantageisthatsince 1 -14.7536 1611–1622, 1 -14.7536 streambased 1 -14.7536 5.2.5 1 -14.7536 earlier,we 1 -14.7536 4.2.8 1 -14.7536 featurefunctionsareoften 1 -14.7536 7.2.2 1 -14.7536 (l1language) 1 -14.7536 persentence 1 -14.7536 wenotethatourmodeloutperforms 1 -14.7536 pinócrates); 1 -14.7536 co-clustering, 1 -14.7536 modeling: 1 -14.7536 arcposfeat, 1 -14.7536 regulariza338 1 -14.7536 ,1997),cder(leusch 1 -14.7536 mbrmax 1 -14.7536 words(b): 1 -14.7536 meeting.), 1 -14.7536 *accounts 1 -14.7536 “tuesday 1 -14.7536 1≤k≤d, 1 -14.7536 cross-word 1 -14.7536 misguided, 1 -14.7536 ∈{t}, 1 -14.7536 highestorder 1 -14.7536 ibm-method 1 -14.7536 take䌬‘hall’ 1 -14.7536 edmonton 1 -14.7536 en->ch 1 -14.7536 :...that 1 -14.7536 pair-based 1 -14.7536 dat, 1 -14.7536 1,792 1 -14.7536 62–71, 1 -14.7536 sub-plot 1 -14.7536 +2.6 1 -14.7536 (usually) 1 -14.7536 646-655. 1 -14.7536 weleaveoutthedetailshereandfocusonparameter 1 -14.7536 reftl: 1 -14.7536 believing 1 -14.7536 4kolmogorov-smirnov 1 -14.7536 coreferenceresolution(core)istheprocessoffinding 1 -14.7536 heartfelt 1 -14.7536 well-calibrated 1 -14.7536 enlisting 1 -14.7536 pcfgθ(1),...,θ(l), 1 -14.7536 shownthatlinguistically-motivatedsyntacticstate-splittingalreadyimprovesparses(kleinandmanning, 1 -14.7536 70.3% 1 -14.7536 huberman 1 -14.7536 acl/ijcnlp’09. 1 -14.7536 non-parameterized 1 -14.7536 glickman 1 -14.7536 183–192, 1 -14.7536 geneexpression,localization,phosphorylation 1 -14.7536 1158–1167, 1 -14.7536 vectors? 1 -14.7536 friberg 1 -14.7536 occurrence-indexed 1 -14.7536 which(ideally)aregeneralandreusable 1 -14.7536 scisumm: 1 -14.7536 maxenttagger 1 -14.7536 (2002b) 1 -14.7536 ∼700 1 -14.7536 methodology:2 1 -14.7536 9.61% 1 -14.7536 thevariabilityalsoextendstotheutility 1 -14.7536 autobiography 1 -14.7536 “mcd” 1 -14.7536 prolog-based 1 -14.7536 5please 1 -14.7536 ultimately,this 1 -14.7536 (“the” 1 -14.7536 chineseusersmaymakeerrorswhentheyaretyping 1 -14.7536 special-title 1 -14.7536 dean, 1 -14.7536 (acl-99). 1 -14.7536 “head-centric” 1 -14.7536 (1h) 1 -14.7536 6338 1 -14.7536 top-center 1 -14.7536 distributions? 1 -14.7536 1464 1 -14.7536 dialectologists: 1 -14.7536 retrospect,this 1 -14.7536 thar 1 -14.7536 card({(i,j)∈a1}) 1 -14.7536 r1,...,r99 1 -14.7536 “only-if” 1 -14.7536 theexamplesaboverequiredeeperunderstanding 1 -14.7536 discussionandconclusion 1 -14.7536 multiple-level 1 -14.7536 tutor-topic 1 -14.7536 “xx-en” 1 -14.7536 additionwe 1 -14.7536 1http://statmt.org/wmt11/results.html 1 -14.7536 scopefiltering 1 -14.7536 rms: 1 -14.7536 mightreasonablyspeculatethatater-tunedsystem 1 -14.7536 thisobservationsuggestsadifferent 1 -14.7536 ‘inc-carbonize’ 1 -14.7536 5.3.3) 1 -14.7536 noλk 1 -14.7536 spm. 1 -14.7536 ldr 1 -14.7536 prestige 1 -14.7536 coling/acl. 1 -14.7536 ontheone 1 -14.7536 xij,1 1 -14.7536 non-interruptions 1 -14.7536 (roche 1 -14.7536 11.6) 1 -14.7536 6roget’s 1 -14.7536 theentailmentsassociatedwiththelatteraremarked 1 -14.7536 tv) 1 -14.7536 zànshí 1 -14.7536 canseethatsupervisedmethodssignificantlyoutperform 1 -14.7536 hypothesis,we 1 -14.7536 85–89, 1 -14.7536 thesystemwastested 1 -14.7536 ms: 1 -14.7536 sim(vw1⊕w2,vw1w2) 1 -14.7536 conll-09. 1 -14.7536 ferrucci 1 -14.7536 54.07% 1 -14.7536 followup, 1 -14.7536 experimental-set-up: 1 -14.7536 74.03 1 -14.7536 hurdle 1 -14.7536 <1773> 1 -14.7536 .004 1 -14.7536 (29)-(30) 1 -14.7536 29(1): 1 -14.7536 c(iprime−i) 1 -14.7536 chunklink7 1 -14.7536 iowa. 1 -14.7536 award-cat 1 -14.7536 (conll-x) 1 -14.7536 title-word 1 -14.7536 96%. 1 -14.7536 ‡lee 1 -14.7536 -lwdp 1 -14.7536 compare: 1 -14.7536 (rgb), 1 -14.7536 cd-based 1 -14.7536 disparity, 1 -14.7536 (roth, 1 -14.7536 goaltest(dependency 1 -14.7536 52.64%, 1 -14.7536 ularization 1 -14.7536 tree-cut 1 -14.7536 language9 1 -14.7536 inordertoconstrain 1 -14.7536 “sense” 1 -14.7536 70.94, 1 -14.7536 hulth 1 -14.7536 gaz-il) 1 -14.7536 accountability: 1 -14.7536 3http://www.ceet.eu/ 1 -14.7536 (yin 1 -14.7536 iis0968450; 1 -14.7536 meansthatlongerdifferenceinscoreswillbeneeded 1 -14.7536 pittsburgh”, 1 -14.7536 thedomainselectorgathers 1 -14.7536 cross-linguistically, 1 -14.7536 multir, 1 -14.7536 phemes 1 -14.7536 nately, 1 -14.7536 slack: 1 -14.7536 oursystemidentified9,984 1 -14.7536 cultural/situational: 1 -14.7536 dberkeleylm 1 -14.7536 asnotedintheintroduction,informationgraphics 1 -14.7536 16please 1 -14.7536 whatthetrend 1 -14.7536 christians 1 -14.7536 large), 1 -14.7536 thisunderspecifiedsemanticformalismwasdesigned 1 -14.7536 mstep, 1 -14.7536 translated-from-source-language 1 -14.7536 1320–1331, 1 -14.7536 allthreesetsoffeaturescontributetoperformanceof 1 -14.7536 scorecard 1 -14.7536 61.1%. 1 -14.7536 significant? 1 -14.7536 summationtextni 1 -14.7536 pling 1 -14.7536 b09 1 -14.7536 thiscomputation 1 -14.7536 crowdflower2 1 -14.7536 fi(c,s,t) 1 -14.7536 fpd 1 -14.7536 (§56) 1 -14.7536 cathcart 1 -14.7536 blog-derived 1 -14.7536 furthemore, 1 -14.7536 iftheipexpertclass 1 -14.7536 inordertoobtainareliablekbwithasufficientcoveragewehaveexploited 1 -14.7536 fay 1 -14.7536 topranking 1 -14.7536 modificationstothealgorithm 1 -14.7536 strates 1 -14.7536 icml’05. 1 -14.7536 joint: 1 -14.7536 45%-55% 1 -14.7536 10uncertainty 1 -14.7536 weonlyusedthetransfer 1 -14.7536 kers 1 -14.7536 extraocular 1 -14.7536 344–350, 1 -14.7536 orthogonally, 1 -14.7536 5ensuring 1 -14.7536 o(d) 1 -14.7536 (incongruent 1 -14.7536 ctb7, 1 -14.7536 tmphr 1 -14.7536 set,then 1 -14.7536 standardly, 1 -14.7536 f(•) 1 -14.7536 page=xconc+suite 1 -14.7536 geo-tagged 1 -14.7536 11results 1 -14.7536 uni-modular 1 -14.7536 201407/2010-8) 1 -14.7536 #1) 1 -14.7536 trec. 1 -14.7536 berkeleyaligner4 1 -14.7536 605–615, 1 -14.7536 refsum: 1 -14.7536 affixations 1 -14.7536 ‘-accurate’ 1 -14.7536 hypothyroidism. 1 -14.7536 dict, 1 -14.7536 arguement, 1 -14.7536 likecohen 1 -14.7536 z-updates. 1 -14.7536 wu(q,u) 1 -14.7536 trilions 1 -14.7536 11.8%, 1 -14.7536 incontrast,ceafe 1 -14.7536 activationi(s) 1 -14.7536 18recall 1 -14.7536 beesley 1 -14.7536 webb 1 -14.7536 [0,m]n, 1 -14.7536 (scored) 1 -14.7536 swier, 1 -14.7536 textrunner-r, 1 -14.7536 295,026 1 -14.7536 naively: 1 -14.7536 labelsthatalignsimilarlybutappearinverydifferent 1 -14.7536 text/permutation 1 -14.7536 tsang, 1 -14.7536 md-ii, 1 -14.7536 despitethelargenumberofwordsensedisambiguationmethodsthathavebeenproposedsofar, 1 -14.7536 reorderingspecific 1 -14.7536 56.13) 1 -14.7536 grafting-10 1 -14.7536 example,measuring 1 -14.7536 (twa) 1 -14.7536 english↔french, 1 -14.7536 ‘aug’ 1 -14.7536 2.5.4 1 -14.7536 λ∈[1,10] 1 -14.7536 1332–1341, 1 -14.7536 so... 1 -14.7536 ledir: 1 -14.7536 6www.statmt.org/wmt10/translation-task.html 1 -14.7536 1258–1267, 1 -14.7536 quasi-arithmetical 1 -14.7536 creased 1 -14.7536 self-selected 1 -14.7536 (2005a)): 1 -14.7536 salience-based 1 -14.7536 heger 1 -14.7536 bedrock 1 -14.7536 stereotypes-in-a-neighborhood: 1 -14.7536 7.1.2 1 -14.7536 (tsi-020110-2009439) 1 -14.7536 lookingupawordorphraseon 1 -14.7536 non-disjunctive) 1 -14.7536 meseq 1 -14.7536 implementation’s 1 -14.7536 ruption(int)andpoker-playing(pkr)ofeachplayer,withtheactualvaluesdisplayed 1 -14.7536 12see 1 -14.7536 perplex3note 1 -14.7536 apartfromtheofficiallyannouncedscoringmethods, 1 -14.7536 word-recognition 1 -14.7536 fool 1 -14.7536 “speed” 1 -14.7536 1,φ(b 1 -14.7536 λt, 1 -14.7536 distributionally) 1 -14.7536 11c, 1 -14.7536 θu 1 -14.7536 acl-97, 1 -14.7536 ‘≈’ 1 -14.7536 maebashi 1 -14.7536 (ptb-style) 1 -14.7536 sogou 1 -14.7536 1109–1116, 1 -14.7536 98.59 1 -14.7536 emnlp’06. 1 -14.7536 theyrequiresomeexternalprocesstoaggregatethemeaningrepresentations 1 -14.7536 8that 1 -14.7536 twitter.com) 1 -14.7536 0.760.530.850.640.900.72average 1 -14.7536 billingual 1 -14.7536 inthisexample, 1 -14.7536 bthe 1 -14.7536 contiguity: 1 -14.7536 “feature-engineering” 1 -14.7536 pnn 1 -14.7536 1427 1 -14.7536 conciseness, 1 -14.7536 tiltman 1 -14.7536 paleographic 1 -14.7536 ‘pperson’) 1 -14.7536 10.0)* 1 -14.7536 essientially 1 -14.7536 eachantecedentoftheoriginalrulemustappearinsomenewruleinthefactorization, 1 -14.7536 323–329, 1 -14.7536 proposeaneffectivepostagpruningmethodwhichcan 1 -14.7536 theprobabilityofsyntaxgiveneithersubsumptiveor 1 -14.7536 company:profitable 1 -14.7536 obstacles, 1 -14.7536 moderate-sized 1 -14.7536 glass. 1 -14.7536 haitiancreole-to-english 1 -14.7536 (u,v)∈v1×v1 1 -14.7536 german→czech 1 -14.7536 ascase 1 -14.7536 350–352, 1 -14.7536 j(d) 1 -14.7536 miner3 1 -14.7536 16–20, 1 -14.7536 vq: 1 -14.7536 7(a) 1 -14.7536 optimistically, 1 -14.7536 inferentially 1 -14.7536 complex/simplified 1 -14.7536 rection 1 -14.7536 six-dot 1 -14.7536 orders-ofmagnitude 1 -14.7536 utterance-label 1 -14.7536 .680 1 -14.7536 94.16 1 -14.7536 customize/ 1 -14.7536 finally,sincementionsthatareinsingleton 1 -14.7536 whileforthiswork 1 -14.7536 non-isomorphism, 1 -14.7536 wealsoanalyzedthecontributionofdifferentdiscourse 1 -14.7536 fc(si,ts)=+ 1 -14.7536 j¨orgensen 1 -14.7536 (princeton 1 -14.7536 pergo 1 -14.7536 pleonastics 1 -14.7536 although: 1 -14.7536 add-on 1 -14.7536 0.797, 1 -14.7536 forgi 1 -14.7536 m34 1 -14.7536 livers 1 -14.7536 henceconfirmisaspecialization 1 -14.7536 86.7%. 1 -14.7536 distancedependent 1 -14.7536 languagespecific, 1 -14.7536 73.13 1 -14.7536 thismaybetosomeextentdependent 1 -14.7536 dunning 1 -14.7536 one-sense-per-x 1 -14.7536 fmeasure) 1 -14.7536 z1···zl 1 -14.7536 -srp 1 -14.7536 categories-map 1 -14.7536 .077). 1 -14.7536 1506–1515, 1 -14.7536 113207. 1 -14.7536 general-purpose, 1 -14.7536 set,andtraintheclassifierusingtheunionsetofgold 1 -14.7536 analysis:1 1 -14.7536 s-pattern 1 -14.7536 correcty(xi) 1 -14.7536 csp, 1 -14.7536 error-agnostic 1 -14.7536 (kalinli 1 -14.7536 vergence 1 -14.7536 bleu9 1 -14.7536 1http://gate.ac.uk/sale/tao/splitch15.html 1 -14.7536 weconsideravariantofthenon-parametricptsg 1 -14.7536 lf(y) 1 -14.7536 6-7), 1 -14.7536 c:implementing 1 -14.7536 contexonyms. 1 -14.7536 pastsuccesseswithconversationalintelligenttutoringsystems(its)(graesseretal.,2001), 1 -14.7536 “sl-full” 1 -14.7536 kyoto-kybot 1 -14.7536 settings.onesettingconsistsof 1 -14.7536 stringlevel 1 -14.7536 ’•’ 1 -14.7536 iob5 1 -14.7536 1 -14.7536 rom(x,y) 1 -14.7536 7both 1 -14.7536 89–94, 1 -14.7536 22–64, 1 -14.7536 theftbsentencesare 1 -14.7536 imbue 1 -14.7536 thereadability 1 -14.7536 .007 1 -14.7536 stopword-list 1 -14.7536 φj(xi) 1 -14.7536 tsg-based 1 -14.7536 conor 1 -14.7536 softclustering 1 -14.7536 overgeneralize, 1 -14.7536 6,620 1 -14.7536 small-change 1 -14.7536 #ttiv 1 -14.7536 columntriangleshows 1 -14.7536 pdftotext 1 -14.7536 theannotatorsstartedfromlocatingopinionexpressions, 1 -14.7536 unecesarily 1 -14.7536 (pet06); 1 -14.7536 criticisms, 1 -14.7536 sorgandcimiano(2008) 1 -14.7536 over-the-counter 1 -14.7536 odels 1 -14.7536 proposer 1 -14.7536 qualitycontrol 1 -14.7536 (kwork 1 -14.7536 furthurmore, 1 -14.7536 donot-resolve-non-anaphors 1 -14.7536 geppetto 1 -14.7536 chaumartin, 1 -14.7536 izing 1 -14.7536 http://www.personal.psu.edu/xxl13/downloads/l2sca. 1 -14.7536 fully-anonymized 1 -14.7536 itaccessestwoknowledgebases, 1 -14.7536 (31m) 1 -14.7536 (articles), 1 -14.7536 tied, 1 -14.7536 were-sample 1 -14.7536 20corpus 1 -14.7536 64–71, 1 -14.7536 1prior 1 -14.7536 wmtyy(t) 1 -14.7536 let’sgeneralize 1 -14.7536 lenzi, 1 -14.7536 ofthembeinggeneralcontextfeaturesthatmightnot 1 -14.7536 31-35, 1 -14.7536 1395–1404, 1 -14.7536 syntacticparsing,whereinmostcasesparsingisperformed 1 -14.7536 bl)ip 1 -14.7536 (granduncle) 1 -14.7536  1 -14.7536 v1(p),p1(p),v2(p),p2(p),v3(p) 1 -14.7536 nclogp: 1 -14.7536 hand-tuned. 1 -14.7536 inthefollowingwewillfirstintroducebiomedical 1 -14.7536 *note 1 -14.7536 704 1 -14.7536 4,334 1 -14.7536 142–150, 1 -14.7536 hehe 1 -14.7536 genres; 1 -14.7536 895–904, 1 -14.7536 wiki-related 1 -14.7536 g46661g4667 1 -14.7536 934–944, 1 -14.7536 illustates 1 -14.7536 whilehwa 1 -14.7536 nk−1 1 -14.7536 http://github.com/mjpost/dptsg. 1 -14.7536 (mfc), 1 -14.7536 ud’s 1 -14.7536 oranchak 1 -14.7536 320–330, 1 -14.7536 59.682.2 1 -14.7536 “correct 1 -14.7536 (canonical 1 -14.7536 theconsequenceofsuchannotationdiscrepanciesis 1 -14.7536 weightsλk 1 -14.7536 embased 1 -14.7536 expert-dependent. 1 -14.7536 esplored 1 -14.7536 {600,800,1000,1200} 1 -14.7536 non-logical 1 -14.7536 random-text, 1 -14.7536 alsothankderrickhiggins 1 -14.7536 word-only 1 -14.7536 cwcd6 1 -14.7536 toy-naming 1 -14.7536 self-consistency) 1 -14.7536 slop. 1 -14.7536 1056 1 -14.7536 web-derivedselectionalpreferencehasnotbeensuccessfully 1 -14.7536 interleaved-uncertainty 1 -14.7536 3hamming 1 -14.7536 case,fn 1 -14.7536 ney-essen 1 -14.7536 shos 1 -14.7536 higher-recall 1 -14.7536 ddts, 1 -14.7536 iis0915327 1 -14.7536 18(4):467–479. 1 -14.7536 proposedmethodefficientlycorrectstheinitial 1 -14.7536 scal1while 1 -14.7536 45m/65m/100m 1 -14.7536 @twitter), 1 -14.7536 thepercentageof“hits” 1 -14.7536 forwam,weonlyshowtheresultsusingtitle-based 1 -14.7536 2less 1 -14.7536 (2010))) 1 -14.7536 scf) 1 -14.7536 decrypt), 1 -14.7536 iis-0968485. 1 -14.7536 vant 1 -14.7536 joshua: 1 -14.7536 transliteration-transcriptionpairs, 1 -14.7536 avenues: 1 -14.7536 001₁₅₀ 1 -14.7536 [4,72] 1 -14.7536 opal 1 -14.7536 urges 1 -14.7536 1467–1478, 1 -14.7536 8,000), 1 -14.7536 classlabel-based 1 -14.7536 gues 1 -14.7536 label→term 1 -14.7536 mean7for 1 -14.7536 dancy), 1 -14.7536 bio-community 1 -14.7536 (case-insensitive) 1 -14.7536 8-dimensional 1 -14.7536 asim 1 -14.7536 frist 1 -14.7536 “non-fringe” 1 -14.7536 oftis 1 -14.7536 consideronlyderivationswithsynchronousspanswhich 1 -14.7536 (#pcdata)> 1 -14.7536 128–136, 1 -14.7536 plato's 1 -14.7536 http://reseach.yahoo.com/academic 1 -14.7536 response/quote/both 1 -14.7536 ...,...eos 1 -14.7536 φd, 1 -14.7536 (w1, 1 -14.7536 thing”, 1 -14.7536 kamal 1 -14.7536 82–88. 1 -14.7536 77–82, 1 -14.7536 “in-between” 1 -14.7536 cles 1 -14.7536 trolled 1 -14.7536 2http://sourceforge.net/projects/hrcf/ 1 -14.7536 23(3):377-403. 1 -14.7536 2008’s 1 -14.7536 mocian 1 -14.7536 91.6%; 1 -14.7536 99.41% 1 -14.7536 non-bottomup 1 -14.7536 sim(vlight,vtraffic 1 -14.7536 1034–1046, 1 -14.7536 sticked 1 -14.7536 4on 1 -14.7536 pos-taggers: 1 -14.7536 khadivi, 1 -14.7536 one-derivation 1 -14.7536 cora(sen 1 -14.7536 dt” 1 -14.7536 h(planckover2pi1(·)) 1 -14.7536 inpe 1 -14.7536 cau(x, 1 -14.7536 deictic: 1 -14.7536 daily) 1 -14.7536 43.03 1 -14.7536 precisionandrecallarecomputed 1 -14.7536 am×mmatrix, 1 -14.7536 4.1.5 1 -14.7536 weadoptedthecase-insensitivebleu4 1 -14.7536 vorhes, 1 -14.7536 cross-lingualsemantictransfer 1 -14.7536 ontolgy 1 -14.7536 sampled; 1 -14.7536 mostlydue 1 -14.7536 (sps) 1 -14.7536 dimensionality-reduction 1 -14.7536 part-of-speeches 1 -14.7536 tanen, 1 -14.7536 “self-training.” 1 -14.7536 (srcc) 1 -14.7536 (74%) 1 -14.7536 p3, 1 -14.7536 acl2007. 1 -14.7536 “e+p*+a*”evaluated 1 -14.7536 6gb 1 -14.7536 100∗(k−c)/k, 1 -14.7536 7indeed, 1 -14.7536 .006 1 -14.7536 burgoon 1 -14.7536 synchronouscontext-freegrammar(scfg),wherea 1 -14.7536 tenfold 1 -14.7536 hk:{x1...xn}→{1...w},1 1 -14.7536 documentpairedtestingforsummarizercomparison.ourmain 1 -14.7536 multi-classifier 1 -14.7536 havelexisting, 1 -14.7536 gen71 1 -14.7536 null0,1nullnullnullnull. 1 -14.7536 0.587) 1 -14.7536 9interpretation 1 -14.7536 wiki/wingbaldridge2011 1 -14.7536 theticketingalgorithm 1 -14.7536 discardfirstmentionsthatstartwithindefinitepronouns 1 -14.7536 markupfeaturesperformremarkablywell 1 -14.7536 9; 1 -14.7536 gleaning 1 -14.7536 214-222. 1 -14.7536 mahesh, 1 -14.7536 acquiescence, 1 -14.7536 theemergenceandchangeindifferenttypesofspatialreferencetophysicallocation(discourse 1 -14.7536 (global)informationalsoprovidesaproperguidance 1 -14.7536 nullv1,...,vknull 1 -14.7536 context-sensitive, 1 -14.7536 infringe 1 -14.7536 29,854 1 -14.7536 .07). 1 -14.7536 xor-with-output 1 -14.7536 rsemh: 1 -14.7536 figure3illustratesthe 1 -14.7536 kaffiles 1 -14.7536 highway...were 1 -14.7536 1990’s, 1 -14.7536 ,class2, 1 -14.7536 lrec-2006, 1 -14.7536 slovko. 1 -14.7536 csidm200805) 1 -14.7536 forestsusuallyencodebettercandidatesmore 1 -14.7536 ”)) 1 -14.7536 sub-process 1 -14.7536 6thestanfordparserisaproductmodel,sotheresultsin§5.1 1 -14.7536 pre-experiments, 1 -14.7536 headtable 1 -14.7536 topic/sense, 1 -14.7536 icwsm, 1 -14.7536 roe 1 -14.7536 traded-off 1 -14.7536 “consensus”, 1 -14.7536 lew-williams 1 -14.7536 becausetheclassificationproblemisusually 1 -14.7536 collective-learning 1 -14.7536 163–171, 1 -14.7536 theseareassignedtoeachwordandthencombinedusingcombinatorial 1 -14.7536 l1-transfer: 1 -14.7536 auxiliary-verbal 1 -14.7536 12to 1 -14.7536 strated 1 -14.7536 id-specific 1 -14.7536 variantwiththehighestsupportforthefinalconsensus 1 -14.7536 codings, 1 -14.7536 mostexperiments 1 -14.7536 bigrambinb(y) 1 -14.7536 givenwvia 1 -14.7536 38–46, 1 -14.7536 riaz 1 -14.7536 setting,we 1 -14.7536 dividual 1 -14.7536 otterbacher 1 -14.7536 (2002a) 1 -14.7536 thoughn 1 -14.7536 (17–22) 1 -14.7536 8(4):567-588. 1 -14.7536 thisrelianceonirrelevantmaterialarisesfromusing 1 -14.7536 +node-external 1 -14.7536 phonepass 1 -14.7536 cg-only 1 -14.7536 hand-write 1 -14.7536 scaleable 1 -14.7536 c3-t, 1 -14.7536 “sq”), 1 -14.7536 boss8 1 -14.7536 ‘uppercase’ 1 -14.7536 intentionallyadjustthemeaningofthreshold(x-axis)forboth 1 -14.7536 erarchical 1 -14.7536 b— 1 -14.7536 1069–1080, 1 -14.7536 tedbased 1 -14.7536 recognizementionboundariesonly 1 -14.7536 cleared), 1 -14.7536 (6-way 1 -14.7536 baseline-to-best 1 -14.7536 moses+del, 1 -14.7536 kedis: 1 -14.7536 yongyi 1 -14.7536 expecteddateonmarket(), 1 -14.7536 maltnullae 1 -14.7536 mixtureθ 1 -14.7536 12fate 1 -14.7536 moredetailsofthecorpus 1 -14.7536 double-circles 1 -14.7536 fine34 1 -14.7536 bensal 1 -14.7536 moilianen 1 -14.7536 1,160. 1 -14.7536 summarises, 1 -14.7536 automatictechniques 1 -14.7536 avgf2: 1 -14.7536 764-767, 1 -14.7536 wecompareourerrorratesto 1 -14.7536 complementarytechnique,increasingtheapplicabilityof 1 -14.7536 roadmap 1 -14.7536 (distinct) 1 -14.7536 wnfv 1 -14.7536 aaai-05 1 -14.7536 m30...4 1 -14.7536 326228 1 -14.7536 procdngs 1 -14.7536 “spaghetti” 1 -14.7536 53,958 1 -14.7536 “()” 1 -14.7536 tld; 1 -14.7536 bes 1 -14.7536 φt. 1 -14.7536 whiletri-nmfisfirstappliedinco-clustering,li 1 -14.7536 oftentimes, 1 -14.7536 supervised-learning 1 -14.7536 79–87, 1 -14.7536 theperformancedifferences 1 -14.7536 wordsets 1 -14.7536 thesetogethersuggestthatsuchconstructionscouldbesimultaneously 1 -14.7536 non-profit 1 -14.7536 104–107, 1 -14.7536 4.2.3, 1 -14.7536 anusaaraka: 1 -14.7536 jan., 1 -14.7536 wi+1...wi+n−1 1 -14.7536 trade-offs, 1 -14.7536 hm(cn−1,cn,wn+2n−2) 1 -14.7536 anne@limsi.fr 1 -14.7536 (katakana) 1 -14.7536 crosscorpus 1 -14.7536 instantiation’s 1 -14.7536 noise1 1 -14.7536 t-generalization, 1 -14.7536 naacl-hlt’07, 1 -14.7536 exploresamethodtobuildahigh-performance 1 -14.7536 reachable, 1 -14.7536 dizer: 1 -14.7536 singlespeaker 1 -14.7536 6due 1 -14.7536 inthecaseofthematrix-spaceword 1 -14.7536 tically 1 -14.7536 theinitialsegmentation 1 -14.7536 46–50, 1 -14.7536 coursebook 1 -14.7536 3hunter 1 -14.7536 non-sensical 1 -14.7536 φmapsto→ψ(p(pi)) 1 -14.7536 10.6%. 1 -14.7536 rerankingfeatures 1 -14.7536 politeness: 1 -14.7536 woszczyna, 1 -14.7536 rec.(%) 1 -14.7536 describeatimeseriesmodelappropriatefortimeseries 1 -14.7536 8934 1 -14.7536 discourse-tagged 1 -14.7536 fruitfulto 1 -14.7536 (saif 1 -14.7536 (43% 1 -14.7536 kathleen, 1 -14.7536 bundschus 1 -14.7536 row5isthephrasebased 1 -14.7536 313-321. 1 -14.7536 aninterestingperspectiveonthemodelsisprovided 1 -14.7536 2007)—has 1 -14.7536 yago: 1 -14.7536 s*: 1 -14.7536 phosphorylation) 1 -14.7536 teslaf, 1 -14.7536 apparentlytheerrorfullabels 1 -14.7536 moses-nogc, 1 -14.7536 dis10http://www.lemurproject.org/ 1 -14.7536 tuation 1 -14.7536 5recall 1 -14.7536 http://opus.lingfil.uu.se/ 1 -14.7536 β(pi,j)α(qi,k)α(rk,j) 1 -14.7536 liwc2007 1 -14.7536 two-levelλ-hybrid 1 -14.7536 6.2.3 1 -14.7536 mnr(x, 1 -14.7536 rose126 1 -14.7536 whileitispossiblelearnruleweightsinanmlndirectlyfromtrainingdata,ourapproachatthistime 1 -14.7536 “8th 1 -14.7536 1ways 1 -14.7536 remember, 1 -14.7536 metformin 1 -14.7536 method,allthedifferentphrasesinonelanguagethat 1 -14.7536 thisislowerthanthemostrecentresultsobtained 1 -14.7536 542–553, 1 -14.7536 1063 1 -14.7536 (markup-dl), 1 -14.7536 trainingon519ksentence 1 -14.7536 changesw.r.t 1 -14.7536 `` 1 -14.7536 specifically,werandomlysampled100reviewsof 1 -14.7536 carlota 1 -14.7536 laborintensive 1 -14.7536 2consequently, 1 -14.7536 7reverse 1 -14.7536 mendes, 1 -14.7536 v516 1 -14.7536 thesetimestendtodiminishsignificantlyafterannotators 1 -14.7536 model-up 1 -14.7536 event-defining 1 -14.7536 evaluaarticles 1 -14.7536 pomdp, 1 -14.7536 materiel, 1 -14.7536 bonie 1 -14.7536 specifically: 1 -14.7536 apsipa 1 -14.7536 tworth: 1 -14.7536 “maxent/ge” 1 -14.7536 13.54% 1 -14.7536 theirapproachneedalargecorpustobuildlanguage 1 -14.7536 posinformed 1 -14.7536 “annotator” 1 -14.7536 e-to-c 1 -14.7536 lnew, 1 -14.7536 lda-5.8s 1 -14.7536 ju_cse_temp: 1 -14.7536 many-to-many, 1 -14.7536 thepaperproposesaprobabilisticapproachto 1 -14.7536 ourworkdiffers 1 -14.7536 taggedwithamanagementsuccessionevent,anduse 1 -14.7536 “group-representing 1 -14.7536 2more 1 -14.7536 twor+ 1 -14.7536 noncanonical 1 -14.7536 sub-metrics 1 -14.7536 (rodrıguez 1 -14.7536 cs:xxxxxx 1 -14.7536 •lis 1 -14.7536 twenty-third 1 -14.7536 321–324, 1 -14.7536 add-text 1 -14.7536 thisisdifferentfromtheofflinebinarizationscheme 1 -14.7536 ysuzuki@yamanashi.ac.jp 1 -14.7536 theflatrolerepresentationdesegregatestheframe 1 -14.7536 nastri 1 -14.7536 near-perfect 1 -14.7536 s-adjunct: 1 -14.7536 {dact,asr,slu,dm,emo}. 1 -14.7536 85-90% 1 -14.7536 content/06_22/b3986120.htm 1 -14.7536 (burkhardt 1 -14.7536 p-n 1 -14.7536 (emnlp’04), 1 -14.7536 38–49, 1 -14.7536 “meta-task” 1 -14.7536 1533 1 -14.7536 254–259, 1 -14.7536 vector/domain 1 -14.7536 2.6.4 1 -14.7536 thetextis 1 -14.7536 1248 1 -14.7536 cross-phenomena 1 -14.7536 causativization 1 -14.7536 mimic, 1 -14.7536 /family.owl), 1 -14.7536 (bionlp-st’11) 1 -14.7536 conll’10. 1 -14.7536 ∼cjlin/libsvm) 1 -14.7536 nsp. 1 -14.7536 (tms). 1 -14.7536 sbmt, 1 -14.7536 wikiwalk: 1 -14.7536 generativity 1 -14.7536 mspan 1 -14.7536 inserts, 1 -14.7536 intricacies 1 -14.7536 human-engineered 1 -14.7536 (world) 1 -14.7536 1077–1086, 1 -14.7536 prob41 1 -14.7536 indef 1 -14.7536 3-month 1 -14.7536 1949) 1 -14.7536 workonmweidentificationhasreliedprimarily 1 -14.7536 429–433, 1 -14.7536 ideaisnotnew,butdespiteseveralpreviousattempts 1 -14.7536 hajiˇcov´a 1 -14.7536 “awesome” 1 -14.7536 relationtype 1 -14.7536 acomparison 1 -14.7536 iot 1 -14.7536 (par): 1 -14.7536 89.6%, 1 -14.7536 thekeytoobtainingprecisionandrecallforreordering 1 -14.7536 distance-independent 1 -14.7536 preferenceoverstops. 1 -14.7536 ¯al/) 1 -14.7536 3despite 1 -14.7536 schwenk’s 1 -14.7536 (storing 1 -14.7536 638–647, 1 -14.7536 bag-of-terms 1 -14.7536 blast: 1 -14.7536 1inflection 1 -14.7536 widely-available 1 -14.7536 besidesbeingacomputationalmodelforthematic 1 -14.7536 abekawa 1 -14.7536 coling-2008, 1 -14.7536 α,λ 1 -14.7536 subject–verb–object 1 -14.7536 ap542 1 -14.7536 betweens 1 -14.7536 kuppev 1 -14.7536 〈s6,c6〉: 1 -14.7536 2,243/75.47 1 -14.7536 responds, 1 -14.7536 (16)-(17) 1 -14.7536 a-data, 1 -14.7536 meshing 1 -14.7536 artificial, 1 -14.7536 dehydroxylation, 1 -14.7536 kienappel 1 -14.7536 hu-i 1 -14.7536 includesfullpaperswhicharenewlyannotated,and 1 -14.7536 nerc/nerc.html); 1 -14.7536 cross-genre 1 -14.7536 lookahead, 1 -14.7536 1352–1362, 1 -14.7536 bard: 1 -14.7536 p(character) 1 -14.7536 morphotactics 1 -14.7536 340–349, 1 -14.7536 wewill 1 -14.7536 mve 1 -14.7536 mis-translated 1 -14.7536 earlystopping 1 -14.7536 at-t(x, 1 -14.7536 caedb 1 -14.7536 ks: 1 -14.7536 bucket-brigade 1 -14.7536 semeval-200, 1 -14.7536 iis0713448 1 -14.7536 lindicates 1 -14.7536 featuresqueeze 1 -14.7536 sitting-down 1 -14.7536 steedman∗ 1 -14.7536 1292 1 -14.7536 rubicon 1 -14.7536 sk+1, 1 -14.7536 haspredecessor; 1 -14.7536 to(y,x)]a1〈e,t〉1 1 -14.7536 polarization 1 -14.7536 tsinghua-sogou 1 -14.7536 intersected. 1 -14.7536 420–429, 1 -14.7536 reconcile: 1 -14.7536 4not 1 -14.7536 6251 1 -14.7536 whilecompositionalsystemsaredesignedtocomputephrasalstancesbottom-up,theircalculistraightforwardly 1 -14.7536 scaleability 1 -14.7536 hereweextendtheapproach 1 -14.7536 172–182, 1 -14.7536 withoutnone, 1 -14.7536 returntree 1 -14.7536 o25 1 -14.7536 coarseposterior 1 -14.7536 19(2):263-311. 1 -14.7536 (0.34 1 -14.7536 introduces, 1 -14.7536 trec-9 1 -14.7536 devote§4 1 -14.7536 450–459, 1 -14.7536 etl: 1 -14.7536 penn2malt, 1 -14.7536 (morency 1 -14.7536 conllx 1 -14.7536 length≤10, 1 -14.7536 singletone 1 -14.7536 thisisdoneinsuchawaythattheresultisonce 1 -14.7536 bouckaert 1 -14.7536 concepts/frames 1 -14.7536 conclusionandfuturework 1 -14.7536 344–354, 1 -14.7536 luong 1 -14.7536 skip-n 1 -14.7536 φngram 1 -14.7536 +deleting 1 -14.7536 inconventionaln-gramlanguagemodels,welook 1 -14.7536 4www.sequencepublishing.com/academic.html. 1 -14.7536 53,960,547 1 -14.7536 high-frequent 1 -14.7536 domestically) 1 -14.7536 social-science 1 -14.7536 revoicing 1 -14.7536 outsidethe9-wordcontext 1 -14.7536 kmeans: 1 -14.7536 unavoidably, 1 -14.7536 21.2, 1 -14.7536 237–249, 1 -14.7536 homology34 1 -14.7536 him/1.000 1 -14.7536 anothersetofre-scoringweightswastunedfor300best 1 -14.7536 0.966] 1 -14.7536 labeling: 1 -14.7536 untangle 1 -14.7536 depen1code.google.com/p/berkeleyparser 1 -14.7536 multi-phrasal 1 -14.7536 thelabelsofthewords 1 -14.7536 +1.4 1 -14.7536 “unrealistic 1 -14.7536 annotation.1 1 -14.7536 2479 1 -14.7536 notethattheclassifierdoesnotrequireanyhand 1 -14.7536 andfinally,disregarding 1 -14.7536 belief-model 1 -14.7536 webrieflycoverrelatedworktopositionour 1 -14.7536 i-cab: 1 -14.7536 q-specifiers 1 -14.7536 368–378, 1 -14.7536 fi,c′(x,c) 1 -14.7536 sentence-nlf 1 -14.7536 mteval-v11b.pl 1 -14.7536 pitch-by-pitch 1 -14.7536 1.2k 1 -14.7536 non-issue: 1 -14.7536 emnlp’02. 1 -14.7536 distances; 1 -14.7536 thesefeaturesare 1 -14.7536 pulls 1 -14.7536 lexicalising 1 -14.7536 xt=1(1)k 1 -14.7536 of〈s,c〉, 1 -14.7536 contrast,mstparser 1 -14.7536 thisobservationis 1 -14.7536 529–539, 1 -14.7536 teddclog 1 -14.7536 adults: 1 -14.7536 lpstwl, 1 -14.7536 444–454, 1 -14.7536 lemma-based, 1 -14.7536 0621, 1 -14.7536 15-point 1 -14.7536 〈f1,...,fm〉 1 -14.7536 (2002).2 1 -14.7536 pennell, 1 -14.7536 (1998)’s 1 -14.7536 14,831 1 -14.7536 3265 1 -14.7536 ‘triangledownsld’ 1 -14.7536 iis-0904684. 1 -14.7536 ntcir-4. 1 -14.7536 bertinetto 1 -14.7536 watermarks 1 -14.7536 “paraphrasing” 1 -14.7536 usinggoldmentionsraisestheoverallscoreby 1 -14.7536 featurama/ 1 -14.7536 innocent 1 -14.7536 subedges, 1 -14.7536 †eisner 1 -14.7536 qij: 1 -14.7536 explaination 1 -14.7536 3http://duc.nist.gov/duc2005/ 1 -14.7536 15964 1 -14.7536 lexification 1 -14.7536 learningthevariousparametervaluestofuturework. 1 -14.7536 re-used. 1 -14.7536 7–8) 1 -14.7536 ijcai-95, 1 -14.7536 constructingdlfs 1 -14.7536 700, 1 -14.7536 lowestscoring 1 -14.7536 mib(v,per) 1 -14.7536 67.85%, 1 -14.7536 lerman, 1 -14.7536 bouma 1 -14.7536 gradient-basedtechniquesrequireadifferentiable 1 -14.7536 numberofpagesindexedbygooglegrows,thenumber 1 -14.7536 ex9as 1 -14.7536 frederik.vaassen@ua.ac.be 1 -14.7536 figure2(b)showsthelearning 1 -14.7536 thetaskofalanguagemodelistoprovideameasure 1 -14.7536 organization: 1 -14.7536 abstains, 1 -14.7536 suchpatternsarebeyondthereachofcurrent 1 -14.7536 (266 1 -14.7536 1426–1436, 1 -14.7536 caseful 1 -14.7536 diversity), 1 -14.7536 defineyprime 1 -14.7536 +2.7 1 -14.7536 communicat1centre 1 -14.7536 thatcombineexpertidentificationandper-user 1 -14.7536 weevaluateourmethodonkeyphrase 1 -14.7536 (2003b)) 1 -14.7536 nonrelevant 1 -14.7536 uimabased 1 -14.7536 v3: 1 -14.7536 575-584. 1 -14.7536 compute-supertreelets, 1 -14.7536 orautomaticallylearnt 1 -14.7536 t−r: 1 -14.7536 ga: 1 -14.7536 (glow) 1 -14.7536 194–199, 1 -14.7536 2...5, 1 -14.7536 shezaf, 1 -14.7536 s∈{1,...,s}we 1 -14.7536 haspopulation; 1 -14.7536 naacl-hlt. 1 -14.7536 10these 1 -14.7536 1,...qn]∧x=qi 1 -14.7536 similarly,mst 1 -14.7536 nullnullnullnullchase, 1 -14.7536 .04). 1 -14.7536 21–30, 1 -14.7536 byproduct 1 -14.7536 273. 1 -14.7536 anullw1bcw2d 1 -14.7536 catalogid=ldc2006t13 1 -14.7536 3a/b 1 -14.7536 detector’s 1 -14.7536 thenodesin 1 -14.7536 2http://www.duden.de/ 1 -14.7536 backtranslated 1 -14.7536 crftagger6, 1 -14.7536 56, 1 -14.7536 languagecommons 1 -14.7536 “d” 1 -14.7536 muc-f1 1 -14.7536 eacl. 1 -14.7536 severalotherstudieshaveexaminedthemainideaof 1 -14.7536 quartiles 1 -14.7536 felix 1 -14.7536 annotatedseed) 1 -14.7536 2077 1 -14.7536 uturku09was 1 -14.7536 31(1):71–106. 1 -14.7536 363–374, 1 -14.7536 162–171, 1 -14.7536 63.10%, 1 -14.7536 auses 1 -14.7536 least: 1 -14.7536 reorderingperformance. 1 -14.7536 lenrat: 1 -14.7536 evalita’09 1 -14.7536 11(1). 1 -14.7536 35.6% 1 -14.7536 theseresultsarecomparableinmagnitudetoour 1 -14.7536 94.70 1 -14.7536 “levels” 1 -14.7536 factrank 1 -14.7536 craig? 1 -14.7536 theseweightsrepresentthedegreeofcontributionof 1 -14.7536 7“precision” 1 -14.7536 position,they 1 -14.7536 thresholdoptimizationmix-upwe 1 -14.7536 thepatternsusedbyconstraintstomatchmention 1 -14.7536 pre-filters 1 -14.7536 i(v,then). 1 -14.7536 automatable: 1 -14.7536 part-of-speechtagourdatausingaperceptrontagger 1 -14.7536 (aa). 1 -14.7536 2nominal 1 -14.7536 4(c). 1 -14.7536 meteornext 1 -14.7536 large- 1 -14.7536 monosemous-to-polysemous 1 -14.7536 o(swt). 1 -14.7536 ejections, 1 -14.7536 100,500 1 -14.7536 (break 1 -14.7536 (50-75%) 1 -14.7536 paper1 1 -14.7536 redoing 1 -14.7536 languages,2 1 -14.7536 log-liner 1 -14.7536 however,feature-based 1 -14.7536 si∈ 1 -14.7536 annotationswe 1 -14.7536 5http://rightwingnews.com/ 1 -14.7536 positions/roles 1 -14.7536 “moses 1 -14.7536 metacontrol 1 -14.7536 openccg3 1 -14.7536 p≤.001. 1 -14.7536 vnp-a. 1 -14.7536 thispaperpresentstwoinstance-leveltransfer 1 -14.7536 2$((7&4")$%& 1 -14.7536 ciphertext:i 1 -14.7536 dense-subgraph 1 -14.7536 “bootstrapping.” 1 -14.7536 entity-related 1 -14.7536 ceist 1 -14.7536 1edges 1 -14.7536 10there 1 -14.7536 onlyprimaryargumentsareconsidered,thefirstfive 1 -14.7536 ducted 1 -14.7536 ex9all 1 -14.7536 forthepurpose 1 -14.7536 rivi`ere.*} 1 -14.7536 pocketsphinx: 1 -14.7536 appropriately, 1 -14.7536 languagemodel(lm)sampling 1 -14.7536 mathieu, 1 -14.7536 #.af: 1 -14.7536 srilm4 1 -14.7536 sincethen, 1 -14.7536 servations 1 -14.7536 dual-query 1 -14.7536 score(n5,n6′) 1 -14.7536 weconstructedtwomachinelearningtaskstoexploittheannotatedspatialinformationtodeterminewhat 1 -14.7536 donthave: 1 -14.7536 interface's 1 -14.7536 n=50. 1 -14.7536 534–544. 1 -14.7536 conv. 1 -14.7536 sc-mv: 1 -14.7536 ah2=b3,ah5=b1 1 -14.7536 “watch”) 1 -14.7536 (sankaranarayanan 1 -14.7536 sethy 1 -14.7536 frep 1 -14.7536 espino 1 -14.7536 bilingualdictionary 1 -14.7536 (18.c) 1 -14.7536 assortivitiy 1 -14.7536 782–792, 1 -14.7536 inull=j 1 -14.7536 generalisation, 1 -14.7536 relative-context 1 -14.7536 non-pca) 1 -14.7536 systemareflects 1 -14.7536 20ng 1 -14.7536 389–398, 1 -14.7536 0,0,,10,11,,5,6,7,8,9,,,,3,4,2,2,12 1 -14.7536 aaai’04. 1 -14.7536 “fine-grained” 1 -14.7536 tivity 1 -14.7536 pages:1012-1020. 1 -14.7536 [bottom] 1 -14.7536 dependency-tree 1 -14.7536 non-vacuous 1 -14.7536 morphadorner6 1 -14.7536 rev.getarticleid( 1 -14.7536 1107 1 -14.7536 8note, 1 -14.7536 δθ, 1 -14.7536 pre-tagging 1 -14.7536 4table 1 -14.7536 true-cased, 1 -14.7536 isr-wn, 1 -14.7536 subgraphgb 1 -14.7536 shinny 1 -14.7536 amitryptyliny 1 -14.7536 (lst) 1 -14.7536 c@1. 1 -14.7536 (formal) 1 -14.7536 speling? 1 -14.7536 blaheta 1 -14.7536 am≻pm 1 -14.7536 german,despitebeinggeneticallyrelatedtoenglish, 1 -14.7536 incorporatingsuchelaboratetechniquesinourframework 1 -14.7536 885–894, 1 -14.7536 forprevention,thehighratioofthesenoun 1 -14.7536 astraightforwardanalysissayingthatmeronymy 1 -14.7536 unary-x 1 -14.7536 pirro, 1 -14.7536 ppois 1 -14.7536 1potential 1 -14.7536 gooood 1 -14.7536 legend 1 -14.7536 maxl 1 -14.7536 2dependence 1 -14.7536 two-component-system: 1 -14.7536 6high-confidence 1 -14.7536 tei,x 1 -14.7536 #twitter) 1 -14.7536 co-task, 1 -14.7536 wearguethatatranslationmetricthatreflectsmeaning 1 -14.7536 xmapsto→z(p(pii)) 1 -14.7536 asked,three 1 -14.7536 nlpba, 1 -14.7536 sjöbergh 1 -14.7536 concordu, 1 -14.7536 1104–1115, 1 -14.7536 czech-specific 1 -14.7536 discourse-todialogue 1 -14.7536 vulnerability 1 -14.7536 162–172, 1 -14.7536 treetagger-chunker 1 -14.7536 (austria), 1 -14.7536 w(θ). 1 -14.7536 1,fmmr 1 -14.7536 83.16 1 -14.7536 173–184, 1 -14.7536 “free-na¨ıve” 1 -14.7536 (ringgaard@google.com) 1 -14.7536 wal, 1 -14.7536 topicz 1 -14.7536 evaluable 1 -14.7536 m¨orschwil: 1 -14.7536 6our 1 -14.7536 svms), 1 -14.7536 label117 1 -14.7536 (liuetal., 1 -14.7536 4formal 1 -14.7536 superpositions) 1 -14.7536 20-topic 1 -14.7536 theconfusionforestapproachmaysufferfromparsing 1 -14.7536 producttextx∧electpresidentnull(y)]3 1 -14.7536 andnon-erasing,whichmeansthatifwewrite 1 -14.7536 word–topic 1 -14.7536 boguraev, 1 -14.7536 weattempttocaptureaspectsofthemodifier 1 -14.7536 expressives: 1 -14.7536 absolute: 1 -14.7536 knowledgeharvesting 1 -14.7536 8,765,444 1 -14.7536 1window 1 -14.7536 304–312, 1 -14.7536 allophones), 1 -14.7536 contable 1 -14.7536 scarce; 1 -14.7536 s-d-tfidf: 1 -14.7536 geometricalmodels/sharedevaluation 1 -14.7536 >arriverà 1 -14.7536 thisparameter 1 -14.7536 fulfiled, 1 -14.7536 lossld 1 -14.7536 thetop 1 -14.7536 26–37, 1 -14.7536 thickness 1 -14.7536 average-linkage. 1 -14.7536 french↔english, 1 -14.7536 1unfortunately, 1 -14.7536 summarization/ 1 -14.7536 1638 1 -14.7536 smt-style 1 -14.7536 φi(a,e,f) 1 -14.7536 8candito 1 -14.7536 remindful 1 -14.7536 r−: 1 -14.7536 ‘signature’ 1 -14.7536 highperforming 1 -14.7536 re69 1 -14.7536 7they 1 -14.7536 −3.5 1 -14.7536 pro1) 1 -14.7536 842 1 -14.7536 marginbased 1 -14.7536 para4this 1 -14.7536 74.40++ 1 -14.7536 1249–1257, 1 -14.7536 ’recent 1 -14.7536 proposedmethodusingdocument-levelcontextiseffec- 1 -14.7536 13a 1 -14.7536 cmu-smt 1 -14.7536 improvment 1 -14.7536 set-ups 1 -14.7536 1318 1 -14.7536 cookandstevenson(2009)employedthe 1 -14.7536 minimal-distance 1 -14.7536 care700 1 -14.7536 hobs, 1 -14.7536 ceauşu, 1 -14.7536 -extract-person-has-treatment-or-medical-diagnostic> 1 -14.7536 4results 1 -14.7536 walter; 1 -14.7536 (usp): 1 -14.7536 31.97 1 -14.7536 soundproof 1 -14.7536 594–604, 1 -14.7536 +8.38 1 -14.7536 outperforms, 1 -14.7536 “random” 1 -14.7536 martineau 1 -14.7536 government‟s 1 -14.7536 rohde 1 -14.7536 〈r,l〉∈exts,σ,a(rankpp0,trivspa,trivlp) 1 -14.7536 weperformed 1 -14.7536 thatsamenounpairdoesnotco-occurwithanyother 1 -14.7536 principle,s 1 -14.7536 mosan, 1 -14.7536 98-letter 1 -14.7536 acl-08. 1 -14.7536 ££geogr-part-selection-criterium 1 -14.7536 f2– 1 -14.7536 log-scores 1 -14.7536 multi6k=50 1 -14.7536 score(n6,n6′) 1 -14.7536 gel: 1 -14.7536 desegmenting 1 -14.7536 relation/length 1 -14.7536 dialoguestyle 1 -14.7536 〈1,=t+wh≈n〉(00221),〈0,d-ep〉(00222000,00222001,00222002),〈1,=d=dv〉(0022221), 1 -14.7536 pac-style 1 -14.7536 rapid, 1 -14.7536 assumedw1 1 -14.7536 univerisity 1 -14.7536 fig.2 1 -14.7536 toencouragethisihaveusedafactoredtranslation 1 -14.7536 temeval-2010 1 -14.7536 6)) 1 -14.7536 c-feel-it, 1 -14.7536 servicize 1 -14.7536 automnlp 1 -14.7536 “transductive 1 -14.7536 mavuno: 1 -14.7536 inves130 1 -14.7536 ecureliesonthehypothesisthatnounsandverbs 1 -14.7536 1356–1364, 1 -14.7536 ^ap 1 -14.7536 ascribes 1 -14.7536 textdriven 1 -14.7536 anotaed 1 -14.7536 thatuniontext 1 -14.7536 annotation/previous/for 1 -14.7536 algoithm 1 -14.7536 acl/eacl 1 -14.7536 chromosome, 1 -14.7536 /; 1 -14.7536 nistweightedld 1 -14.7536 32nd 1 -14.7536 case-marking, 1 -14.7536 wethink 1 -14.7536 coments 1 -14.7536 ever-expandingdatasetderivedfromstate-of-the-art 1 -14.7536 scilex) 1 -14.7536 conference-6: 1 -14.7536 concludingremarks 1 -14.7536 so-learned 1 -14.7536 group-average 1 -14.7536 dybala 1 -14.7536 webkb(sen 1 -14.7536 k1;j 1 -14.7536 theq(a) 1 -14.7536 improved-kneser-ney 1 -14.7536 98.06 1 -14.7536 matrax, 1 -14.7536 1405–1415, 1 -14.7536 segmentator 1 -14.7536 waiting/*waited 1 -14.7536 fusco 1 -14.7536 5they 1 -14.7536 s-ary 1 -14.7536 multilanguage 1 -14.7536 causality-style 1 -14.7536 anselmope˜nas 1 -14.7536 663–672, 1 -14.7536 prob653 1 -14.7536 sundance’s 1 -14.7536 message-topic: 1 -14.7536 f+m: 1 -14.7536 “checkpoint” 1 -14.7536 termlabel 1 -14.7536 berkeley+lattice 1 -14.7536 human-likeness: 1 -14.7536 2named 1 -14.7536 rebuttals). 1 -14.7536 emnlpconll-07. 1 -14.7536 guish 1 -14.7536 nulljnulljnull[1,j]null 1 -14.7536 difference.n 1 -14.7536 4.44% 1 -14.7536 2lis, 1 -14.7536 copy-pasted 1 -14.7536 tanakaishii 1 -14.7536 determiningtheoptimalsegmentationissufficientfor 1 -14.7536 sinceonlyautomaticparsetrees 1 -14.7536 ijcnlp2008. 1 -14.7536 first-letter 1 -14.7536 (l+r) 1 -14.7536 maximum-accuracy 1 -14.7536 trainingtesting 1 -14.7536 65-73. 1 -14.7536 (tsi-020110-2009-439) 1 -14.7536 (1994)discussthetemporalinertiaofnarrativetexts-timemovesforward 1 -14.7536 (hltemnlp), 1 -14.7536 disimproved 1 -14.7536 mismatchedcomparisons 1 -14.7536 high-scored 1 -14.7536 7bansalandklein(2010)alsoexperimentedwithsymbolrefinement 1 -14.7536 5.94% 1 -14.7536 ’+...’ 1 -14.7536 “dichotomizer”) 1 -14.7536 followingpropbank,therearenosemanticframes 1 -14.7536 serials 1 -14.7536 nips2002. 1 -14.7536 ysis 1 -14.7536 h3(x) 1 -14.7536 r+: 1 -14.7536 mac/pc 1 -14.7536 annonymization 1 -14.7536 prose? 1 -14.7536 savas 1 -14.7536 oracle-then-deletion 1 -14.7536 8according 1 -14.7536 wumpus, 1 -14.7536 wealsoinvestigatedtheeffect 1 -14.7536 thelekbotrobotcanmoveforwardandbackward, 1 -14.7536 5.5.1 1 -14.7536 12(5):563-598. 1 -14.7536 electron. 1 -14.7536 interarticle 1 -14.7536 irr 1 -14.7536 wordsneednot 1 -14.7536 partof-speech-wise 1 -14.7536 the12mostcommonlymisusedprepositionsbyesl 1 -14.7536 8.1.2 1 -14.7536 kestemont 1 -14.7536 16this 1 -14.7536 glenn 1 -14.7536 500-sentence 1 -14.7536 counter-training 1 -14.7536 4.7], 1 -14.7536 advantageofemployingpackedforesttoenhancethemen- 1 -14.7536 proliferation) 1 -14.7536 morrissey, 1 -14.7536 trans6: 1 -14.7536 concatenating: 1 -14.7536 4−voting 1 -14.7536 59–63, 1 -14.7536 1451 1 -14.7536 attr-c: 1 -14.7536 southeastern 1 -14.7536 complementgeneration 1 -14.7536 acl-ijcnlp. 1 -14.7536 um, 1 -14.7536 37.3(23) 1 -14.7536 volving 1 -14.7536 plan; 1 -14.7536 elabo164 1 -14.7536 refertofigure3,theimprovementofpolarityclassification 1 -14.7536 developingtoolstocleanthegazetteers. 1 -14.7536 reranker: 1 -14.7536 onlyseed: 1 -14.7536 gwb16 1 -14.7536 c0−80 1 -14.7536 qpdg, 1 -14.7536 f(e) 1 -14.7536 ecir-2009, 1 -14.7536 vector11 1 -14.7536 “decode” 1 -14.7536 evaluation7 1 -14.7536 ipsini: 1 -14.7536 in-stance 1 -14.7536 theere 1 -14.7536 typesrepresentdifferentlevelsofcomplexity. 1 -14.7536 •so-tp+gi: 1 -14.7536 lmsr 1 -14.7536 rename, 1 -14.7536 algorithms,the 1 -14.7536 furthermore,therelationsamongindividualopinions 1 -14.7536 barriere(1997) 1 -14.7536 questionnaire: 1 -14.7536 engine1 1 -14.7536 acl’06, 1 -14.7536 strikes 1 -14.7536 341–343, 1 -14.7536 2885 1 -14.7536 hua, 1 -14.7536 “≥others”(resp. 1 -14.7536 corenlp.shtml 1 -14.7536 thetwoversionsdifferintheir 1 -14.7536 d0–d6, 1 -14.7536 minconf, 1 -14.7536 relaxcor, 1 -14.7536 scl-mi, 1 -14.7536 procee 1 -14.7536 scored: 1 -14.7536 memory-efficiency. 1 -14.7536 gellerstam 1 -14.7536 automaticcally 1 -14.7536 cherry-pick 1 -14.7536 19–26, 1 -14.7536 (matthias, 1 -14.7536 6356 1 -14.7536 observers, 1 -14.7536 semi-manually 1 -14.7536 figure1:chinese 1 -14.7536 indirect-hmmbased 1 -14.7536 ual.pdf 1 -14.7536 13part 1 -14.7536 (sager 1 -14.7536 4terry 1 -14.7536 anagha 1 -14.7536 trees〈α,β〉being 1 -14.7536 trialled 1 -14.7536 (simplifiedne−originalne)originalne 1 -14.7536 werankthelabelcandidates 1 -14.7536 ctb5, 1 -14.7536 manifest. 1 -14.7536 beavers 1 -14.7536 wholesentence(bag-of-words)informationwhichis 1 -14.7536 mainguy 1 -14.7536 3assuming 1 -14.7536 stringe 1 -14.7536 11–17, 1 -14.7536 epr: 1 -14.7536 textrank: 1 -14.7536 tokenization/morpheme 1 -14.7536 (synt) 1 -14.7536 259–265, 1 -14.7536 niesler 1 -14.7536 (gfss) 1 -14.7536 semisupervised, 1 -14.7536 ciations 1 -14.7536 distortion-limit 1 -14.7536 tachment 1 -14.7536 internally: 1 -14.7536 process714 1 -14.7536 experimentinfig.5,wechangedthemaximumnumberof 1 -14.7536 kricfalusi, 1 -14.7536 referent: 1 -14.7536 theoverallperformanceofthetwostrategiesisabout 1 -14.7536 1numbers 1 -14.7536 caseunigram 1 -14.7536 heideltime: 1 -14.7536 kul: 1 -14.7536 ¯φ(nonea)←∅for 1 -14.7536 (a)-(c)), 1 -14.7536 s→q(s)≥q(sprime) 1 -14.7536 therulesaysthatiftheatomfortheverb“forgetto”appears 1 -14.7536 87.160 1 -14.7536 china↔china) 1 -14.7536 increasingly, 1 -14.7536 “objective 1 -14.7536 weigths 1 -14.7536 (2010a)and 1 -14.7536 methodolgy 1 -14.7536 1559 1 -14.7536 outlets 1 -14.7536 4oneofthereviewerssuggestedthatevaluatingtheaccuracy 1 -14.7536 thus,thenumberofdimensionsof 1 -14.7536 inflected. 1 -14.7536 weaddedthepos 1 -14.7536 nullnull=nullnull1null 1 -14.7536 frank(2004)discussesthepotentialofapplyingvariousgeneralisation‘filters’tocorpus-inducedsyntaxsemantics 1 -14.7536 slipshod 1 -14.7536 well-know 1 -14.7536 ilustrates, 1 -14.7536 notarrowrightc2: 1 -14.7536 12.2) 1 -14.7536 floorleftyfloorright 1 -14.7536 voting-inspired 1 -14.7536 ifc={1...n}, 1 -14.7536 getting-giving 1 -14.7536 8once 1 -14.7536 fencepost 1 -14.7536 boxer, 1 -14.7536 ourprocedureextendsthecontextofthesentence 1 -14.7536 mitchellcountshowofteneachwordtypeappearsin 1 -14.7536 pre-compute 1 -14.7536 5http://www.cs.pitt.edu/mpqa/databaserelease/ 1 -14.7536 0.3186 1 -14.7536 mrw. 1 -14.7536 4-core 1 -14.7536 comic-de, 1 -14.7536 (context) 1 -14.7536 usually: 1 -14.7536 biboot: 1 -14.7536 functionalities, 1 -14.7536 ngram-based 1 -14.7536 dekang, 1 -14.7536 confusion-network-based 1 -14.7536 datta 1 -14.7536 multi-classification, 1 -14.7536 de– 1 -14.7536 temp-pp: 1 -14.7536 conclusively 1 -14.7536 f8-12, 1 -14.7536 355–362, 1 -14.7536 systemwithsoon’straining 1 -14.7536 ―extraction‖ 1 -14.7536 impracticality 1 -14.7536 10www.ilc.cnr.it/eagles/home.html 1 -14.7536 defacto 1 -14.7536 14,827 1 -14.7536 physicochemical 1 -14.7536 complementizer), 1 -14.7536 relationsthan 1 -14.7536 gwet 1 -14.7536 8022 1 -14.7536 structurallearningmethodisusedtoproduce 1 -14.7536 /inko/, 1 -14.7536 interrgb 1 -14.7536 con3past 1 -14.7536 t-scores, 1 -14.7536 (ssl), 1 -14.7536 73–79, 1 -14.7536 tempeval-2. 1 -14.7536 stanfill 1 -14.7536 pragmatic: 1 -14.7536 extracts, 1 -14.7536 externalpositives(pcbc): 1 -14.7536 fthe 1 -14.7536 ipsyn 1 -14.7536 68–72, 1 -14.7536 inneighbor 1 -14.7536 analyfeatures 1 -14.7536 newsgroups6 1 -14.7536 monosemous-tomonosemous 1 -14.7536 strategy/199/07/01/microsoft-sued-for-racist-application2072468/ 1 -14.7536 “self-regulation,” 1 -14.7536 474–485, 1 -14.7536 =2, 1 -14.7536 (2007b)), 1 -14.7536 expanding, 1 -14.7536 3audio 1 -14.7536 mihalceaand 1 -14.7536 .002. 1 -14.7536 incorrect/inexact 1 -14.7536 65–75, 1 -14.7536 ‘+cluster’ 1 -14.7536 23,000 1 -14.7536 372–378, 1 -14.7536 koyama 1 -14.7536 erbach 1 -14.7536 adjacent-utterance 1 -14.7536 thentheoccurrencesof 1 -14.7536 richer/noisier 1 -14.7536 (bayesian) 1 -14.7536 stil, 1 -14.7536 piatetsky-shapiro 1 -14.7536 taxon&cid=77643 1 -14.7536 toimprovethespeedofthealgorithm,thetop500 1 -14.7536 concisereportof 1 -14.7536 thescopeprocessingtaskisconcernedwithdeterminingatasentencelevelwhichtokensareaffected 1 -14.7536 1dylan 1 -14.7536 simplistically 1 -14.7536 ax.t 1 -14.7536 s−kps+m, 1 -14.7536 97,813 1 -14.7536 above/below 1 -14.7536 theta-marking. 1 -14.7536 8subsequent 1 -14.7536 thisstudyusedavariety 1 -14.7536 (keywords) 1 -14.7536 graphstructured 1 -14.7536 (mart´ınezromero 1 -14.7536 non-transitive 1 -14.7536 lowestperplexity 1 -14.7536 “backoff” 1 -14.7536 itisworthnotingthatthesemanticrolefeaturesin 1 -14.7536 difficult-to-translate 1 -14.7536 tab.1). 1 -14.7536 reduce–binary–fl/rg–x, 1 -14.7536 mispronunciations, 1 -14.7536 “father” 1 -14.7536 componentc3 1 -14.7536 567-575. 1 -14.7536 ax3 1 -14.7536 lexemes: 1 -14.7536 pt,l(〈e,f〉) 1 -14.7536 theydonotreceivetheuncertaintyadaptation, 1 -14.7536 c3-f: 1 -14.7536 generative1 1 -14.7536 72.19 1 -14.7536 itshowsaperformanceoptimizedforprecisionwith 1 -14.7536 perfectrankingwouldplaceallthecorrecttweetsbefore 1 -14.7536 sub-entry 1 -14.7536 s[a] 1 -14.7536 disprefers 1 -14.7536 harmonization, 1 -14.7536 fames 1 -14.7536 trade: 1 -14.7536 6743 1 -14.7536 “-lr” 1 -14.7536 223–229, 1 -14.7536 nextwetestourn-bestrescoringapproachusinga 1 -14.7536 minimun 1 -14.7536 thisspecifiesthatdaysarepartsof 1 -14.7536 improvementofaveragefunctionisnotsignificantat 1 -14.7536 attributes-words 1 -14.7536 (§2.1) 1 -14.7536 +1.06 1 -14.7536 newwire 1 -14.7536 antecedent-anaphora 1 -14.7536 kinetic 1 -14.7536 lumps 1 -14.7536 speaker12prn: 1 -14.7536 sócrates); 1 -14.7536 4,359 1 -14.7536 “proportional 1 -14.7536 thesewerethetopscores 1 -14.7536 proposalimprovestheaccuracyoftheextractedentities. 1 -14.7536 matrixl0×00nullnullmodnull 1 -14.7536 (pdtb), 1 -14.7536 plexity 1 -14.7536 #0705832. 1 -14.7536 grene 1 -14.7536 (haitian-english) 1 -14.7536 affirm 1 -14.7536 of-.html 1 -14.7536 481–490, 1 -14.7536 3,490 1 -14.7536 splitting-by-variables 1 -14.7536 short-comings 1 -14.7536 somasundaran, 1 -14.7536 lwlm+wordnet: 1 -14.7536 buildv 1 -14.7536 10-topic 1 -14.7536 sqp 1 -14.7536 left-to-right: 1 -14.7536 for+wbow: 1 -14.7536 889–898, 1 -14.7536 experimentaions 1 -14.7536 ghodke 1 -14.7536 642–652, 1 -14.7536 (srs) 1 -14.7536 construction-error 1 -14.7536 s3vm, 1 -14.7536 maxscore: 1 -14.7536 (novel) 1 -14.7536 etts: 1 -14.7536 themethodisonlyapplicable 1 -14.7536 1 1 -14.7536 440–449, 1 -14.7536 (sgwater@inf.ed.ac.uk) 1 -14.7536 sample-per-sentence 1 -14.7536 cas 1 -14.7536 4nakov 1 -14.7536 w−1−n+1wh−1−mg 1 -14.7536 wordw2 1 -14.7536 mispeling 1 -14.7536 recentwork 1 -14.7536 wewillrelease 1 -14.7536 acl-04. 1 -14.7536 5-7 1 -14.7536 mapem 1 -14.7536 ensuing, 1 -14.7536 “j-tweet 1 -14.7536 one-sample 1 -14.7536 1264 1 -14.7536 4words 1 -14.7536 commenton 1 -14.7536 bleun4 1 -14.7536 6.1.3 1 -14.7536 (wm,wc), 1 -14.7536 finvestigate 1 -14.7536 wheremostofthetreeletsarenon-emptyiswhenthe 1 -14.7536 15www.ist-luna.eu 1 -14.7536 e-shops 1 -14.7536 1hill 1 -14.7536 notethatconventionalprecision-andrecall-based 1 -14.7536 [ball]a2 1 -14.7536 14additional 1 -14.7536 7conclusion 1 -14.7536 648–653, 1 -14.7536 clair::bio::gin::interaction 1 -14.7536 ‘cn’ 1 -14.7536 location=tasks&taskid=9 1 -14.7536 .......................................... 1 -14.7536 sentenceselection 1 -14.7536 basili, 1 -14.7536 partially-judged 1 -14.7536 143–146, 1 -14.7536 ρ(dikhataa) 1 -14.7536 willalsoneed,foreachtriggeriandproteinpairp,q, 1 -14.7536 experimentseti:wecompareoursimilaritymeasure 1 -14.7536 1993a) 1 -14.7536 ha-1: 1 -14.7536 73.0, 1 -14.7536 oneislearnedfrom 1 -14.7536 thereason 1 -14.7536 gre3d7: 1 -14.7536 distij 1 -14.7536 “patient” 1 -14.7536 geometric-mean 1 -14.7536 608.74 1 -14.7536 66"1232452" 1 -14.7536 theaqlannotatortagssomeviewsasoutputviews, 1 -14.7536 light6(joachims, 1 -14.7536 26%, 1 -14.7536 scp, 1 -14.7536 additionally,riedel 1 -14.7536 configura289 1 -14.7536 goudbeek 1 -14.7536 miniature 1 -14.7536 29-iteration 1 -14.7536 136–141, 1 -14.7536 8q0 1 -14.7536 alot 1 -14.7536 extendthisworktoothertypesofauthorityclaimsin 1 -14.7536 “production 1 -14.7536 previousstudies 1 -14.7536 corpuses: 1 -14.7536 742–751, 1 -14.7536 instrument-agency: 1 -14.7536 λ∈[0, 1 -14.7536 data-driven/statistical 1 -14.7536 lexical/pos 1 -14.7536 koskenniemi 1 -14.7536 ,.ie, 1 -14.7536 13972 1 -14.7536 totaly 1 -14.7536 n=1000 1 -14.7536 h,p, 1 -14.7536 themaximumnumberofmertiterations 1 -14.7536 pu(2,3) 1 -14.7536 glosaries 1 -14.7536 ...,nk(...))...) 1 -14.7536 blanking 1 -14.7536 collectingandorganizingthisdiverseinformationisanimportant 1 -14.7536 hierarchical: 1 -14.7536 458,505 1 -14.7536 (9c) 1 -14.7536 tread 1 -14.7536 polemic 1 -14.7536 3.3.4 1 -14.7536 s0pn0wp 1 -14.7536 3σ 1 -14.7536 binning, 1 -14.7536 l=0 1 -14.7536 changetrend(1997, 1 -14.7536 880–888, 1 -14.7536 9other 1 -14.7536 227–236, 1 -14.7536 nowo-pcfg, 1 -14.7536 ins, 1 -14.7536 9-gram 1 -14.7536 europarl), 1 -14.7536 [vectorβt=1,vectorβt=2]. 1 -14.7536 (chinese-to-english, 1 -14.7536 thesequencemodelsoutperformthecclbenchmarkatbothtasksand 1 -14.7536 thattheseextratitleswereusefulforconstructingthe 1 -14.7536 instancelevel 1 -14.7536 prescher 1 -14.7536 f→e 1 -14.7536 forty-eight 1 -14.7536 agendas, 1 -14.7536 81-88. 1 -14.7536 alignmentsthatarefoundinbothdirections. 1 -14.7536 icml’99. 1 -14.7536 paraphrasegeneration 1 -14.7536 299–308, 1 -14.7536 oscillation 1 -14.7536 scorings 1 -14.7536 top1 1 -14.7536 framenetincludesmanually 1 -14.7536 i(x;y) 1 -14.7536 surveymonkey1, 1 -14.7536 quantityp(w) 1 -14.7536 somers. 1 -14.7536 rom(x,nyc)∧f(x) 1 -14.7536 addi88 1 -14.7536 365–371, 1 -14.7536 acto 1 -14.7536 declension 1 -14.7536 typos/spelling 1 -14.7536 bionlpshared 1 -14.7536 7340) 1 -14.7536 corpus3 1 -14.7536 unravelling 1 -14.7536 arguments1 1 -14.7536 emo107 1 -14.7536 textrankalgorithm(mihalceaandtarau,2004)toiden- 1 -14.7536 prefixprobabilitiescanbeusedtocomputeprobability 1 -14.7536 x&z(p(pii)) 1 -14.7536 white≤0.823 1 -14.7536 (3)–(5), 1 -14.7536 945–954, 1 -14.7536 dependen1144 1 -14.7536 7885, 1 -14.7536 lrec-2004. 1 -14.7536 pertinence 1 -14.7536 software/treebank_converter/ 1 -14.7536 undistinguished, 1 -14.7536 1607–1615, 1 -14.7536 3paired 1 -14.7536 annotators' 1 -14.7536 gensaf(y) 1 -14.7536 icml, 1 -14.7536 uncensored 1 -14.7536 mappingλfrom 1 -14.7536 712–724, 1 -14.7536 non-hierarchical) 1 -14.7536 kathleen.r. 1 -14.7536 naturaly, 1 -14.7536 1308–1317, 1 -14.7536 practicality 1 -14.7536 shao, 1 -14.7536 person-indicating 1 -14.7536 cunyblender 1 -14.7536 fory(i) 1 -14.7536 lscript2-norm 1 -14.7536 typedin 1 -14.7536 clusteringofthefeatures 1 -14.7536 ktimeml: 1 -14.7536 therefore,theweightslearned 1 -14.7536 inourexperiments,werestrict 1 -14.7536 polak 1 -14.7536 603–608, 1 -14.7536 aslscript−k, 1 -14.7536 (second) 1 -14.7536 phonotactics, 1 -14.7536 clda’s 1 -14.7536 1092–1103, 1 -14.7536 shemtov 1 -14.7536 wiktionary3), 1 -14.7536 suchlistscanbeextractedfrommonolingualdatafor 1 -14.7536 hieroor 1 -14.7536 unnanotated 1 -14.7536 polgren 1 -14.7536 ranker-level: 1 -14.7536 russell, 1 -14.7536 exmbest 1 -14.7536 sentencestart 1 -14.7536 grab, 1 -14.7536 time-stamp 1 -14.7536 ineffectiveness 1 -14.7536 escobar 1 -14.7536 3hartmann(2005) 1 -14.7536 importantreasonwhythishappensisthatnounsmod- 1 -14.7536 2.usability: 1 -14.7536 in-laboratory 1 -14.7536 pair〈n,d〉leads 1 -14.7536 6,6,7_8,4,3,3,3,3,3,0,0,0,0,0,1,1,9,9 1 -14.7536 alsoimaginesuchinformationwouldbeusefulforidentifyingcommunicationbreakdowns. 1 -14.7536 systemscanbesplitintoapproachesbasedonstatistical 1 -14.7536 2.619] 1 -14.7536 lambda-values 1 -14.7536 wholeheartedly 1 -14.7536 (58.39 1 -14.7536 thefrenchberkeleyparserhasasingletag, 1 -14.7536 uml’s 1 -14.7536 emnlp2010. 1 -14.7536 difficult-to-classify 1 -14.7536 (2002): 1 -14.7536 #(t,d) 1 -14.7536 (terman, 1 -14.7536 smallscaled 1 -14.7536 weird_banana, 1 -14.7536 �(fahui)/demonstrate” 1 -14.7536 ‘pennconverter’5 1 -14.7536 cprim(s) 1 -14.7536 out-of-dictionary 1 -14.7536 meteor11 1 -14.7536 five→ 1 -14.7536 emotion-2010, 1 -14.7536 #463: 1 -14.7536 t3’s 1 -14.7536 converted8 1 -14.7536 thesegazetteersrepresentgeneralknowledge 1 -14.7536 use→+ 1 -14.7536 fas. 1 -14.7536 datapreparation 1 -14.7536 variableˆclass. 1 -14.7536 0.0175). 1 -14.7536 good  1 -14.7536 0.7801. 1 -14.7536 primarly 1 -14.7536 “phosphorylates”) 1 -14.7536 type-checking 1 -14.7536 379–385, 1 -14.7536 tlinks, 1 -14.7536 notagdictionary 1 -14.7536 disclaimers)or 1 -14.7536 comparativenewssummarizationaimstohighlight 1 -14.7536 sequel, 1 -14.7536 conjuncts, 1 -14.7536 tested-half 1 -14.7536 arewidelyusedbyexistingnersystemstoevaluate 1 -14.7536 usingthesamemethod,wefound18incoherentgraphs 1 -14.7536 nbar 1 -14.7536 martinson, 1 -14.7536 4typically, 1 -14.7536 conll’02. 1 -14.7536 n(u) 1 -14.7536 fbk-irst. 1 -14.7536 folklore#n#1 1 -14.7536 wi2006. 1 -14.7536 jeremiahezekiel 1 -14.7536 forq 1 -14.7536 sweep, 1 -14.7536 glarf-style 1 -14.7536 experiential: 1 -14.7536 in-spite 1 -14.7536 error-detection, 1 -14.7536 willshowhowtheexpectationsofanagent-verbpair 1 -14.7536 re-exerts 1 -14.7536 583–593, 1 -14.7536 thus,this 1 -14.7536 traded 1 -14.7536 there688 1 -14.7536 x@p 1 -14.7536 silviu 1 -14.7536 uppertailed 1 -14.7536 1500–1511, 1 -14.7536 7thanks 1 -14.7536 aggressivemerge 1 -14.7536 shopping) 1 -14.7536 p<0.05 1 -14.7536 amsterdam/philadelphia: 1 -14.7536 itlackstheflexibilityof 1 -14.7536 dialectonly 1 -14.7536 svore 1 -14.7536 host/chair 1 -14.7536 naacl’06. 1 -14.7536 1425–1434, 1 -14.7536 alvisnlp/ml 1 -14.7536 inordertoevaluatetheeffectivenessofourproposed 1 -14.7536 wereport 1 -14.7536 modality=“andare”> 1 -14.7536 non-standard; 1 -14.7536 -constraint 1 -14.7536 “turkey 1 -14.7536 138147. 1 -14.7536 exp6: 1 -14.7536 tokenizer.perl 1 -14.7536 semlink3 1 -14.7536 797.42% 1 -14.7536 psycho-biology 1 -14.7536 70–78, 1 -14.7536 theresultsshow 1 -14.7536 definitenounphrase:comparedto 1 -14.7536 wecanassume 1 -14.7536 revu-overall 1 -14.7536 boldin 1 -14.7536 medievallanguage,alexicalanalyserwasaddedinclud- 1 -14.7536 +1.3 1 -14.7536 hvs-based 1 -14.7536 thosearecurrentlydownloadableforenglish, 1 -14.7536 sc(dr) 1 -14.7536 pos–tagging 1 -14.7536 http://www.isi.edu/natural- 1 -14.7536 inappropriateness 1 -14.7536 533–541, 1 -14.7536 lightweight, 1 -14.7536 2www.cis.udel.edu/~carberry/graphs/viewallgraphs.php 1 -14.7536 tucker 1 -14.7536 verle 1 -14.7536 motivationofusingsucha 1 -14.7536 “textual 1 -14.7536  impression: 1 -14.7536 interspeech. 1 -14.7536 :)” 1 -14.7536 coordinative 1 -14.7536 (arbitrarily) 1 -14.7536 sure: 1 -14.7536 earlier,malt 1 -14.7536 inthisexperiment,weproposetousestrtmtoaddress 1 -14.7536 educators, 1 -14.7536 -f. 1 -14.7536 edu/˜myleott/op_spam 1 -14.7536 surprising—our 1 -14.7536 qualityoftranslationsisevaluatedbythe 1 -14.7536 advanced, 1 -14.7536 <114> 1 -14.7536 wehaveformallyproventhatwithinthislargesearch 1 -14.7536 bold): 1 -14.7536 jpda 1 -14.7536 reanalyses 1 -14.7536 ciated 1 -14.7536 pc4, 1 -14.7536 schema_cluster_matchn 1 -14.7536 intoasequenceofcardsegments,inwhichplayersdiscusswhichcardtodiscard,or 1 -14.7536 145–151, 1 -14.7536 kepat: 1 -14.7536 modestsized 1 -14.7536 1but 1 -14.7536 (alm 1 -14.7536 bachman’s 1 -14.7536 “and”-splicing. 1 -14.7536 eeg-signal 1 -14.7536 103316 1 -14.7536 richnes 1 -14.7536 hiearchical 1 -14.7536 7work 1 -14.7536 apartofevaluating 1 -14.7536 non-trivial: 1 -14.7536 (b-input), 1 -14.7536 cessfully 1 -14.7536 dagstuhl 1 -14.7536 wmt08n) 1 -14.7536 ofsentences 1 -14.7536 experiments,each 1 -14.7536 nus-i2r: 1 -14.7536 34(1):35–80. 1 -14.7536 pluperfect 1 -14.7536 enulli 1 -14.7536 manysetexpansionalgorithmsarebasedonbootstrapping 1 -14.7536 riensche 1 -14.7536 tocombineresponses,wetookthemodalresponse 1 -14.7536 thetypologiesthatcomewith 1 -14.7536 primitivesindicateifacertain 1 -14.7536 (2005)3 1 -14.7536 phrase-hit(1) 1 -14.7536 1616–1625, 1 -14.7536 titovetal.(2008)exploredatopicmodelingmethod 1 -14.7536 usethetermvocabularylistsynonymouslyforawordfrequency 1 -14.7536 experiences.) 1 -14.7536 930–940, 1 -14.7536 inseam 1 -14.7536 “i-tweet 1 -14.7536 medco-b, 1 -14.7536 simmetrics 1 -14.7536 arnott 1 -14.7536 annotateddataset,andshowthatourmethodoutper- 1 -14.7536 performancesuggested 1 -14.7536 weightsware 1 -14.7536 training-set 1 -14.7536 chicken-and-egg 1 -14.7536 suggeststhatthestructuredrolerepresentationcould 1 -14.7536 range.n, 1 -14.7536 ordinals– 1 -14.7536 implemented5 1 -14.7536 replace(ns, 1 -14.7536 8workshop 1 -14.7536 tense: 1 -14.7536 cyan, 1 -14.7536 pennsylvania: 1 -14.7536 blinking 1 -14.7536 tigate 1 -14.7536 130–137, 1 -14.7536 however,document 1 -14.7536 chen2010all. 1 -14.7536 s± 1 -14.7536 49–58, 1 -14.7536 〈s3,c3〉: 1 -14.7536 7,025,587 1 -14.7536 anintuitionthatexperimentaldevices,whichwillbe 1 -14.7536 fastandaccuratejointinferenceisprovidedby 1 -14.7536 .so 1 -14.7536 “character-based” 1 -14.7536 97–104, 1 -14.7536 tw(i,j) 1 -14.7536 superuser’s 1 -14.7536 “jp+fr+de” 1 -14.7536 “data-driven 1 -14.7536 nips. 1 -14.7536 fλx.f(x)∧to(x,y) 1 -14.7536 objectofverb, 1 -14.7536 8(3):423-459. 1 -14.7536 thematrixof 1 -14.7536 tsunakawa 1 -14.7536 matchvectors 1 -14.7536 wordalignments 1 -14.7536 fig.3illustratesthedictionaryextractionoperator 1 -14.7536 agreement-trained 1 -14.7536 x▽z(p(pii)) 1 -14.7536 troubling, 1 -14.7536 outperfoms 1 -14.7536 parameterψc,t 1 -14.7536 ds-cascade: 1 -14.7536 newark: 1 -14.7536 are→is) 1 -14.7536 fication, 1 -14.7536 transgressive 1 -14.7536 030 1 -14.7536 #query 1 -14.7536 lowbow, 1 -14.7536 succinctness, 1 -14.7536 10application-specific 1 -14.7536 mt-summit 1 -14.7536 probabilitypa(w) 1 -14.7536 constructs±∞are 1 -14.7536 (α’s) 1 -14.7536 punctuate 1 -14.7536 496–500, 1 -14.7536 'let 1 -14.7536 temperature(summer) 1 -14.7536 this705 1 -14.7536 130–141, 1 -14.7536 86–91, 1 -14.7536 vinnova2 1 -14.7536 casesthatamodifierconnectstomorethanoneopinion 1 -14.7536 referee) 1 -14.7536 labourd) 1 -14.7536 θseq 1 -14.7536 apropertyhavingdirect 1 -14.7536 strength; 1 -14.7536 penicillin) 1 -14.7536 thiserrormodelhasoneparameterforevery 1 -14.7536 uence 1 -14.7536 thesecostsarenotcorrect, 1 -14.7536 multivalued 1 -14.7536 trees.1 1 -14.7536 expl(ain) 1 -14.7536 pafel, 1 -14.7536 frame), 1 -14.7536 score(n2,n2′) 1 -14.7536 wn-fn 1 -14.7536 skip+ 1 -14.7536 match(d1,d2) 1 -14.7536 t2t 1 -14.7536 delexicalizing 1 -14.7536 −is 1 -14.7536 440-447, 1 -14.7536 pm1 1 -14.7536 toy-scale 1 -14.7536 eration 1 -14.7536 weextractasfeaturesanywordsintheinputsentences 1 -14.7536 sentence-beginning 1 -14.7536 con68 1 -14.7536 sumo-based 1 -14.7536 forexample,mullen 1 -14.7536 700–711, 1 -14.7536 “fullgraph,” 1 -14.7536 ceptron 1 -14.7536 pregroups 1 -14.7536 inwhichweadd 1 -14.7536 herman’s 1 -14.7536 .981** 1 -14.7536 mision, 1 -14.7536 (judges) 1 -14.7536 mainlyfollowing 1 -14.7536 attribute-value), 1 -14.7536 tutions, 1 -14.7536 language-pairdependentmethodscarryovertoother 1 -14.7536 syntac115 1 -14.7536 non-bigrams 1 -14.7536 pursuant 1 -14.7536 max=0.9) 1 -14.7536 3mle 1 -14.7536 3less 1 -14.7536 555–596, 1 -14.7536 (p,a,e) 1 -14.7536 (cws), 1 -14.7536 mid(v,per) 1 -14.7536 rerankingmodelsaregenerativeversionsofkooand 1 -14.7536 tobi-style 1 -14.7536 pings 1 -14.7536 perform∼109 1 -14.7536 1k. 1 -14.7536 defendant 1 -14.7536 revalidates 1 -14.7536 manually-added 1 -14.7536 creasing 1 -14.7536 homologene4 1 -14.7536 .11). 1 -14.7536 svm_light/ 1 -14.7536 pos-tags: 1 -14.7536 5pitler 1 -14.7536 emlp 1 -14.7536 ganter 1 -14.7536 sakaki 1 -14.7536 idicates 1 -14.7536 o(max(u,v)3) 1 -14.7536 152–161, 1 -14.7536 sood 1 -14.7536 name-mining 1 -14.7536 theinitialdictionarycontained 1 -14.7536 noninvasive 1 -14.7536 4concatenating 1 -14.7536 memory-wise, 1 -14.7536 36.5m 1 -14.7536 wefindthat 1 -14.7536 samardˇzi´c 1 -14.7536 neighbor-of-a-neighbor: 1 -14.7536 annesi 1 -14.7536 srilm—an 1 -14.7536 smuggle 1 -14.7536 hm-decoding 1 -14.7536 theirmethodisalso 1 -14.7536 indistinguishability 1 -14.7536 witnesed 1 -14.7536 tokenlevel 1 -14.7536 1222–1233, 1 -14.7536 apertium, 1 -14.7536 ourcomparativeanalysisofchineseandenglishtypos 1 -14.7536 4aligning 1 -14.7536 (knight,1999)showedthattspisasub-classofmt 1 -14.7536 φjk(xi) 1 -14.7536 tutorp6: 1 -14.7536 concisely,we 1 -14.7536 instancecentered 1 -14.7536 dugast 1 -14.7536 ness 1 -14.7536 24(2):123–140. 1 -14.7536 (rei 1 -14.7536 agnostic, 1 -14.7536 continuation: 1 -14.7536 pro8http://www.cs.berkeley.edu/˜pliang/ 1 -14.7536 (iccpol-2010), 1 -14.7536 concurent 1 -14.7536 sessing 1 -14.7536 accept.02; 1 -14.7536 graph-based, 1 -14.7536 -nkcp 1 -14.7536 finally,we 1 -14.7536 (germesin 1 -14.7536 re-investigate 1 -14.7536 searching... 1 -14.7536 test-reveals-problem 1 -14.7536 lengthening-aware 1 -14.7536 mostfreq-base 1 -14.7536 a0predicate 1 -14.7536 performedby 1 -14.7536 473–478, 1 -14.7536 battig, 1 -14.7536 kittler 1 -14.7536 wsdm 1 -14.7536 lexique~2: 1 -14.7536 44.9% 1 -14.7536 133–138, 1 -14.7536 108k 1 -14.7536 bermingham, 1 -14.7536 expected-log-gain 1 -14.7536 399–404, 1 -14.7536 predicate造成in 1 -14.7536 pdt, 1 -14.7536 44–49, 1 -14.7536 3.in 1 -14.7536 (2010)5 1 -14.7536 1-135 1 -14.7536 wealsoplantoconductexperiments 1 -14.7536 “non-null” 1 -14.7536 〈0,+ept,-ep,-wh〉(0022201,002221,002222,0022200,002200),〈1,=t+wh≈n〉(00221),〈0,=dv〉(20,21,23), 1 -14.7536 carbon 1 -14.7536 pinyincorrection 1 -14.7536 unlinked, 1 -14.7536 proposetodirectlymeasurethequeryimportanceinstead 1 -14.7536 hdt; 1 -14.7536 (0.14%) 1 -14.7536 unfortunificantly 1 -14.7536 [2007], 1 -14.7536 differentfromtheprevious 1 -14.7536 otero, 1 -14.7536 jindalandliu(2006)studiedtheproblemofidentifying 1 -14.7536 c[s:e] 1 -14.7536 rolesent 1 -14.7536 minitacitus 1 -14.7536 (on-line), 1 -14.7536 thecollectionofanalyzedtextsconsists,semantically, 1 -14.7536 top-2000 1 -14.7536 wl/arrival 1 -14.7536 3070 1 -14.7536 17since 1 -14.7536 kageura 1 -14.7536 proximity: 1 -14.7536 ==": 1 -14.7536 3splitting 1 -14.7536 showsthereferencetranslationwithreconstructedsemantic 1 -14.7536 (1964), 1 -14.7536 28(3):245–288. 1 -14.7536 calf) 1 -14.7536 pages:440-447. 1 -14.7536 suchverycommon 1 -14.7536 genarator 1 -14.7536 6,590 1 -14.7536 keyfeatures 1 -14.7536 subtype), 1 -14.7536 rsc 1 -14.7536 prefaces 1 -14.7536 backtranslation 1 -14.7536 reciprocally, 1 -14.7536 cyc: 1 -14.7536 high-density 1 -14.7536 paraphraseevaluationmetrics 1 -14.7536 experimentedwithanumberofwaystocombinethefivehuman 1 -14.7536 lexiconmodels 1 -14.7536 5similar 1 -14.7536 §2.2 1 -14.7536 accord154 1 -14.7536 wikinyt, 1 -14.7536 00-034. 1 -14.7536 state-of-the-artmt 1 -14.7536 2http://sourceforge.net/projects/ngram/ 1 -14.7536 onlyoneanswer 1 -14.7536 shift-x 1 -14.7536 rellda1 1 -14.7536 gareraandyarowsky(2008)proposeanapproach 1 -14.7536 wsj+biomed+n-gram 1 -14.7536 (bls) 1 -14.7536 “anonymous”. 1 -14.7536 area’s 1 -14.7536 4thater 1 -14.7536 prefixes 1 -14.7536 plmrf, 1 -14.7536 re-implementation, 1 -14.7536 moses+del: 1 -14.7536 antecedent-pronoun 1 -14.7536 7from 1 -14.7536 60925008) 1 -14.7536 1notice 1 -14.7536 contrast,longer 1 -14.7536 set1390 1 -14.7536 nercb−knn 1 -14.7536 tac-09. 1 -14.7536 b+ti+c+pi 1 -14.7536 19see 1 -14.7536 501–511, 1 -14.7536 92–98, 1 -14.7536 con7provided 1 -14.7536 15a 1 -14.7536 0.08% 1 -14.7536 φdice 1 -14.7536 (grl) 1 -14.7536 3artstein 1 -14.7536 pro592 1 -14.7536 uncertaintyadaptivesystemyielded120dialogues(1957student 1 -14.7536 userguided 1 -14.7536 i-j-same-sentdiff-clause=true, 1 -14.7536 48–57, 1 -14.7536 abstractive: 1 -14.7536 romera 1 -14.7536 288–293, 1 -14.7536 11here, 1 -14.7536 clearlydemonstratethatthelookaheadsearchboosts 1 -14.7536 differentvaluesofαandreportonlythebestaccu- 1 -14.7536 .18. 1 -14.7536 leters 1 -14.7536 atranslationdictionary 1 -14.7536 (s\np)/np, 1 -14.7536 (kanaris 1 -14.7536 elling 1 -14.7536 3due 1 -14.7536 shallow-semantics 1 -14.7536 ‘labeled’ 1 -14.7536 nle 1 -14.7536 ofρrange 1 -14.7536 acessibility 1 -14.7536 hypothesisshouldbe 1 -14.7536 514–518, 1 -14.7536 r,x)∧state(x), 1 -14.7536 http:/ww.medilexicon.com 1 -14.7536 5641 1 -14.7536 pos_out: 1 -14.7536 24.11% 1 -14.7536 weusefactored 1 -14.7536 context(k) 1 -14.7536 nn-train 1 -14.7536 fn(m′;m) 1 -14.7536 1-tweet 1 -14.7536 stituent 1 -14.7536 game.in 1 -14.7536 titles/summaries 1 -14.7536 “fullgraph” 1 -14.7536 finalclassifierbyusingtheenrichedtraining 1 -14.7536 expatiatory 1 -14.7536 mada-produced 1 -14.7536 recscore, 1 -14.7536 74–81, 1 -14.7536 opaqueness 1 -14.7536 (0.270), 1 -14.7536 adaptations: 1 -14.7536 “unsupervised,” 1 -14.7536 hr0011-08-c-0004 1 -14.7536 (56.3 1 -14.7536 “injected” 1 -14.7536 (14)) 1 -14.7536 873 1 -14.7536 35.70 1 -14.7536 drawback—they 1 -14.7536 medium-frequency 1 -14.7536 states(wn1) 1 -14.7536 bi-normal 1 -14.7536 first-step 1 -14.7536 ’d 1 -14.7536 445–449, 1 -14.7536 null2.88; 1 -14.7536 objective/neutral. 1 -14.7536 combinedsystem:classiferplus 1 -14.7536 pre-transforming 1 -14.7536 6based 1 -14.7536 1 -14.7536 allmodelswere 1 -14.7536 rouge-4 1 -14.7536 speculation: 1 -14.7536 resolu809 1 -14.7536 (2000, 1 -14.7536 “chester”), 1 -14.7536 (declarative 1 -14.7536 respect) 1 -14.7536 ―list‖ 1 -14.7536 buses, 1 -14.7536 a150a151a152a143a144a155a156 1 -14.7536 5tb: 1 -14.7536 showsdifferenttypesofscoringpartsusedincurrent 1 -14.7536 thismeans 1 -14.7536 (bf): 1 -14.7536 #iis-0910884. 1 -14.7536 -swdp 1 -14.7536 un-able 1 -14.7536 98-07, 1 -14.7536 on), 1 -14.7536 output:h∗,h∗: 1 -14.7536 printer) 1 -14.7536 dict(y)returns 1 -14.7536 gwc2010, 1 -14.7536 20150 1 -14.7536 experimentable 1 -14.7536 interaction-0, 1 -14.7536 1http://www.kyoto-project.eu/ 1 -14.7536 thisisbecause 1 -14.7536 (12,349) 1 -14.7536 cr: 1 -14.7536 20(2):233–287. 1 -14.7536 kaminskas, 1 -14.7536 lazzyff 1 -14.7536 drop-offs 1 -14.7536 ruleid 1 -14.7536 controversy: 1 -14.7536 ”awesome” 1 -14.7536 slowdown, 1 -14.7536 618 1 -14.7536 interestingness: 1 -14.7536 cussed 1 -14.7536 #5–6 1 -14.7536 cvem 1 -14.7536 free-style 1 -14.7536 2integer 1 -14.7536 24.8%—comparable 1 -14.7536 data-collection 1 -14.7536 (minimum-cost) 1 -14.7536 4perplexity 1 -14.7536 discordances 1 -14.7536 ’stem’ 1 -14.7536 (right:) 1 -14.7536 ‘learning 1 -14.7536 dis-allowing 1 -14.7536 sent/sec* 1 -14.7536 mwetoolkitandthecfgfromthewhichthetsg 1 -14.7536 338–340, 1 -14.7536 ‘=’) 1 -14.7536 extractionof 1 -14.7536 drown, 1 -14.7536 score1 1 -14.7536 stephvent 1 -14.7536 inflective 1 -14.7536 m=summationtext 1 -14.7536 4hinrichs 1 -14.7536 part-of-speechinthetargettranslationisa. 1 -14.7536 1interpolating 1 -14.7536 14a 1 -14.7536 hashingbased 1 -14.7536 stmt(a2,a1,¬at(a2,home),yet) 1 -14.7536 ehn@cs.cmu.edu 1 -14.7536 mis-selected 1 -14.7536 monoboot: 1 -14.7536 mawat, 1 -14.7536 blairgoldensohn 1 -14.7536 reasons; 1 -14.7536 under246 1 -14.7536 individualization: 1 -14.7536 nonetheles, 1 -14.7536 dollaghan 1 -14.7536 cfrs 1 -14.7536 17figure 1 -14.7536 intrinsic: 1 -14.7536 (context-insensitive) 1 -14.7536 setc⊆{1,2,...,n}, 1 -14.7536 thus,most 1 -14.7536 michel.gagnon@polymtl.ca 1 -14.7536 eukaryotes, 1 -14.7536 fill-in-the-blank: 1 -14.7536 “logical” 1 -14.7536 1276 1 -14.7536 92,415 1 -14.7536 3,163 1 -14.7536 45.82 1 -14.7536 coling-08. 1 -14.7536 syntax-guided 1 -14.7536 intractable; 1 -14.7536 4including 1 -14.7536 bag-ofword 1 -14.7536 factorgraph 1 -14.7536 fx2, 1 -14.7536 website6 1 -14.7536 programmingbased 1 -14.7536 (mean±standard 1 -14.7536 coincided 1 -14.7536 formalizedalgorithmand 1 -14.7536 11an 1 -14.7536 o-coverage). 1 -14.7536 dev.) 1 -14.7536 94,448 1 -14.7536 distsim: 1 -14.7536 warnke 1 -14.7536 thishypothesis 1 -14.7536 word-overlap 1 -14.7536 62–72, 1 -14.7536 item-types 1 -14.7536 metaph: 1 -14.7536 naradowsky 1 -14.7536 1557–1567, 1 -14.7536 agr0 1 -14.7536 15,689 1 -14.7536 synecdoche: 1 -14.7536 much-desired 1 -14.7536 ∀uv 1 -14.7536 reproductions 1 -14.7536 have)utterancesintaskinterruptions, 1 -14.7536 fromthesesentenceswekeeponlythosethatcontain 1 -14.7536 lin’s: 1 -14.7536 second(“other-unmatched”)containssentencesinwhich 1 -14.7536 3.798e-08 1 -14.7536 choinicolov’s 1 -14.7536 thepubliclyavailableeuroparlcorpus 1 -14.7536 [jj+nn+in]. 1 -14.7536 svm-hmm3 1 -14.7536 weassumethiscouldinprinciplebedoneautomatically, 1 -14.7536 42–47, 1 -14.7536 multi-decision 1 -14.7536 csr: 1 -14.7536 8(a) 1 -14.7536 ‘time’ 1 -14.7536 be8applying 1 -14.7536 corplex/treetagger/ 1 -14.7536 es→en, 1 -14.7536 commensurable 1 -14.7536 downspeak: 1 -14.7536 domain-independent, 1 -14.7536 wikify!: 1 -14.7536 (“slots”) 1 -14.7536 correlations,the 1 -14.7536 entities): 1 -14.7536 ‘alien⇒space’ 1 -14.7536 (zhikov 1 -14.7536 qris 1 -14.7536 κstatistics 1 -14.7536 10–21, 1 -14.7536 knp 1 -14.7536 dunning, 1 -14.7536 899–908, 1 -14.7536 (article) 1 -14.7536 twitter_nlp 1 -14.7536 thesemanticclassofmembers 1 -14.7536 715–719, 1 -14.7536 i!j; 1 -14.7536 weevaluateourmodelsonseveraltracksofthe2009 1 -14.7536 languagesareintrinsicallycohesiveandcoherent, 1 -14.7536 fxd-gap+ 1 -14.7536 intruded: 1 -14.7536 possprn: 1 -14.7536 “four”) 1 -14.7536 extractive: 1 -14.7536 mweextracting 1 -14.7536 dohsaka 1 -14.7536 inter-annotatoragreement 1 -14.7536 americas. 1 -14.7536 hmeantusingtheimprovedstructuredrolerepresentation 1 -14.7536 869–879, 1 -14.7536 (easier) 1 -14.7536 -csbp 1 -14.7536 t−h: 1 -14.7536 rewritting 1 -14.7536 [12] 1 -14.7536 applicabletothebigrammodelowingtosparsityand 1 -14.7536 (1∗)) 1 -14.7536 formatting-related 1 -14.7536 sentences:1 1 -14.7536 [rock]a2 1 -14.7536 cessive 1 -14.7536 adependencyparse,asalreadyindicated,isthe 1 -14.7536 human-labels 1 -14.7536 aguilar 1 -14.7536 n(wt,si) 1 -14.7536 fromthis,wecomputedascorebyaveragingover 1 -14.7536 isavalidtree 1 -14.7536 thestructureofadiscourseisacombinationoflinguisticstructure,intentionalstructure, 1 -14.7536 nullnullo(w) 1 -14.7536 influx 1 -14.7536 sub-processes: 1 -14.7536 2lmi 1 -14.7536 5.2.3), 1 -14.7536 douban.com, 1 -14.7536 vignet’s 1 -14.7536 uk/s0453356/share 1 -14.7536 {yij} 1 -14.7536 chored 1 -14.7536 ,phrasesu 1 -14.7536 (zweigenbaum 1 -14.7536 2.1.1, 1 -14.7536 kiyoki 1 -14.7536 examinees’ 1 -14.7536 (s[wq]/s[inv])/n 1 -14.7536 yahoo!japan 1 -14.7536 timent 1 -14.7536 6,891 1 -14.7536 de-emphasizing 1 -14.7536 bleucompared 1 -14.7536 hlt-naacl-07. 1 -14.7536 136–145, 1 -14.7536 federalist. 1 -14.7536 setmt(at) 1 -14.7536 cateogories 1 -14.7536 maybe, 1 -14.7536 (omaprox2), 1 -14.7536 reordering-specific 1 -14.7536 1-(a) 1 -14.7536 perfor2we 1 -14.7536 4provided 1 -14.7536 nopred 1 -14.7536 label-propagated 1 -14.7536 780–788. 1 -14.7536 1623 1 -14.7536 ={0,1}. 1 -14.7536 hypernyms: 1 -14.7536 ‘match(k)’ 1 -14.7536 [fluency] 1 -14.7536 “mrf” 1 -14.7536 csrl 1 -14.7536 1531 1 -14.7536 8http://pascallin.ecs.soton.ac.uk/challenges/rte2/ 1 -14.7536 712–721, 1 -14.7536 clc-download 1 -14.7536 reconcile-internalscorers 1 -14.7536 lind 1 -14.7536 paradoxical 1 -14.7536 precaution, 1 -14.7536 183–191, 1 -14.7536 4.1length 1 -14.7536 wasp−1++, 1 -14.7536 methodhadverylittleinfluenceonthetestsetscores 1 -14.7536 (isr-wn) 1 -14.7536 thisproblemisanoptimizationproblem 1 -14.7536 attribute: 1 -14.7536 151–161, 1 -14.7536 liwc+. 1 -14.7536 linguatec’s 1 -14.7536 δ1,δ2 1 -14.7536 toensurethatwe 1 -14.7536 “word-tree” 1 -14.7536 detect/correct 1 -14.7536 precision-trained 1 -14.7536 mecab2, 1 -14.7536 r(r)s(s) 1 -14.7536 9,d′s 1 -14.7536 acl02, 1 -14.7536 table2.we 1 -14.7536 76.29% 1 -14.7536 9wordnet 1 -14.7536 isomer 1 -14.7536 thispaperisconcernedwiththeautomaticrecognitionofdialogueactsbasedonpartiallyavailable 1 -14.7536 stan-annotation, 1 -14.7536 clc, 1 -14.7536 t13) 1 -14.7536 therefor 1 -14.7536 "gold" 1 -14.7536 nullnullj. 1 -14.7536 vmm’s 1 -14.7536 0.2n. 1 -14.7536 1f 1 -14.7536 inter_cfg 1 -14.7536 sumption 1 -14.7536 nontensed 1 -14.7536 (candidatecount) 1 -14.7536 (206 1 -14.7536 p(vectorθ),p(−→φt),p(αt),p(αprimet) 1 -14.7536 graphtheoretic 1 -14.7536 finally,sinceoursystemusesaninitialdictionary,whichwetriedtominimizebyonlyhaving 1 -14.7536 3bipartite 1 -14.7536 (2009)usemitriggersintheirconfidence 1 -14.7536 losscost 1 -14.7536 bag-of-visualwords 1 -14.7536 halts, 1 -14.7536 393–398, 1 -14.7536 interestrate, 1 -14.7536 57–62, 1 -14.7536 1grawltcq 1 -14.7536 421–432, 1 -14.7536 “becoming” 1 -14.7536 swat 1 -14.7536 subsetdof 1 -14.7536 went→go), 1 -14.7536 instill 1 -14.7536 .60). 1 -14.7536 98–106, 1 -14.7536 24–28, 1 -14.7536 aswiththepreviouslymentioned 1 -14.7536 iis0910664, 1 -14.7536 thetagisusedtoannotate 1 -14.7536 301–306, 1 -14.7536 (andrade 1 -14.7536 72.4* 1 -14.7536 261–271, 1 -14.7536 thisrepresentationismotivated 1 -14.7536 log-factor, 1 -14.7536 865–874, 1 -14.7536 orange: 1 -14.7536 condenses 1 -14.7536 42:47-54. 1 -14.7536 "null 1 -14.7536 (repeatedly) 1 -14.7536 paperpresentsamethodofiterativelycoarsening 1 -14.7536 participants), 1 -14.7536 a... 1 -14.7536 4both 1 -14.7536 tp+tn2004 1 -14.7536 re-weigh 1 -14.7536 2.6.5.2 1 -14.7536 τr−2 1 -14.7536 1998)) 1 -14.7536 combine-x 1 -14.7536 (swsdgold) 1 -14.7536 uri4 1 -14.7536 co.uk) 1 -14.7536 220–229, 1 -14.7536 5here 1 -14.7536 rcv1: 1 -14.7536 (section4)thatourmatrix-spacemodelsignificantly 1 -14.7536 impartiality, 1 -14.7536 5unlike 1 -14.7536 585-592. 1 -14.7536 9part 1 -14.7536 insufficiencies 1 -14.7536 nist03 1 -14.7536 sentimentspecific 1 -14.7536 sub-collection. 1 -14.7536 lexeme-template 1 -14.7536 59.44 1 -14.7536 “cheatingexperiment”in 1 -14.7536 offprompt 1 -14.7536 dpphr 1 -14.7536 np-cluster 1 -14.7536 data-set, 1 -14.7536 mwetoolkit/ 1 -14.7536 beinteresting 1 -14.7536 pitch: 1 -14.7536 mate 1 -14.7536 21(2):203–225. 1 -14.7536 39:189-196. 1 -14.7536 weselectednormalized 1 -14.7536 non-1-to-1 1 -14.7536 surprisingasthereisaclearmathematicaldeficiency 1 -14.7536 looks, 1 -14.7536 part-fac(y) 1 -14.7536 8.074, 1 -14.7536 expressivity-efficiency 1 -14.7536 linguistically-annotated 1 -14.7536 esther 1 -14.7536 preprocessng 1 -14.7536 givenatweetasinput,ourtaskistoidentifyboththe 1 -14.7536 weexperiment 1 -14.7536 dot/dash 1 -14.7536 r(eprime,e) 1 -14.7536 (raina 1 -14.7536 0null2 1 -14.7536 transcriber: 1 -14.7536 ei: 1 -14.7536 .ser(727) 1 -14.7536 1).3 1 -14.7536 evaluatedbymeansofconfidenceweightedscore(cws), 1 -14.7536 cricticism 1 -14.7536 20-way 1 -14.7536 02] 1 -14.7536 restatements 1 -14.7536 entity-destination: 1 -14.7536 parser/dependency 1 -14.7536 “upper-bound”, 1 -14.7536 modelensuresconsistencybetweenarguments 1 -14.7536 1373–1383, 1 -14.7536 8000-instance 1 -14.7536 maxall 1 -14.7536 tokeniser 1 -14.7536 1583–1592, 1 -14.7536 lingua::en::stopwords 1 -14.7536 1012–1022, 1 -14.7536 jp-en 1 -14.7536 emotiblog: 1 -14.7536 connectivewise 1 -14.7536 tokens— 1 -14.7536 hand,an 1 -14.7536 “e-business” 1 -14.7536 cokaine, 1 -14.7536 graphically, 1 -14.7536 420–425, 1 -14.7536 expected-loggain, 1 -14.7536 queryinthetaskisassociatedwithanamestringand 1 -14.7536 boosters3 1 -14.7536 back-proped 1 -14.7536 (173 1 -14.7536 originalverb,tensentencescontainingthatverbwith 1 -14.7536 non-binary 1 -14.7536 scriptbased 1 -14.7536 zburg, 1 -14.7536 iwcs-5, 1 -14.7536 mitlm 1 -14.7536 baseline—i.e. 1 -14.7536 ap-proach 1 -14.7536 coling’02. 1 -14.7536 iwata 1 -14.7536 stroppa 1 -14.7536 denselypopulated 1 -14.7536 letxf 1 -14.7536 wealsoplantousedualdecompositiontoencourage 1 -14.7536 in-corpus 1 -14.7536 4smrž 1 -14.7536 theformulastocomputeentity 1 -14.7536 116–125, 1 -14.7536 k0 1 -14.7536 firthien 1 -14.7536 postedit 1 -14.7536 machinetagger 1 -14.7536 reversely-trained 1 -14.7536 sneiders 1 -14.7536 hyperbolic 1 -14.7536 ourexperimentsshowthatcombiningalltheseelements 1 -14.7536 “encode” 1 -14.7536 wmtsmall+pb 1 -14.7536 ijcnlp. 1 -14.7536 cyk-2 1 -14.7536 (heinrich, 1 -14.7536 s(e1,e2) 1 -14.7536 7degenerate 1 -14.7536 †lee 1 -14.7536 tradaboost, 1 -14.7536 43(2):161-179. 1 -14.7536 discriminated, 1 -14.7536 deterministically. 1 -14.7536 solvingtheatsp(whichisawellknownnphard 1 -14.7536 variationalapproximateframeworkcanbeappliedtoothertasks 1 -14.7536 (es+er+pc). 1 -14.7536 1-to1 1 -14.7536 out(vi)) 1 -14.7536 mert) 1 -14.7536 sourceconditional 1 -14.7536 themvariables 1 -14.7536 rule-learning 1 -14.7536 funql 1 -14.7536 mergingbothtrainingsetsisreasonable 1 -14.7536 franke 1 -14.7536 not-activate-labeled 1 -14.7536 calculatek 1 -14.7536 cl-lsi, 1 -14.7536 wethencombinethepoissonbased 1 -14.7536 asitssecond 1 -14.7536 sentence-pair, 1 -14.7536 (mesh36), 1 -14.7536 470–477, 1 -14.7536 13null 1 -14.7536 paragraph6 1 -14.7536 with↓. 1 -14.7536 ks·kp. 1 -14.7536 (2276 1 -14.7536 1066–1076, 1 -14.7536 chevelu 1 -14.7536 expected,as 1 -14.7536 bibliome, 1 -14.7536 (40%), 1 -14.7536 owed 1 -14.7536 (left: 1 -14.7536 534–539, 1 -14.7536 .all, 1 -14.7536 experience”, 1 -14.7536 furtherpresentaneffectivepruningstrategytore- 1 -14.7536 cachepipe: 1 -14.7536 parsings 1 -14.7536 43–52, 1 -14.7536 wordnet-affect, 1 -14.7536 conv(hi;h1...hn) 1 -14.7536 ana-maria 1 -14.7536 intra/inter-date 1 -14.7536 [experiment-v: 1 -14.7536 [louis 1 -14.7536 thisweighting, 1 -14.7536 inacompanionworkwedevelopanenhancementofthisoperator 1 -14.7536 tagdict: 1 -14.7536 vindicate 1 -14.7536 vectorseare 1 -14.7536 rious 1 -14.7536 90.1, 1 -14.7536 r5(s,p) 1 -14.7536 a-patterns, 1 -14.7536 bc), 1 -14.7536 resource-oriented 1 -14.7536 analysis-transfer-synthesis 1 -14.7536 15determining 1 -14.7536 ra, 1 -14.7536 kilgariff 1 -14.7536 3555 1 -14.7536 hmsvms 1 -14.7536 severe, 1 -14.7536 aschematicdescription 1 -14.7536 findthatitwouldbeextremelytime-consumingeven 1 -14.7536 cooooooooooooooollllllllllllll!!!!!!!!!!!!!! 1 -14.7536 info(h) 1 -14.7536 hbouamor@limsi.fr 1 -14.7536 popovi¸c 1 -14.7536 maxsim: 1 -14.7536 2,160 1 -14.7536 3,311 1 -14.7536 deeper. 1 -14.7536 ssel, 1 -14.7536 thecomputationalcomplexityofoptimalfactorizationwithtreedecompositionis 1 -14.7536 6–10, 1 -14.7536 (indi93 1 -14.7536 surfacestring 1 -14.7536 (weight 1 -14.7536 αvalue 1 -14.7536 tp+fp12) 1 -14.7536 classifer-based 1 -14.7536 top-frequent 1 -14.7536 s1-s3 1 -14.7536 w(c)hh 1 -14.7536 ijcnlp-05. 1 -14.7536 unionsq 1 -14.7536 de-capitalised 1 -14.7536 68.1%. 1 -14.7536 27th, 1 -14.7536 coling’ 1 -14.7536 nothing, 1 -14.7536 em-lp-n 1 -14.7536 second151 1 -14.7536 94.1. 1 -14.7536 aaai-caaw 1 -14.7536 wikiambi: 1 -14.7536 1213–1221, 1 -14.7536 rationalize 1 -14.7536 voyatzi, 1 -14.7536 pattsvm 1 -14.7536 31(4):477–504. 1 -14.7536 430–439, 1 -14.7536 (proux 1 -14.7536 non-lattice 1 -14.7536 +0.2 1 -14.7536 (best-source) 1 -14.7536 ldas 1 -14.7536 occam 1 -14.7536 projects/wikulu 1 -14.7536 freq(w,k) 1 -14.7536 venkatapathy, 1 -14.7536 (92:83%) 1 -14.7536 alsofromtable4weseethatthereisnoimprovementbyaddingmorefeatures, 1 -14.7536 (ml)-based 1 -14.7536 <16> 1 -14.7536 θ0. 1 -14.7536 961-968. 1 -14.7536 multiwordnet: 1 -14.7536 43) 1 -14.7536 b+tree 1 -14.7536 resource6 1 -14.7536 fb3(rs3: 1 -14.7536 mallet’s 1 -14.7536 "some" 1 -14.7536 deteriorated 1 -14.7536 early-update, 1 -14.7536 p−qp+q 1 -14.7536 contexts1 1 -14.7536 97.44% 1 -14.7536 http://glaros.dtc.umn.edu/gkhome/cluto/cluto/overview 1 -14.7536 representating 1 -14.7536 (n=317) 1 -14.7536 ppmi 1 -14.7536 eitrich, 1 -14.7536 13.2k 1 -14.7536 ak-dimensional 1 -14.7536 ”opennlp” 1 -14.7536 english.communications 1 -14.7536 {te,f}in 1 -14.7536 mrrs 1 -14.7536 includeany 1 -14.7536 91/100 1 -14.7536 that3 1 -14.7536 ”obj” 1 -14.7536 theseformsarethensubmittedas 1 -14.7536 sincethealignmentofsimplifiedsentences 1 -14.7536 mint: 1 -14.7536 outline: 1 -14.7536 ˜schmid/ 1 -14.7536 (evalv) 1 -14.7536 479–484, 1 -14.7536 xt=1(n)k 1 -14.7536 lan298 1 -14.7536 wordseye(coyneandsproat,2001),(coyneetal.,2010)isasystemforautomaticallyconvertingnatural 1 -14.7536 accidental, 1 -14.7536 damani@cse.iitb.ac.in 1 -14.7536 evalu10 1 -14.7536 span({}) 1 -14.7536 multi-feature 1 -14.7536 twitterusersareinterestedinnamedentities,such 1 -14.7536 303) 1 -14.7536 interactions”) 1 -14.7536 divide-by-zero 1 -14.7536 edges(u,v)∈e,thereexistsani∈iwithu,v∈x 1 -14.7536 9training 1 -14.7536 sampson, 1 -14.7536 5.6×5.9 1 -14.7536 rocio 1 -14.7536 manyoftheirworksincludingtheonebygeandmooney 1 -14.7536 pretence 1 -14.7536 similarpattern-basedapproaches 1 -14.7536 pseudo-label 1 -14.7536 theseproblemsareespeciallycrucialwhen 1 -14.7536 interactions”), 1 -14.7536 belief-propagation 1 -14.7536 surfacerealizationwith 1 -14.7536 pos-3gram, 1 -14.7536 thesecondproblem 1 -14.7536 itmayshowaninfluence 1 -14.7536 translingual 1 -14.7536 realm, 1 -14.7536 construction’s 1 -14.7536 parser/ 1 -14.7536 prp(x, 1 -14.7536 “attack”) 1 -14.7536 björneet 1 -14.7536 reductionfeature: 1 -14.7536 mwe: 1 -14.7536 somestudieshave 1 -14.7536 managers, 1 -14.7536 〈n1,n2〉 1 -14.7536 biordf 1 -14.7536 planning: 1 -14.7536 pairing, 1 -14.7536 mstislav 1 -14.7536 e6,binding 1 -14.7536 mirowski 1 -14.7536 tl], 1 -14.7536 added). 1 -14.7536 in§6, 1 -14.7536 harper 1 -14.7536 θd,l∼dir(α). 1 -14.7536 linguistics-2003, 1 -14.7536 1token 1 -14.7536 ictclas: 1 -14.7536 softmax-margin, 1 -14.7536 histf−1,lscript−j, 1 -14.7536 ijcnl 1 -14.7536 newsblaster2) 1 -14.7536 ψ(0)i 1 -14.7536 daß. 1 -14.7536 hand-vetting 1 -14.7536 textrank,expandrankperformsbetterwhenfacing 1 -14.7536 (las, 1 -14.7536 valid/invalid 1 -14.7536 baumwelch 1 -14.7536 eegolofssonandknutsson(2003)usedrule-basedmeth- 1 -14.7536 troduce 1 -14.7536 self-consistency), 1 -14.7536 cost(r): 1 -14.7536 pcj(wi) 1 -14.7536 sigurd, 1 -14.7536 multi-sitesystemsdevelopmentwhichmore 1 -14.7536 word-colour 1 -14.7536 bionlp11st-ge1 1 -14.7536 lose2 1 -14.7536 1643 1 -14.7536 pres.)=breche, 1 -14.7536 hand-full 1 -14.7536 ginsberg 1 -14.7536 feedback: 1 -14.7536 parameterλin 1 -14.7536 lnlr, 1 -14.7536 entities2 1 -14.7536 1546–1556, 1 -14.7536 3/9 1 -14.7536 vanhalteren 1 -14.7536 allergies: 1 -14.7536 wk4f+wn-4f, 1 -14.7536 616–627, 1 -14.7536 end,this 1 -14.7536 atlas.ti, 1 -14.7536 (camon 1 -14.7536 cellxij 1 -14.7536 1,898 1 -14.7536 in1according 1 -14.7536 3generating 1 -14.7536 2because, 1 -14.7536 cor1491 1 -14.7536 hand-built, 1 -14.7536 thesuggestedkeyphrases 1 -14.7536 thisisconfirmedbyobservingtheconstituent 1 -14.7536 joint-global: 1 -14.7536 preform 1 -14.7536 revilp. 1 -14.7536 institutional: 1 -14.7536 countoverlap(λ) 1 -14.7536 connotations. 1 -14.7536 zdi,kij 1 -14.7536 similar8 1 -14.7536 admissible) 1 -14.7536 anon-exhaustive 1 -14.7536 detmar meurers 1 -14.7536 label1. 1 -14.7536 gold2second-order, 1 -14.7536 “benign 1 -14.7536 finalreward:-18701.43 1 -14.7536 overweighting 1 -14.7536 (4d) 1 -14.7536 script-based 1 -14.7536 outgrowth 1 -14.7536 aparaphrasecandidatetendstobeavalidparaphrase 1 -14.7536 (globalvision 1 -14.7536 ficult” 1 -14.7536 taskofpredictingdocument-levelstarratingswas 1 -14.7536 p(word), 1 -14.7536 wizards' 1 -14.7536 “buzzing 1 -14.7536 maloof, 1 -14.7536 loaded, 1 -14.7536 tango: 1 -14.7536 (approximate)paraphrases. 1 -14.7536 ‡microsoft 1 -14.7536 sessions: 1 -14.7536 66.5* 1 -14.7536 ifyou’dliketotalkthepositionthatquotesfrom 1 -14.7536 273).1 1 -14.7536 verbnet: 1 -14.7536 carol: 1 -14.7536 confirst: 1 -14.7536 partiallyannotated 1 -14.7536 newbie 1 -14.7536 se/oe, 1 -14.7536 (c@1) 1 -14.7536 83.71% 1 -14.7536 four-condition 1 -14.7536 186–187, 1 -14.7536 51,680 1 -14.7536 1632 1 -14.7536 312-319. 1 -14.7536 tree-to-x 1 -14.7536 matures, 1 -14.7536 wasforthetac2010updatedocumentsets,similar 1 -14.7536 .38). 1 -14.7536 uppermost 1 -14.7536 acl-1989, 1 -14.7536 8if 1 -14.7536 path-toroot 1 -14.7536 (9/100) 1 -14.7536 1318–1326, 1 -14.7536 1298–1307, 1 -14.7536 cumbersome, 1 -14.7536 mysqldriven 1 -14.7536 561–569, 1 -14.7536 satis162 1 -14.7536 http://acl.ldc.upenn.edu/j/j93/j93-2004.pdf 1 -14.7536 pairwise: 1 -14.7536 “inconsistent” 1 -14.7536 neill, 1 -14.7536 iis0534700 1 -14.7536 bold: 1 -14.7536 49–59, 1 -14.7536 dispel 1 -14.7536 one-onone 1 -14.7536 corb´ıbellot 1 -14.7536 tuf), 1 -14.7536 experiment’s 1 -14.7536 optimally, 1 -14.7536 non-medical 1 -14.7536 laughs), 1 -14.7536 http://nlp.comp.nus. 1 -14.7536 phrasealigned 1 -14.7536 gtrec06 1 -14.7536 distributionality 1 -14.7536 ‘drug⇒medicine’, 1 -14.7536 177–186, 1 -14.7536 nondisambiguation 1 -14.7536 german-to-english. 1 -14.7536 wehavealso 1 -14.7536 cardinality, 1 -14.7536 1-4, 1 -14.7536 pinyin-correctionandchineseword-correctionpairs 1 -14.7536 lv-nounobj: 1 -14.7536 guidelines.pdf 1 -14.7536 litvak 1 -14.7536 situation-specific 1 -14.7536 joint-global-retrained: 1 -14.7536 landman 1 -14.7536 22(1):39-71. 1 -14.7536 (2)leftpositionalfamilysizeand 1 -14.7536 (1)agglomerative 1 -14.7536 (president) 1 -14.7536 1∑= 1 -14.7536 attardo 1 -14.7536 nelkenandshieberraisethisvalueto83.1% 1 -14.7536 aronson. 1 -14.7536 pfcseanavery 1 -14.7536 1310–1319, 1 -14.7536 hybridizations 1 -14.7536 bilingual/multilingual 1 -14.7536 8http://developer.yahoo.com/search/boss/ 1 -14.7536 over-emphasized 1 -14.7536 self-learner 1 -14.7536 similarily 1 -14.7536 187–197, 1 -14.7536 compression-specific 1 -14.7536 440–446, 1 -14.7536 classifier.5 1 -14.7536 entropy-pruning 1 -14.7536 csc, 1 -14.7536 zikanova 1 -14.7536 avenues, 1 -14.7536 2,911 1 -14.7536 (cicling), 1 -14.7536 particular〈t,t〉-lexicographic 1 -14.7536 contextual: 1 -14.7536 2figure 1 -14.7536 forty-four 1 -14.7536 1266 1 -14.7536 amulti-agentdomainwhereagentscompeteagainsteachotherinasimulatedsoccergame. 1 -14.7536 sadamitsu 1 -14.7536 dds 1 -14.7536 b-nn 1 -14.7536 yes/no: 1 -14.7536 mas: 1 -14.7536 eliminateda 1 -14.7536 -1.1 1 -14.7536 4,192 1 -14.7536 sptks, 1 -14.7536 s[wq]/n 1 -14.7536 node-internal 1 -14.7536 tac_kbp: 1 -14.7536 showthatthecorrelationofhmeant,thehuman 1 -14.7536 however,itisimpracticalforpeople 1 -14.7536 weseparatelycomparedictionary 1 -14.7536 273–283, 1 -14.7536 “hypergraph” 1 -14.7536 (18.c), 1 -14.7536 1283 1 -14.7536 iterates, 1 -14.7536 beamed 1 -14.7536 acl’07. 1 -14.7536 z2,3 1 -14.7536 scrapes 1 -14.7536 lindsey 1 -14.7536 treetag1http://hunspell.sourceforge.net/ 1 -14.7536 majority-class 1 -14.7536 0.709. 1 -14.7536 mwu-forests 1 -14.7536 135–139, 1 -14.7536 dependencystructuretothetargetsideviawordalignment, 1 -14.7536 hlt-naacl. 1 -14.7536 1058–1068, 1 -14.7536 twitterandothermicro-bloggingservicesarehighly 1 -14.7536 thestandardmethodfortreating 1 -14.7536 sense-appropriateness 1 -14.7536 bozapalidis 1 -14.7536 ‘divide-and-conquer’ 1 -14.7536 (concrete) 1 -14.7536 ungraspable 1 -14.7536 pre-processed, 1 -14.7536 path* 1 -14.7536 wrap-up 1 -14.7536 659–669, 1 -14.7536 null2.96; 1 -14.7536 outputsyare 1 -14.7536 iied 1 -14.7536 inthisarticle,wewishtoprovideageneralframeworkforfactorizationofdeductive 1 -14.7536 b= 1 -14.7536 ‘testavg’ 1 -14.7536 h¨aufig 1 -14.7536 temporallyoriented 1 -14.7536 hurum, 1 -14.7536 (brin, 1 -14.7536 (†: 1 -14.7536 to the learning algorithm (also svm) are simple  1 -14.7536 (§5.2), 1 -14.7536 parametersλi 1 -14.7536 magenta, 1 -14.7536 gtrec05 1 -14.7536 439–444, 1 -14.7536 diaconis 1 -14.7536 (aa) 1 -14.7536 http://twitter.com/about. 1 -14.7536 coli) 1 -14.7536 mendoza 1 -14.7536 fn) 1 -14.7536 hand-vetted 1 -14.7536 initializaion 1 -14.7536 clickable 1 -14.7536 16100 1 -14.7536 si∈s, 1 -14.7536 tuitively, 1 -14.7536 insteadofusingtheprosodicmodelinthefirstpassde- 1 -14.7536 tsvmm 1 -14.7536 mtis 1 -14.7536 ££description 1 -14.7536 φ(w,i,j) 1 -14.7536 (wl,tl,gl,d) 1 -14.7536 empha498 1 -14.7536 oval 1 -14.7536 http://www.nlm.nih.gov/mesh/ 1 -14.7536 7due 1 -14.7536 weproposeaformalisationwhichcaptures 1 -14.7536 radday 1 -14.7536 35-40, 1 -14.7536 wallonie-bruxelles 1 -14.7536 28.90 1 -14.7536 ethnologue 1 -14.7536 2]), 1 -14.7536 macrotextual 1 -14.7536 soricut, 1 -14.7536 compare7 1 -14.7536 3,391 1 -14.7536 multi-headedness: 1 -14.7536 biweekly 1 -14.7536 cherry-picked 1 -14.7536 simplelexicalconstraintswithoutsemanticinformation, 1 -14.7536 2(a)) 1 -14.7536 peirsman 1 -14.7536 1http://www.d.umn.edu/∼tpederse/data.html 1 -14.7536 6,379 1 -14.7536 127) 1 -14.7536 cuni.cz/morce/) 1 -14.7536 ”svp”), 1 -14.7536 sigir-08. 1 -14.7536 iwashita 1 -14.7536 definitionlookup. 1 -14.7536 1268 1 -14.7536 3online 1 -14.7536 4short 1 -14.7536 {xi,j}, 1 -14.7536 how1this 1 -14.7536 roget-like 1 -14.7536 wen-oin 1 -14.7536 82–90, 1 -14.7536 2pos-tags 1 -14.7536 nulex: 1 -14.7536 frameworkisthefirstcoresystemthatisdesignedtosup- 1 -14.7536 landman’s 1 -14.7536 (a-c), 1 -14.7536 disambiguation: 1 -14.7536 legitmize 1 -14.7536 ingoverlap 1 -14.7536 operationalised 1 -14.7536 §2.3 1 -14.7536 knowledge-drivenfeatures,wepresentbothversions 1 -14.7536 simplicity; 1 -14.7536 err_disj_r 1 -14.7536 ’$’ 1 -14.7536 schematically, 1 -14.7536 puled 1 -14.7536 twoparticulareffectiveimplementationsofcrfs 1 -14.7536 nee 1 -14.7536 otherhand,weseethattheperformancewiththespeaker-leave-one-out(m=78.3%) 1 -14.7536 arc,we 1 -14.7536 conduct five different experiments, each time using  1 -14.7536 english-to-hebrew 1 -14.7536 lasso’s 1 -14.7536 ganisations, 1 -14.7536 svmlight, 1 -14.7536 bodenstab 1 -14.7536 themodel 1 -14.7536 non-literary 1 -14.7536 6.140, 1 -14.7536 α/β: 1 -14.7536 contribution.n 1 -14.7536 emielkrahmer 1 -14.7536 1225 1 -14.7536 transcript’s 1 -14.7536 3these, 1 -14.7536 secdi 1 -14.7536 correponds 1 -14.7536 witen, 1 -14.7536 discussion/conclusions 1 -14.7536 complexityofckydecodinglimitsthetrainingdatainto 1 -14.7536 teases 1 -14.7536 author-based 1 -14.7536 60.38 1 -14.7536 pains 1 -14.7536 antecipate 1 -14.7536 instruc124 1 -14.7536 inthissetofexperiments,wetrainalogistic 1 -14.7536 rushton, 1 -14.7536 negativealignmentmoves 1 -14.7536 bioeve: 1 -14.7536 wealsowouldliketoexplore 1 -14.7536 4b), 1 -14.7536 ˆsare 1 -14.7536 388-395. 1 -14.7536 mturkers. 1 -14.7536 “^p 1 -14.7536 6.4, 1 -14.7536 nearly-unsupervised, 1 -14.7536 non-dmv-based 1 -14.7536 multi-classifications, 1 -14.7536 (letters 1 -14.7536 (reyes 1 -14.7536 asynchronous: 1 -14.7536 skovde 1 -14.7536 parser-agnostic, 1 -14.7536 unnamed 1 -14.7536 2sentences 1 -14.7536 traversals, 1 -14.7536 8235 1 -14.7536 1315 1 -14.7536 e0183287 1 -14.7536 t4) 1 -14.7536 paraphraseacquisition 1 -14.7536 #499: 1 -14.7536 interconnections 1 -14.7536 denotenullenullis 1 -14.7536 248–253, 1 -14.7536 wods 1 -14.7536 role-based 1 -14.7536 85.33 1 -14.7536 supposethelengthsofadocument 1 -14.7536 hence,low 1 -14.7536 (itshouldbenotedthatnotallofthetargetandresponses 1 -14.7536 envmine: 1 -14.7536 beforewediscusstherankingmodelsandtheempirical 1 -14.7536 unigram-filtered 1 -14.7536 sys1-4 1 -14.7536 c4.5+generaldown-sampling 1 -14.7536 decryption 1 -14.7536 thediscoveredtemplateshouldcontainasetofslots 1 -14.7536 (regularized) 1 -14.7536 #410: 1 -14.7536 2.4.5 1 -14.7536 85–91, 1 -14.7536 10also 1 -14.7536 homogeneous), 1 -14.7536 t-subsumption, 1 -14.7536 (helper 1 -14.7536 webelieve 1 -14.7536 runigramdclwdp 1 -14.7536 97-08, 1 -14.7536 machinelearning-inspired 1 -14.7536 tively 1 -14.7536 avarietyofdifferenttextfiletypes 1 -14.7536 32(4):1–14. 1 -14.7536 (p´erywoodley 1 -14.7536 (memm, 1 -14.7536 christoper 1 -14.7536 (17); 1 -14.7536 recentworkonemailhasconsideredthetaskofhighlightingmessageswithinaninboxwhichrequire 1 -14.7536 3,741 1 -14.7536 re-310 1 -14.7536 γu 1 -14.7536 8/10 1 -14.7536 ⁄⁄ 1 -14.7536 (pst), 1 -14.7536 letlbe 1 -14.7536 63.9%, 1 -14.7536 condemn) 1 -14.7536 theo(n6) 1 -14.7536 f(d)=pi 1 -14.7536 prerequisites, 1 -14.7536 p(w3nullw1,w2) 1 -14.7536 sigir1995. 1 -14.7536 (“dictionary”) 1 -14.7536 {d1,...,dk}denote 1 -14.7536 11any 1 -14.7536 parents1 1 -14.7536 conclusive: 1 -14.7536 aspect” 1 -14.7536 idation 1 -14.7536 236-241, 1 -14.7536 heart: 1 -14.7536 shallow7bos 1 -14.7536 eachamt 1 -14.7536 withour 1 -14.7536 in-paragraph 1 -14.7536 t∗b2 1 -14.7536 ifαutters 1 -14.7536 set6. 1 -14.7536 first-role 1 -14.7536 207–216, 1 -14.7536 dptsg 1 -14.7536 “earth-shattering” 1 -14.7536 epc 1 -14.7536 thealignernaturally 1 -14.7536 1282 1 -14.7536 asr-confidence 1 -14.7536 +label 1 -14.7536 nlp-oriented 1 -14.7536 optimizer/ 1 -14.7536 f(c) 1 -14.7536 gr. 1 -14.7536 2,3,4,5 1 -14.7536 clash 1 -14.7536 linguistic-inspired 1 -14.7536 re-construction 1 -14.7536 and〈k〉vary 1 -14.7536 conll2007 1 -14.7536 seah 1 -14.7536 overell’s 1 -14.7536 re-recreate 1 -14.7536 scgis 1 -14.7536 principle). 1 -14.7536 contrain 1 -14.7536 alternatively,one 1 -14.7536 hsmq-learning, 1 -14.7536 gulz 1 -14.7536 umist/ucl 1 -14.7536 t-scores 1 -14.7536 ‘warship⇒ship’ 1 -14.7536 wmeb’s 1 -14.7536 143–151, 1 -14.7536 152–158, 1 -14.7536 ace) 1 -14.7536 a.trainable 1 -14.7536 (atkins, 1 -14.7536 assignment〈z1,...,zs〉is 1 -14.7536 cov(α, 1 -14.7536 iis0713448. 1 -14.7536 84.59% 1 -14.7536 “†” 1 -14.7536 4http://ucrel.lancs.ac.uk/claws/ 1 -14.7536 thresholdγ, 1 -14.7536 tice 1 -14.7536 annotator‘s 1 -14.7536 syntactic-path 1 -14.7536 (interpreting 1 -14.7536 translationhypothesisproducedby 1 -14.7536 maamouri, 1 -14.7536 tection 1 -14.7536 noting, 1 -14.7536 system-agnostic 1 -14.7536 inotherwords,vertices 1 -14.7536 shenem��q� 1 -14.7536 -1.4 1 -14.7536 featurenullj 1 -14.7536 theproposedprior 1 -14.7536 230–238, 1 -14.7536 nodesuandv 1 -14.7536 wi∼mult(ηsi). 1 -14.7536 catalogueordering 1 -14.7536 pseudorandomized 1 -14.7536 gabrieli@limsi.fr 1 -14.7536 aspointedoutbybergerandlafferty 1 -14.7536 "bow: 1 -14.7536 champollion: 1 -14.7536 jcluster 1 -14.7536 solv62 1 -14.7536 wefind 1 -14.7536 disambigua571 1 -14.7536 phosphorylation–regulation) 1 -14.7536 35–42, 1 -14.7536 (1000) 1 -14.7536 wealsointroduce 1 -14.7536 esse: 1 -14.7536 tospoterrorsinwn,ourbasicideaistocontrasttwosourcesofknowledgeandautomaticallychecktheir 1 -14.7536 127-133, 1 -14.7536 100101102103104105106 1 -14.7536 postmodification 1 -14.7536 wiki5 1 -14.7536 37:207213. 1 -14.7536 upparse, 1 -14.7536 decidable, 1 -14.7536 architectural) 1 -14.7536 thelasthypothesisinfigure1(a)hasapassivevoicegrammaticalconstructionwhiletheothers 1 -14.7536 3http://kdd.ics.uci.edu/databases/ 1 -14.7536 nlcc 1 -14.7536 949–958, 1 -14.7536 ucphwas 1 -14.7536 2010/07/lda1.pdf 1 -14.7536 those〈a,n〉in 1 -14.7536 sycara 1 -14.7536 jectives 1 -14.7536 “multiword 1 -14.7536 ws/ns) 1 -14.7536 setbyagridsearchonthetrainingset. 1 -14.7536 evaluationshowsthatourmethodobtainsbetterrouge 1 -14.7536 differentmethodshavebeenproposedformaking 1 -14.7536 thisapproach 1 -14.7536 1per 1 -14.7536 3http://crfpp.sourceforge.net/ 1 -14.7536 1202–1212, 1 -14.7536 following(chiang 1 -14.7536 anddis 1 -14.7536 (+tb), 1 -14.7536 dbpedia: 1 -14.7536 methodweshowthecumulativeresultsfor 1 -14.7536 nphard, 1 -14.7536 ah4=b2,ah5=b1 1 -14.7536 138–142, 1 -14.7536 obtaininghigh-qualitywordalignmentsthatwerequire 1 -14.7536 σ=100 1 -14.7536 combo, 1 -14.7536 temp-np: 1 -14.7536 relatedworks 1 -14.7536 mini-tacitustriestoprovethelogical 1 -14.7536 ruleand 1 -14.7536 systemsconfidenceself-score(herreraetal., 1 -14.7536 reportedinthesecondpartoftable5. 1 -14.7536 functionr(s) 1 -14.7536 561k 1 -14.7536 constituencylike 1 -14.7536 hapax) 1 -14.7536 18superlative 1 -14.7536 boosting-based 1 -14.7536 2,296 1 -14.7536 551–560. 1 -14.7536 reshuffling. 1 -14.7536 543-565. 1 -14.7536 (nakanishi, 1 -14.7536 2.2e-16. 1 -14.7536 4.2graph 1 -14.7536 rouge-1.5.5 1 -14.7536 1.2.4 1 -14.7536 f(x)=0.5x+0.5 1 -14.7536 window: 1 -14.7536 bothsetscontain 1 -14.7536 ortiz-mart´ınez 1 -14.7536 null5.51; 1 -14.7536 ggps 1 -14.7536 insummary,oursystem’sexperimentalpipelineis 1 -14.7536 −20), 1 -14.7536 2008.peter 1 -14.7536 325–331, 1 -14.7536 educationally 1 -14.7536 +36 1 -14.7536 wecanseethatstrtm 1 -14.7536 macrof 1 -14.7536 294–303, 1 -14.7536 mackinnon 1 -14.7536 1thomas 1 -14.7536 (dandapat 1 -14.7536 naist, 1 -14.7536 metamap: 1 -14.7536 3http://www.factiva.com/ 1 -14.7536 rescorer 1 -14.7536 emnlpconll. 1 -14.7536 5opennlp 1 -14.7536 497–507, 1 -14.7536 null3.99; 1 -14.7536 interestingly,however,we 1 -14.7536 majorityclass 1 -14.7536 524–528, 1 -14.7536 ypq 1 -14.7536 1280 1 -14.7536 3)weapplytheaspecthierarchytothetaskofimplicit 1 -14.7536 6.9005** 1 -14.7536 correlationnaturallydoes 1 -14.7536 al-sabbagh 1 -14.7536 catib’s 1 -14.7536 ability) 1 -14.7536 52,265 1 -14.7536 4c), 1 -14.7536 r−h: 1 -14.7536 pseudo-relevant: 1 -14.7536 language/ 1 -14.7536 1.2.3 1 -14.7536 “holistically” 1 -14.7536 testes 1 -14.7536 2brown 1 -14.7536 side-effect, 1 -14.7536 1278–1287, 1 -14.7536 t12: 1 -14.7536 between-stimulus 1 -14.7536 iwsd-based 1 -14.7536 pronomial 1 -14.7536 formosan’s 1 -14.7536 1301–1309, 1 -14.7536 coling’10. 1 -14.7536 sentiful: 1 -14.7536 licensers 1 -14.7536 1435–1444, 1 -14.7536 d-tfidf: 1 -14.7536 ijcai’07. 1 -14.7536 typicalslotsincludeopinionholder,opinionexpression, 1 -14.7536 becauselogisticcanonlyuseoneruleatatime, 1 -14.7536 popularised 1 -14.7536 umcc-dlsi: 1 -14.7536 rameter 1 -14.7536 desc:manner 1 -14.7536 〈s10,c10〉: 1 -14.7536 (ananthakrishnan 1 -14.7536 make.03) 1 -14.7536 5mechanical 1 -14.7536 cd14surface 1 -14.7536 ‘clust-7-tgt’ 1 -14.7536 (segmenter, 1 -14.7536 id_1443, 1 -14.7536 coarsest 1 -14.7536 all/most/four 1 -14.7536 heeman’s 1 -14.7536 sideeffect 1 -14.7536 under-prediction 1 -14.7536 ports 1 -14.7536 ωj-proximal 1 -14.7536 sapporo. 1 -14.7536 non-task 1 -14.7536 sim(ci,cj) 1 -14.7536 bon. 1 -14.7536 ourbasicsystemcanconstrain 1 -14.7536 democracy 1 -14.7536 reinitialize 1 -14.7536 87m 1 -14.7536 low-accuracy 1 -14.7536 twitter-blog 1 -14.7536 classifiedinto 1 -14.7536 equaly, 1 -14.7536 lst;8 1 -14.7536 14c 1 -14.7536 intheformercase 1 -14.7536 =summationtextp:v2(p)=v 1 -14.7536 experiencers 1 -14.7536 easy-to-readversion 1 -14.7536 md-ii’s 1 -14.7536 3rd/3 1 -14.7536 (+cl) 1 -14.7536 com249 1 -14.7536 of/of) 1 -14.7536 stbi22 1 -14.7536 121–129, 1 -14.7536 opinionoriented 1 -14.7536 “type” 1 -14.7536 fordependency 1 -14.7536 hlt/emnlp-2005. 1 -14.7536 (stutsman 1 -14.7536 (2005b)), 1 -14.7536 13google 1 -14.7536 0-alignments 1 -14.7536 thisenforcestheconstraintthatatmost 1 -14.7536 vernimb, 1 -14.7536 non-terminalnodes 1 -14.7536 (2001b) 1 -14.7536 htp:/ww.csie.ntu.edu.tw/~cjlin/libsvm. 1 -14.7536 (3)mimacr 1 -14.7536 onework 1 -14.7536 “reuse” 1 -14.7536 imass 1 -14.7536 (torgesen 1 -14.7536 380,000 1 -14.7536 iis0812297 1 -14.7536 inevaluatingagainstathree-way 1 -14.7536 11l-types 1 -14.7536 road-testing 1 -14.7536 1,2,4 1 -14.7536 sjl= 1 -14.7536 1 -14.7536 attr-pp: 1 -14.7536 (ter-bleu)/2 1 -14.7536 treetagger1, 1 -14.7536 frobenious 1 -14.7536 (2006): 1 -14.7536 oversimplified 1 -14.7536 theseobservationsmotivateustoutilizeaspect 1 -14.7536 relative: 1 -14.7536 95–102, 1 -14.7536 qsbp: 1 -14.7536 675, 1 -14.7536 7mechanical 1 -14.7536 108–115, 1 -14.7536 8please 1 -14.7536 (hanapin) 1 -14.7536 wsem∩wsyn 1 -14.7536 infact, 1 -14.7536 +1.1 1 -14.7536 37%, 1 -14.7536 verbnet) 1 -14.7536 bleuscoreofusingtheseadditionaltreeletsforonthe-flyrulegenerationinoursimplemtsystem. 1 -14.7536 ifz 1 -14.7536 sytems, 1 -14.7536 5there 1 -14.7536 computetq 1 -14.7536 ourbackground 1 -14.7536 .507 1 -14.7536 imposition: 1 -14.7536 tight. 1 -14.7536 toaccesstheinformationencodedinrelationsbetweendrslabels, 1 -14.7536 single-syllable 1 -14.7536 problably 1 -14.7536 techniqes 1 -14.7536 mended 1 -14.7536 toeachpairandthengreedily 1 -14.7536 f15) 1 -14.7536 larson 1 -14.7536 ontonotesisanexcerptofnewsfromtheontonotes 1 -14.7536 however,thisfigureisartificiallylowasqa-syssupplies 1 -14.7536 fukumoto@yamanashi.ac.jp 1 -14.7536 k-statistic 1 -14.7536 diminishers, 1 -14.7536 meteor-next-* 1 -14.7536 long-distancewords 1 -14.7536 94.24%) 1 -14.7536 logic-basedsemantics. 1 -14.7536 habitat-lite: 1 -14.7536 thesummeddiagonalofconfusionmatrixisthepercentage 1 -14.7536 “promoter”) 1 -14.7536 re-usability 1 -14.7536 [...]in 1 -14.7536 “discovers” 1 -14.7536 compatibility, 1 -14.7536 pris 1 -14.7536 1011 1 -14.7536 daneman, 1 -14.7536 isdifficulttofindananalyticsolution. 1 -14.7536 developers’ 1 -14.7536 60s-to90s 1 -14.7536 weakly/distantly 1 -14.7536 pp→adjective 1 -14.7536 or-oov 1 -14.7536 alsowewantcheckwhether 1 -14.7536 b/d 1 -14.7536 appear; 1 -14.7536 >initial 1 -14.7536 115–120, 1 -14.7536 regula1210 1 -14.7536 91files 1 -14.7536 −log(x) 1 -14.7536 rel-c: 1 -14.7536 θis, 1 -14.7536 syn-low, 1 -14.7536 align: 1 -14.7536 5http://www.ldc.upenn.edu/ 1 -14.7536 probed 1 -14.7536 hwc’s 1 -14.7536 m>2 1 -14.7536 (bresnan 1 -14.7536 voyer 1 -14.7536 exteme 1 -14.7536 q(r) 1 -14.7536 tics 1 -14.7536 apa 1 -14.7536 error-based 1 -14.7536 t-chunk’s 1 -14.7536 combinationwhichexploitsthesyntacticsimilarityofsystem 1 -14.7536 (enabling 1 -14.7536 ‘degree.’ 1 -14.7536 (7/100) 1 -14.7536 (sconstraints) 1 -14.7536 privacy-reasons 1 -14.7536 fragmentlevel 1 -14.7536 〈s9,c9〉: 1 -14.7536 breiman 1 -14.7536 author/creator, 1 -14.7536 volume. 1 -14.7536 approximatestringsearchhasbeenstudiedbymany 1 -14.7536 (fr2), 1 -14.7536 categoriesthataregrammaticallyintimatelyconnectedtonouns 1 -14.7536 30(4):479–511. 1 -14.7536 “follow 1 -14.7536 translationrulesinvolvingflatphrasesare 1 -14.7536 especiallythe 1 -14.7536 bloated 1 -14.7536 s-guidelines_v5.4.3.pdf 1 -14.7536 contrastthiswith20minuteson32 1 -14.7536 newswire), 1 -14.7536 structed 1 -14.7536 11561159, 1 -14.7536 s-dtfidf: 1 -14.7536 othershaveconcrete 1 -14.7536 648–658, 1 -14.7536 document-query 1 -14.7536 1,303 1 -14.7536 paraphrased, 1 -14.7536 cereproc 1 -14.7536 cobleu, 1 -14.7536 238–246, 1 -14.7536 http://sites.google.com/site/bionlpst/. 1 -14.7536 “transmission”, 1 -14.7536 insteadoftask-independentexploratorydataanalysis 1 -14.7536 ‘granularity’ 1 -14.7536 23,651 1 -14.7536 chalendar 1 -14.7536 uai 1 -14.7536 1147–1157, 1 -14.7536 thetrainingandtestdatawerecreatedfromthewall 1 -14.7536 approximatedictionarymatching. 1 -14.7536 thismodellearnsfromthedevelopmentsetaseparate 1 -14.7536 dp-based 1 -14.7536 open-source1. 0.333333 -14.7536 0.62 0.60 -0.00 0.333333 -14.7536 (test) 0.333333 -14.7536 strained 0.333333 -14.7536 (agichtein 0.5 -14.7536 subontology 0.5 -14.7536 inu 0.5 -14.7536 jdkim@dbcls.rois.ac.jp 0.5 -14.7536 hiragana, 0.5 -14.7536 (asl+gi) 0.5 -14.7536 ˜null, 0.5 -14.7536 pos-taging 0.5 -14.7536 is-anullnullnull 0.5 -14.7536 visible. 0.5 -14.7536 “trivial” 0.5 -14.7536 π32 0.5 -14.7536 turns) 0.5 -14.7536 (kenny 0.5 -14.7536 (ast-cova) 0.5 -14.7536 −d 0.5 -14.7536 iis-0953149 0.5 -14.7536 g3(%) 0.5 -14.7536 (corston-oliver 0.5 -14.7536 owe 0.5 -14.7536 pisa, 0.5 -14.7536 rwk 0.5 -14.7536 cucchiarini, 0.5 -14.7536 ms,t 0.5 -14.7536 pizzas 0.5 -14.7536 0.7057 0.5 -14.7536 age-wise 0.5 -14.7536 re-score 0.5 -14.7536 γ(φn(wi−11 0.5 -14.7536 cocitation 0.5 -14.7536 gold-pos 0.5 -14.7536 0.5 -14.7536 allauzen 0.5 -14.7536 inv.edge-counts 0.5 -14.7536 beijing 100190, 0.5 -14.7536 st) 0.5 -14.7536 l∃ 0.5 -14.7536 roots. 0.5 -14.7536 11.5% 0.5 -14.7536 ntcir-7. 0.5 -14.7536 distribution). 0.5 -14.7536 normalform 0.5 -14.7536 obs 0.5 -14.7536 heart. 0.5 -14.7536 aggregations 0.5 -14.7536 callers, 0.5 -14.7536 (vm) 0.5 -14.7536 〈bl cr〉 0.5 -14.7536 d(x,y) 0.5 -14.7536 ubertwitter 0.5 -14.7536 “listens” 0.5 -14.7536 qsbp 0.5 -14.7536 contingency, 0.5 -14.7536 evaluationresultson 0.5 -14.7536 10−20 0.5 -14.7536 ensemble. 0.5 -14.7536 nil) 0.5 -14.7536 normalization; 0.5 -14.7536 commission, 0.5 -14.7536 de montr´eal 0.5 -14.7536 lakoff 0.5 -14.7536 143,580 0.5 -14.7536 summarizedin 0.5 -14.7536 tempex 0.5 -14.7536 finies 0.5 -14.7536 dev testbleu 0.5 -14.7536 superfluous 0.5 -14.7536 1,935 0.5 -14.7536 suburb 0.5 -14.7536 enter-s1.1 0.5 -14.7536 86.307 0.5 -14.7536 〈source〉(c 0.5 -14.7536 pre-clustersrefer 0.5 -14.7536 thine 0.5 -14.7536 0.99, 0.5 -14.7536 kaczynski 0.5 -14.7536 xx-enter 0.5 -14.7536 90920004. 0.5 -14.7536 top3 0.5 -14.7536 stipulates 0.5 -14.7536 meronyms, 0.5 -14.7536 cross-classified 0.5 -14.7536 alpage, 0.5 -14.7536 pittsburgh’s 0.5 -14.7536 mention) 0.5 -14.7536 1929-1930 0.5 -14.7536 pak, 0.5 -14.7536 0.119 0.5 -14.7536 (beck 0.5 -14.7536 (hancock 0.5 -14.7536 n-n 0.5 -14.7536 programmatically 0.5 -14.7536 b(t 0.5 -14.7536 obozinski 0.5 -14.7536 festgenommen , 0.5 -14.7536 passive) 0.5 -14.7536 gouws 0.5 -14.7536 pavel 0.5 -14.7536 hri 0.5 -14.7536 (passive 0.5 -14.7536 sleep? 0.5 -14.7536 1...n}, 0.5 -14.7536 iis-0844507, 0.5 -14.7536 prince. 0.5 -14.7536 (literal) 0.5 -14.7536 brighton 0.5 -14.7536 ppalli 0.5 -14.7536 (brockett 0.5 -14.7536 “that 0.5 -14.7536 suprising 0.5 -14.7536 (as), 0.5 -14.7536 ip1 0.5 -14.7536 gimenez, 0.5 -14.7536 f(t 0.5 -14.7536 urduenglish 0.5 -14.7536 cousot, 0.5 -14.7536 springer. 0.5 -14.7536 contaminated 0.5 -14.7536 unseen, 0.5 -14.7536 strzalkowski, 0.5 -14.7536 msdr 0.5 -14.7536 svmbased 0.5 -14.7536 w/constrained 0.5 -14.7536 propositions; 0.5 -14.7536 mistype 0.5 -14.7536 sentence-spanning 0.5 -14.7536 “christopher” 0.5 -14.7536 amani: 0.5 -14.7536 0.4470 0.5 -14.7536 rocket 0.5 -14.7536 abduction. 0.5 -14.7536 scfs 0.5 -14.7536 pixels 0.5 -14.7536 (stein 0.5 -14.7536 retuned 0.5 -14.7536 oppose 0.5 -14.7536 faruqui 0.5 -14.7536 φ(null) 0.5 -14.7536 root-morph 0.5 -14.7536 0.333333 -14.7547 s32s85s76s73s83s83s69 0.333333 -14.7547 graehl, 0.333333 -14.7547 θl+gq 0.172414 -14.7548 (wilson 0.4 -14.755 0.783 0.235294 -14.7552 task organizers 0.5 -14.7552 rumshisky 0.5 -14.7552 nose” 0.5 -14.7552 nes; 0.5 -14.7552 in§5.1. 0.5 -14.7552 canopies 0.5 -14.7552 sanders, 0.5 -14.7552 problem). 0.5 -14.7552 glimpse 0.5 -14.7552 harvested, 0.5 -14.7552 ptg2(usquigglerightr) 0.5 -14.7552 ciprimej. 0.5 -14.7552 woz. 0.5 -14.7552 preterminal-level 0.5 -14.7552 co-svm. 0.5 -14.7552 mary. 0.5 -14.7552 acceptable). 0.5 -14.7552 s-neut. 0.5 -14.7552 teporal 0.5 -14.7552 training.’ 0.5 -14.7552 8,666 0.5 -14.7552 cliques. 0.5 -14.7552 deprel. 0.5 -14.7552 corner, 0.5 -14.7552 acts), 0.5 -14.7552 uria 0.5 -14.7552 ssr-based 0.5 -14.7552 history-interpolated 0.5 -14.7552 accents 0.5 -14.7552 (f-score) 0.5 -14.7552 (tsvm) 0.5 -14.7552 0.8%. 0.5 -14.7552 bility 0.5 -14.7552 (scannell, 0.5 -14.7552 clauset 0.5 -14.7552 composition-decomposition 0.5 -14.7552 multi-hop 0.5 -14.7552 2–7, 0.5 -14.7552 (to) 0.5 -14.7552 canot 0.5 -14.7552 cool. 0.5 -14.7552 ,...,y 0.5 -14.7552 type-3 0.5 -14.7552 (tumasjan 0.5 -14.7552 0.1%. 0.5 -14.7552 -fb) 0.5 -14.7552 wm,n. 0.5 -14.7552 and−1 0.5 -14.7552 agrees. 0.5 -14.7552 (slda) 0.5 -14.7552 λi. 0.5 -14.7552 7” 0.5 -14.7552 asynch, 0.5 -14.7552 fresh. 0.5 -14.7552 accept. 0.5 -14.7552 sj2, 0.5 -14.7552 (larkey 0.5 -14.7552 95.64 0.5 -14.7552 fseed 0.5 -14.7552 (wainwright 0.5 -14.7552 n(key) 0.5 -14.7552 fully-fledged 0.5 -14.7552 based). 0.5 -14.7552 96.39% 0.5 -14.7552 67.4%, 0.5 -14.7552 lacostejulien 0.5 -14.7552 btstxdmwn1 0.5 -14.7552 wants. 0.5 -14.7552 q6, 0.5 -14.7552 3-grams. 0.5 -14.7552 wachsmuth, 0.5 -14.7552 5): 0.5 -14.7552 77.2% 0.5 -14.7552 class-labels 0.5 -14.7552 cer, 0.5 -14.7552 et.al, 0.5 -14.7552 subdivision 0.5 -14.7552 (steedman 0.5 -14.7552 mˆeme 0.5 -14.7552 [liang 0.5 -14.7552 compressors, 0.5 -14.7552 pre-tagged 0.5 -14.7552 jedi. 0.5 -14.7552 berkum, 0.5 -14.7552 (mad) 0.5 -14.7552 9.1, 0.5 -14.7552 geo250 0.5 -14.7552 flexible; 0.5 -14.7552 mrf. 0.5 -14.7552 (linguistic 0.5 -14.7552 (0.91 0.5 -14.7552 arg0. 0.5 -14.7552 mouse. 0.5 -14.7552 093 0.5 -14.7552 eroneously 0.5 -14.7552 schaback 0.5 -14.7552 sch. 0.5 -14.7552 “part-of-speech” 0.5 -14.7552 l(nullnullnull) 0.5 -14.7552 (4.05%) 0.5 -14.7552 64,000 0.5 -14.7552 skipn-gram 0.5 -14.7552 fixed), 0.5 -14.7552 deepsyntactic 0.5 -14.7552 halverson 0.5 -14.7552 w(s), 0.5 -14.7552 york) 0.5 -14.7552 co-refering 0.5 -14.7552 datato-text 0.5 -14.7552 language-like 0.5 -14.7552 m2. 0.5 -14.7552 (ei), 0.5 -14.7552 (unidirectional 0.5 -14.7552 supertagging: 0.5 -14.7552 sure. 0.5 -14.7552 3(c) 0.5 -14.7552 74.5% 0.5 -14.7552 four). 0.5 -14.7552 carlon 0.5 -14.7552 cross-validation; 0.5 -14.7552 1426 0.5 -14.7552 79.14 0.5 -14.7552 fk(x) 0.5 -14.7552 knlu, 0.5 -14.7552 (fvs) 0.5 -14.7552 η(t) 0.5 -14.7552 detonate, 0.5 -14.7552 infection. 0.5 -14.7552 “zooms 0.5 -14.7552 5k) 0.5 -14.7552 (contrastive) 0.5 -14.7552 gorithm 0.5 -14.7552 (2000); 0.5 -14.7552 turkdev, 0.5 -14.7552 one.” 0.5 -14.7552 47.10. 0.5 -14.7552 v-ing; 0.5 -14.7552 72.0% 0.5 -14.7552 third) 0.5 -14.7552 n-grams; 0.5 -14.7552 field: 0.5 -14.7552 135–144, 0.5 -14.7552 corected 0.5 -14.7552 tack 0.5 -14.7552 [nn1admiration] 0.5 -14.7552 expert-dependent 0.5 -14.7552 0.807 0.5 -14.7552 1124 0.5 -14.7552 aspect; 0.5 -14.7552 con1we 0.5 -14.7552 classify, 0.5 -14.7552 we(q,e) 0.5 -14.7552 (ca) 0.5 -14.7552 nonscorable 0.5 -14.7552 ortho 0.5 -14.7552 cross-loading 0.5 -14.7552 mt-chat, 0.5 -14.7552 y∈r 0.5 -14.7552 exemplarbased 0.5 -14.7552 (50.3%) 0.5 -14.7552 flue 0.5 -14.7552 discrepancy.n, 0.5 -14.7552 doris 0.5 -14.7552 (4.2) 0.5 -14.7552 0.0) 0.5 -14.7552 ensheathment 0.5 -14.7552 remembered 0.5 -14.7552 ζi 0.5 -14.7552 estimate) 0.5 -14.7552 overgeneration. 0.5 -14.7552 (vaassen 0.5 -14.7552 w.-accuracy 0.5 -14.7552 handwriting. 0.5 -14.7552 85.91 0.5 -14.7552 warranted. 0.5 -14.7552 grasshopper 0.5 -14.7552 :y≻y(cpt(y)). 0.5 -14.7552 1918) 0.5 -14.7552 comparable), 0.5 -14.7552 chain) 0.5 -14.7552 29.62 0.5 -14.7552 mexico. 0.5 -14.7552 (bdc) 0.5 -14.7552 one.) 0.5 -14.7552 pos-tagged. 0.5 -14.7552 0.0. 0.5 -14.7552 induction) 0.5 -14.7552 0.5◦ 0.5 -14.7552 -20 0.5 -14.7552 huberman, 0.5 -14.7552 59.55 0.5 -14.7552 5.2), 0.5 -14.7552 wprime1 0.5 -14.7552 p1,...,pk 0.5 -14.7552 32.1m 0.5 -14.7552 historian’s 0.5 -14.7552 2.0) 0.5 -14.7552 identical; 0.5 -14.7552 non-final 0.5 -14.7552 shirai 0.5 -14.7552 connectiveonly 0.5 -14.7552 band, 0.5 -14.7552 adequacy: 0.5 -14.7552 same-unit 0.5 -14.7552 darkness 0.5 -14.7552 referee? 0.5 -14.7552 (pet). 0.5 -14.7552 71.54 0.5 -14.7552 context-target 0.5 -14.7552 spelling-correction 0.5 -14.7552 7880 0.5 -14.7552 n-gram/plsa 0.5 -14.7552 reverb¬lex. 0.5 -14.7552 (0.3) 0.5 -14.7552 surname, 0.5 -14.7552 dredze. 0.5 -14.7552 bonfire 0.5 -14.7552 originals. 0.5 -14.7552 impairing 0.5 -14.7552 194) 0.5 -14.7552 (explicit 0.5 -14.7552 ei,t, 0.5 -14.7552 tent 0.5 -14.7552 bleuscore. 0.5 -14.7552 -0.83 0.5 -14.7552 4.1): 0.5 -14.7552 rpen 0.5 -14.7552 el-kahlout 0.5 -14.7552 strains 0.5 -14.7552 embassy 0.5 -14.7552 european-american 0.5 -14.7552 double-exponential 0.5 -14.7552 miscellaneous, 0.5 -14.7552 osteoporosis) 0.5 -14.7552 aliasing 0.5 -14.7552 kamp 0.5 -14.7552 1c. 0.5 -14.7552 α1. 0.5 -14.7552 device). 0.5 -14.7552 stopword, 0.5 -14.7552 wraps 0.5 -14.7552 warranty 0.5 -14.7552 gargett 0.5 -14.7552 2000)). 0.5 -14.7552 crumbled 0.5 -14.7552 (23% 0.5 -14.7552 2,715 0.5 -14.7552 considered), 0.5 -14.7552 wna. 0.5 -14.7552 ugaritic. 0.5 -14.7552 nps-baa-03 0.5 -14.7552 woepos’s 0.5 -14.7552 (arcs 0.5 -14.7552 tunes 0.5 -14.7552 unsuccessfully 0.5 -14.7552 cxyij 0.5 -14.7552 regeneration. 0.5 -14.7552 hillard 0.5 -14.7552 “no.” 0.5 -14.7552 currencies. 0.5 -14.7552 coarse-grained). 0.5 -14.7552 “remodeling”, 0.5 -14.7552 oneagainst-others 0.5 -14.7552 difficult) 0.5 -14.7552 rule-specific 0.5 -14.7552 lexically. 0.5 -14.7552 approximately), 0.5 -14.7552 86.1k 0.5 -14.7552 opportunistically 0.5 -14.7552 wikipedia5. 0.5 -14.7552 lemmatisation. 0.5 -14.7552 prague. 0.5 -14.7552 conferences. 0.5 -14.7552 autobi 0.5 -14.7552 pseudo-inverse 0.5 -14.7552 (+). 0.5 -14.7552 g0,f0. 0.5 -14.7552 (symmetric) 0.5 -14.7552 flowers. 0.5 -14.7552 grade. 0.5 -14.7552 man’s 0.5 -14.7552 semcor: 0.5 -14.7552 n¨oth 0.5 -14.7552 φ) 0.5 -14.7552 (iq) 0.5 -14.7552 (bouma 0.5 -14.7552 coder, 0.5 -14.7552 3.01 0.5 -14.7552 q1,...,qn. 0.5 -14.7552 ccb@cs.jhu.edu 0.5 -14.7552 (ca. 0.5 -14.7552 precedent 0.5 -14.7552 distributionally, 0.5 -14.7552 mauser 0.5 -14.7552 yamamoto, 0.5 -14.7552 wiley. 0.5 -14.7552 bayen, 0.5 -14.7552 finer-grained, 0.5 -14.7552 m&l-like 0.5 -14.7552 1892). 0.5 -14.7552 reformulation. 0.5 -14.7552 teu. 0.5 -14.7552 spanish; 0.5 -14.7552 action/event 0.5 -14.7552 oneto-one 0.5 -14.7552 personphone 0.5 -14.7552 43.73 0.5 -14.7552 purchased 0.5 -14.7552 assessors. 0.5 -14.7552 (cicling-2002), 0.5 -14.7552 non-propositional 0.5 -14.7552 blatantly 0.5 -14.7552 refreshing 0.5 -14.7552 typesetting. 0.5 -14.7552 eeg-olofsson 0.5 -14.7552 “word”. 0.5 -14.7552 ‘philip 0.5 -14.7552 (0/0=1) 0.5 -14.7552 sleeping 0.5 -14.7552 traction 0.5 -14.7552 (unsurprisingly) 0.5 -14.7552 (i>j), 0.5 -14.7552 edge-only 0.5 -14.7552 peaks. 0.5 -14.7552 122–126, 0.5 -14.7552 linguist, 0.5 -14.7552 “chaos” 0.5 -14.7552 (to). 0.5 -14.7552 lbj) 0.5 -14.7552 degree) 0.5 -14.7552 .am. 0.5 -14.7552 softly, 0.5 -14.7552 34.0% 0.5 -14.7552 92.09 0.5 -14.7552 lightturnstileleftn 0.5 -14.7552 (long) 0.5 -14.7552 metaphors. 0.5 -14.7552 specializing 0.5 -14.7552 helped. 0.5 -14.7552 1-34. 0.5 -14.7552 designing, 0.5 -14.7552 1–3 0.5 -14.7552 unplugged’ 0.5 -14.7552 lexicon-grammar. 0.5 -14.7552 classifiers‟ 0.5 -14.7552 “part,” 0.5 -14.7552 “documents” 0.5 -14.7552 dependant 0.5 -14.7552 duc-07, 0.5 -14.7552 2,284 0.5 -14.7552 naive. 0.5 -14.7552 dev10 0.5 -14.7552 (passonneau, 0.5 -14.7552 init, 0.5 -14.7552 h−t 0.5 -14.7552 wasat 0.5 -14.7552 cows. 0.5 -14.7552 publication; 0.5 -14.7552 respectively.) 0.5 -14.7552 “none” 0.5 -14.7552 consequence: 0.5 -14.7552 last). 0.5 -14.7552 stm: 0.5 -14.7552 duluth1 0.5 -14.7552 uturku. 0.5 -14.7552 tractability. 0.5 -14.7552 psychosentiwordnet(s) 0.5 -14.7552 71.14 0.5 -14.7552 (ades 0.5 -14.7552 “cheap” 0.5 -14.7552 “specific” 0.5 -14.7552 deu→en→es 0.5 -14.7552 proceeds. 0.5 -14.7552 word-aligned. 0.5 -14.7552 accompaniment. 0.5 -14.7552 plausibility. 0.5 -14.7552 inits. 0.5 -14.7552 pseudo-relation 0.5 -14.7552 niehren 0.5 -14.7552 25.74 0.5 -14.7552 (jones, 0.5 -14.7552 〈we〉 0.5 -14.7552 mbr-sc-e/c/extended 0.5 -14.7552 (ilisei 0.5 -14.7552 modernizing 0.5 -14.7552 importan 0.5 -14.7552 vista. 0.5 -14.7552 proceses 0.5 -14.7552 ninds, 0.5 -14.7552 ££ordinal-selector. 0.5 -14.7552 world”, 0.5 -14.7552 “entire” 0.5 -14.7552 subtree). 0.5 -14.7552 hedberg, 0.5 -14.7552 200; 0.5 -14.7552 opportunity. 0.5 -14.7552 (0.5). 0.5 -14.7552 mcfate 0.5 -14.7552 gerard 0.5 -14.7552 modeller, 0.5 -14.7552 ˆθik,φ 0.5 -14.7552 (little 0.5 -14.7552 completed; 0.5 -14.7552 juvenile 0.5 -14.7552 (6039) 0.5 -14.7552 strained.” 0.5 -14.7552 oier 0.5 -14.7552 bd’s 0.5 -14.7552 portfolio 0.5 -14.7552 (lison 0.5 -14.7552 “deceptive” 0.5 -14.7552 (callan 0.5 -14.7552 1970). 0.5 -14.7552 42m 0.5 -14.7552 hardness 0.5 -14.7552 clichés 0.5 -14.7552 selects, 0.5 -14.7552 “clean” 0.5 -14.7552 (judging 0.5 -14.7552 pivots. 0.5 -14.7552 (context). 0.5 -14.7552 m1. 0.5 -14.7552 [w2,w3] 0.5 -14.7552 overview), 0.5 -14.7552 iob2 0.5 -14.7552 task.) 0.5 -14.7552 4.the 0.5 -14.7552 45.6% 0.5 -14.7552 arrow. 0.5 -14.7552 features.4 0.5 -14.7552 flexible. 0.5 -14.7552 (0.1 0.5 -14.7552 logs). 0.5 -14.7552 prokaryotes. 0.5 -14.7552 76.6% 0.5 -14.7552 schoenemann, 0.5 -14.7552 labels? 0.5 -14.7552 (n.b. 0.5 -14.7552 sketch[k,hk(x)]←sketch[k,hk(x)] 0.5 -14.7552 semrol. 0.5 -14.7552 (1999)) 0.5 -14.7552 exclusively. 0.5 -14.7552 2.01 0.5 -14.7552 [0,1]; 0.5 -14.7552 unsel). 0.5 -14.7552  caught 0.5 -14.7552 (chunk 0.5 -14.7552 48–52, 0.5 -14.7552 elghaari 0.5 -14.7552 filling. 0.5 -14.7552 drop-off 0.5 -14.7552 instead: 0.5 -14.7552 lighten 0.5 -14.7552 promoted, 0.5 -14.7552 misc). 0.5 -14.7552 broadcast, 0.5 -14.7552 56.97 0.5 -14.7552 non-parallel, 0.5 -14.7552 encima 0.5 -14.7552 people). 0.5 -14.7552 bold). 0.5 -14.7552 fgen 0.5 -14.7552 1923). 0.5 -14.7552 4.4, 0.5 -14.7552 sportsteam. 0.5 -14.7552 3.72 0.5 -14.7552 artificialy 0.5 -14.7552 parts-of-speech-wise 0.5 -14.7552 industry. 0.5 -14.7552 concrete-abstract 0.5 -14.7552 three), 0.5 -14.7552 non-targets. 0.5 -14.7552 mbr-sc-e/conjoin 0.5 -14.7552 forecasts. 0.5 -14.7552 well-proven 0.5 -14.7552 (mccray 0.5 -14.7552 subsidiary. 0.5 -14.7552 patern” 0.5 -14.7552 pres, 0.5 -14.7552 timeframe 0.5 -14.7552 (30%). 0.5 -14.7552 (icwsm). 0.5 -14.7552 panlingual 0.5 -14.7552 bird. 0.5 -14.7552 metric-based 0.5 -14.7552 prp. 0.5 -14.7552 opinion). 0.5 -14.7552 re-orderings. 0.5 -14.7552 pl-mrf, 0.5 -14.7552 (x). 0.5 -14.7552 11,600 0.5 -14.7552 303k 0.5 -14.7552 postsim 0.5 -14.7552 ukwac+clc 0.5 -14.7552 derivation). 0.5 -14.7552 culture: 0.5 -14.7552 als. 0.5 -14.7552 §4.3. 0.5 -14.7552 stringto-tree 0.5 -14.7552 differentk-best 0.5 -14.7552 791 0.5 -14.7552 overestimated. 0.5 -14.7552 transliterating 0.5 -14.7552 (german). 0.5 -14.7552 omissions. 0.5 -14.7552 flesch 0.5 -14.7552 oracles, 0.5 -14.7552 rv. 0.5 -14.7552 parabronchi. 0.5 -14.7552 1589 0.5 -14.7552 (nec) 0.5 -14.7552 hash. 0.5 -14.7552 uni-grams 0.5 -14.7552 (1996, 0.5 -14.7552 metagenomic 0.5 -14.7552 n=10, 0.5 -14.7552 scheme-sets, 0.5 -14.7552 59.83 0.5 -14.7552 (q)↓} 0.5 -14.7552 g(·) 0.5 -14.7552 alia), 0.5 -14.7552 traces, 0.5 -14.7552 garbled 0.5 -14.7552 41.08 0.5 -14.7552 ceafe, 0.5 -14.7552 (soto 0.5 -14.7552 2here 0.5 -14.7552 diverse: 0.5 -14.7552 lump 0.5 -14.7552 (tweets) 0.5 -14.7552 mcconkie 0.5 -14.7552 acronym, 0.5 -14.7552 7722 0.5 -14.7552 2006)? 0.5 -14.7552 lexical). 0.5 -14.7552 (fp7-ict-4-24853). 0.5 -14.7552 pas. 0.5 -14.7552 (bottou, 0.5 -14.7552 s∈st. 0.5 -14.7552 exemplifying 0.5 -14.7552 restricted-many-to-many. 0.5 -14.7552 unborn 0.5 -14.7552 (yerva 0.5 -14.7552 p+hp 0.5 -14.7552 questionable, 0.5 -14.7552 f1,f2 0.5 -14.7552 sarajevo 0.5 -14.7552 (function 0.5 -14.7552 object-oriented 0.5 -14.7552 (biber 0.5 -14.7552 ψi. 0.5 -14.7552 cognizer 0.5 -14.7552 recased 0.5 -14.7552 z; 0.5 -14.7552 (ccg). 0.5 -14.7552 unkown 0.5 -14.7552 tutorial. 0.5 -14.7552 20676 0.5 -14.7552 56%. 0.5 -14.7552 90.0%. 0.5 -14.7552 vbg; 0.5 -14.7552 coarsened 0.5 -14.7552 query-log 0.5 -14.7552 headnounonly 0.5 -14.7552 familiarity, 0.5 -14.7552 rel-lda1. 0.5 -14.7552 1.763 0.5 -14.7552 kpbl1, 0.5 -14.7552 book). 0.5 -14.7552 opinions: 0.5 -14.7552 devtest, 0.5 -14.7552 bfgs 0.5 -14.7552 ?y2 0.5 -14.7552 understod 0.5 -14.7552 barerel. 0.5 -14.7552 integral, 0.5 -14.7552 bcn, 0.5 -14.7552 matching-based 0.5 -14.7552 ffop 0.5 -14.7552 (benson 0.5 -14.7552 (typical 0.5 -14.7552 featuresdescribingtherelationshipbetweenci 0.5 -14.7552 hripcsak. 0.5 -14.7552 affiliation. 0.5 -14.7552 kbow 0.5 -14.7552 (patrick 0.5 -14.7552 hospital. 0.5 -14.7552 (1+2+3) 0.5 -14.7552 application-based 0.5 -14.7552 constraints). 0.5 -14.7552 re-generation 0.5 -14.7552 estonia. 0.5 -14.7552 school”; 0.5 -14.7552 consequences. 0.5 -14.7552 aligned; 0.5 -14.7552 (below). 0.5 -14.7552 (2001)) 0.5 -14.7552 (clustering) 0.5 -14.7552 metathesaurus. 0.5 -14.7552 bleutuned 0.5 -14.7552 (0.9, 0.5 -14.7552 tunable-metric 0.5 -14.7552 lisävuodevaatteet 0.5 -14.7552 inverts 0.5 -14.7552 ..c 0.5 -14.7552 (l). 0.5 -14.7552 thursday. 0.5 -14.7552 munro, 0.5 -14.7552 oh...!! 0.5 -14.7552 baldridge’s 0.5 -14.7552 scalability, 0.5 -14.7552 imputing 0.5 -14.7552 listing. 0.5 -14.7552 (aac). 0.5 -14.7552 (bl) 0.5 -14.7552 λx.f(g(x)) 0.5 -14.7552 training.) 0.5 -14.7552 stem+morph. 0.5 -14.7552 baluja 0.5 -14.7552 inflection-wise 0.5 -14.7552 mirrored 0.5 -14.7552 (5.2%) 0.5 -14.7552 levantine, 0.178571 -14.7554 anthology 0.333333 -14.7558 procesing. 0.4 -14.7563 0.138 0.4 -14.7563 h∗ 0.4 -14.7563 “al” 0.4 -14.7563 0.994 0.5 -14.7569 jacent 0.5 -14.7569 annotators‟ 0.5 -14.7569 familiarizing 0.5 -14.7569 vitanyi 0.5 -14.7569 farrow 0.5 -14.7569 0.953 0.5 -14.7569 husband, 0.5 -14.7569 3:6a 0.5 -14.7569 exp(pmi(wk,wi)) 0.5 -14.7569 james.henderson@unige.ch 0.5 -14.7569 felzenszwalb 0.5 -14.7569 narrator 0.5 -14.7569 ce-reduction 0.5 -14.7569 work.’ 0.5 -14.7569 non-nested 0.5 -14.7569 karrowbothv(g(+g2)1 0.5 -14.7569 victims, 0.5 -14.7569 biscuits 0.5 -14.7569 hmm; 0.5 -14.7569 62.17 0.5 -14.7569 tpr, 0.5 -14.7569 s-factors. 0.5 -14.7569 $uidccmean 0.5 -14.7569 (chunks) 0.5 -14.7569 mcallester, 0.5 -14.7569 20–26, 0.5 -14.7569 (young, 0.5 -14.7569 non-anaphor 0.5 -14.7569 pairwise. 0.5 -14.7569 s(a(pu),ps) 0.5 -14.7569 tt[k] 0.5 -14.7569 methylation, 0.5 -14.7569 lr) 0.5 -14.7569 subsectionheader 0.5 -14.7569 ’screening’, 0.5 -14.7569 ˜s. 0.5 -14.7569 contrasted. 0.5 -14.7569 synonymity. 0.5 -14.7569 j(i). 0.5 -14.7569 2010)2. 0.5 -14.7569 58%. 0.5 -14.7569 weijer, 0.5 -14.7569 blogger. 0.5 -14.7569 (7.7 0.5 -14.7569 51.05 0.5 -14.7569 bericht 0.5 -14.7569 expandrank, 0.5 -14.7569 researches, 0.5 -14.7569 1due 0.5 -14.7569 daunting. 0.5 -14.7569 untrusted} 0.5 -14.7569 lists) 0.5 -14.7569 plementation 0.5 -14.7569 .673 0.5 -14.7569 perplexity-based 0.5 -14.7569 (jsps). 0.5 -14.7569 93.86 0.5 -14.7569 sub-grammar 0.5 -14.7569 119,783 0.5 -14.7569 pas) 0.5 -14.7569 semester 0.5 -14.7569 aai 0.5 -14.7569 generalized. 0.5 -14.7569 18m 0.5 -14.7569 50.4% 0.5 -14.7569 96.64 0.5 -14.7569 kpbl2, 0.5 -14.7569 endocentric 0.5 -14.7569 (blume, 0.5 -14.7569 (possible 0.5 -14.7569 hierarchic 0.5 -14.7569 tolkit 0.5 -14.7569 client’s 0.5 -14.7569 fish”). 0.5 -14.7569 dp-tsg. 0.5 -14.7569 air. 0.5 -14.7569 down-casing 0.5 -14.7569 gap−fill 0.5 -14.7569 candidatec∈e(s) 0.5 -14.7569 worlds, 0.5 -14.7569 1045 0.5 -14.7569 did/do/does 0.5 -14.7569 86.62 0.5 -14.7569 (§2) 0.5 -14.7569 (schegloff, 0.5 -14.7569 (21%) 0.5 -14.7569 al: 0.5 -14.7569 espinosa 0.5 -14.7569 namasivayam, 0.5 -14.7569 (incorrectly) 0.5 -14.7569 33.68 0.5 -14.7569 (162 0.5 -14.7569 (eula) 0.5 -14.7569 “share” 0.5 -14.7569 hiding 0.5 -14.7569 0.78). 0.5 -14.7569 closet. 0.5 -14.7569 star-free, 0.5 -14.7569 澳洲 0.5 -14.7569 64.08 0.5 -14.7569 fulfilled; 0.5 -14.7569 quantified, 0.5 -14.7569 earlyupdate 0.5 -14.7569 callaway 0.5 -14.7569 sampler’s 0.5 -14.7569 objectives) 0.5 -14.7569 *ne* 0.5 -14.7569 y+1 0.5 -14.7569 administering 0.5 -14.7569 1133 0.5 -14.7569 g− 0.5 -14.7569 “off” 0.5 -14.7569 enlightenment 0.5 -14.7569 nonpropagation 0.5 -14.7569 inflects 0.5 -14.7569 okuda 0.5 -14.7569 edu; 0.5 -14.7569 ”translation”. 0.5 -14.7569 depicted. 0.5 -14.7569 (ed). 0.5 -14.7569 frame-evoking 0.5 -14.7569 many-toone 0.5 -14.7569 threads) 0.5 -14.7569 top-50 0.5 -14.7569 fear. 0.5 -14.7569 world”. 0.5 -14.7569 abbrev.100 0.5 -14.7569 .726 0.5 -14.7569 ter(w/o 0.5 -14.7569 cecandidates 0.5 -14.7569 sv-lvcs. 0.5 -14.7569 goggin, 0.5 -14.7569  some 0.5 -14.7569 ass, 0.5 -14.7569 ,g(+g1)2 0.5 -14.7569 use; 0.5 -14.7569 7476 0.5 -14.7569 4because 0.5 -14.7569 @?x 0.5 -14.7569 cross-entropies 0.5 -14.7569 1http://www.statmt.org/wmt10 0.5 -14.7569 95.66 0.5 -14.7569 redundancy: 0.5 -14.7569 76.2%) 0.5 -14.7569 l(δθ(xi),yi). 0.5 -14.7569 wherew 0.5 -14.7569 trinken. 0.5 -14.7569 “audi” 0.5 -14.7569 disagreeing. 0.5 -14.7569 c2‟, 0.5 -14.7569 %gain 0.5 -14.7569 915 0.5 -14.7569 larsen-freeman, 0.5 -14.7569 40). 0.5 -14.7569 (cettolo 0.5 -14.7569 s-factor). 0.5 -14.7569 3.6, 0.5 -14.7569 objective). 0.5 -14.7569 rte’s 0.5 -14.7569 z−j 0.5 -14.7569 ready-made 0.5 -14.7569 declarations, 0.5 -14.7569 (ni 0.5 -14.7569 sociolects, 0.5 -14.7569 recscore+bias, 0.5 -14.7569 n(f) 0.5 -14.7569 (david 0.5 -14.7569 5001000150020002500 0.5 -14.7569 null2). 0.5 -14.7569 time(sec) 0.5 -14.7569 (liwc) 0.5 -14.7569 jobs). 0.5 -14.7569 11%) 0.5 -14.7569 human-likeness. 0.5 -14.7569 vm-lr 0.5 -14.7569 400m 0.5 -14.7569 misestimations 0.5 -14.7569 (unsupervised 0.5 -14.7569 (desr, 0.5 -14.7569 heaven 0.5 -14.7569 above”. 0.5 -14.7569 feedback} 0.5 -14.7569 (el 0.5 -14.7569 nullnull2) 0.5 -14.7569 epsilon1/2 0.5 -14.7569 (figure, 0.5 -14.7569 eanlu, 0.5 -14.7569 conciliated 0.5 -14.7569 arrest 0.5 -14.7569 49.72 0.5 -14.7569 dict). 0.5 -14.7569 anchored, 0.5 -14.7569 bayesian. 0.5 -14.7569 disproportionally 0.5 -14.7569 emergencies. 0.5 -14.7569 loops. 0.5 -14.7569 grow. 0.5 -14.7569 “origtweets” 0.5 -14.7569 weave 0.5 -14.7569 twiter. 0.5 -14.7569 (rdm) 0.5 -14.7569 {1,...,n}. 0.5 -14.7569 frozen, 0.5 -14.7569 (rodrigo 0.5 -14.7569 co-task 0.5 -14.7569 proteinlocation 0.5 -14.7569 74.86 0.5 -14.7569 0.86. 0.5 -14.7569 (na). 0.5 -14.7569 [6]. 0.5 -14.7569 planted 0.5 -14.7569 he). 0.5 -14.7569 smile. 0.5 -14.7569 odel 0.5 -14.7569 role-syntax 0.5 -14.7569 db:13 0.5 -14.7569 ibm-style 0.5 -14.7569 ideational 0.5 -14.7569 22.20 0.5 -14.7569 xprime←x; 0.5 -14.7569 opposites, 0.5 -14.7569 recalculation 0.5 -14.7569 (@) 0.5 -14.7569 ···wn} 0.5 -14.7569 survives, 0.5 -14.7569 focused. 0.5 -14.7569 variations). 0.5 -14.7569 nrsnl 0.5 -14.7569 0.034 0.5 -14.7569 (rokach, 0.5 -14.7569 (schiller 0.5 -14.7569 strokes 0.5 -14.7569 2-3% 0.5 -14.7569 hiroya 0.5 -14.7569 ···bj. 0.5 -14.7569 time=abs, 0.5 -14.7569 mark.dras@mq.edu.au 0.5 -14.7569 kwek 0.5 -14.7569 (ept) 0.5 -14.7569 linguistics; 0.5 -14.7569 1)-th 0.5 -14.7569 context-appropriate 0.5 -14.7569 milin, 0.5 -14.7569 allowance 0.5 -14.7569 localizations 0.5 -14.7569 fsas 0.5 -14.7569 rsttool, 0.5 -14.7569 2009)for 0.5 -14.7569 non-coverage 0.5 -14.7569 spliting 0.5 -14.7569 http://projects.ldc.upenn.edu/ace/ 0.5 -14.7569 qian. 0.5 -14.7569 67.9% 0.5 -14.7569 bayen 0.5 -14.7569 disciplines, 0.5 -14.7569 retrieval) 0.5 -14.7569 non-coordinated 0.5 -14.7569 bad’ 0.5 -14.7569 .800 0.5 -14.7569 macedonia 0.5 -14.7569 mxterminator 0.5 -14.7569 wa’s 0.5 -14.7569 acts; 0.5 -14.7569 recommends 0.5 -14.7569 reentered 0.5 -14.7569 (thousands) 0.5 -14.7569 work.3 0.5 -14.7569 λlw 0.5 -14.7569 fracture). 0.5 -14.7569 ko. 0.5 -14.7569 a2: 0.5 -14.7569 from/to 0.5 -14.7569 (r): 0.333333 -14.757 wan, 0.0081171 -14.757 which 0.272727 -14.7582 48th annual 0.5 -14.7586 54.08 0.5 -14.7586 marc21 0.5 -14.7586 50.62 0.5 -14.7586 maxg 0.5 -14.7586 (caseli 0.5 -14.7586 s∪ 0.5 -14.7586 grayscale. 0.5 -14.7586 mvm. 0.5 -14.7586 gold.1 0.5 -14.7586 first-orderword 0.5 -14.7586 movies. 0.5 -14.7586 on” 0.5 -14.7586 #outputs-assessed 0.5 -14.7586 lea 0.5 -14.7586 〈x,y〉 0.5 -14.7586 texas. 0.5 -14.7586 sandra. 0.5 -14.7586 paik 0.5 -14.7586 ˜xijnegationslash= 0.5 -14.7586 (strong) 0.5 -14.7586 jwpl, 0.5 -14.7586 pinwith和 0.5 -14.7586 boring, 0.5 -14.7586 today's 0.5 -14.7586 (1972) 0.5 -14.7586 linguistics). 0.5 -14.7586 concatenated, 0.5 -14.7586 best-k 0.5 -14.7586 1.048 0.5 -14.7586 performing. 0.5 -14.7586 extensibility: 0.5 -14.7586 4(2). 0.5 -14.7586 syllabification. 0.5 -14.7586 hyphened 0.5 -14.7586 pre-test. 0.5 -14.7586 investors. 0.5 -14.7586 1592 0.5 -14.7586 (llda) 0.5 -14.7586 1−p. 0.5 -14.7586 equated 0.5 -14.7586 accommodation. 0.5 -14.7586 glue. 0.5 -14.7586 august. 0.5 -14.7586 data.5 0.5 -14.7586 english-urdu 0.5 -14.7586 learning), 0.5 -14.7586 51.0% 0.5 -14.7586 fulfil 0.5 -14.7586 idsc. 0.5 -14.7586 extend, 0.5 -14.7586 64.96 0.5 -14.7586 genia/bionlp 0.5 -14.7586 discourses. 0.5 -14.7586 term–document 0.5 -14.7586 (multi-proj.) 0.5 -14.7586 hindi) 0.5 -14.7586 ips. 0.5 -14.7586 (fossati 0.5 -14.7586 stay. 0.5 -14.7586 93.67 0.5 -14.7586 crabb´e 0.5 -14.7586 (weir, 0.5 -14.7586 meaningless. 0.5 -14.7586 nd-l2 0.5 -14.7586 tutor) 0.5 -14.7586 hyperonyms, 0.5 -14.7586 screens. 0.5 -14.7586 (azmp). 0.5 -14.7586 {s=4000}. 0.5 -14.7586 guidelines). 0.5 -14.7586 coreferenceresolution,as 0.5 -14.7586 (seq/symbol) 0.5 -14.7586 eye-blink 0.5 -14.7586 x-1 0.5 -14.7586 66.72 0.5 -14.7586 here.) 0.5 -14.7586 erinnern. 0.5 -14.7586 discuss, 0.5 -14.7586 at-mentions 0.5 -14.7586 digits). 0.5 -14.7586 surname. 0.5 -14.7586 1442 0.5 -14.7586 playable 0.5 -14.7586 iraqi. 0.5 -14.7586 hve-3e 0.5 -14.7586 ttest 0.5 -14.7586 gories 0.5 -14.7586 intentionality. 0.5 -14.7586 terrorism, 0.5 -14.7586 couples. 0.5 -14.7586 merit) 0.5 -14.7586 dependent’s 0.5 -14.7586 weblog, 0.5 -14.7586 coreferential). 0.5 -14.7586 26.75 0.5 -14.7586 6.4). 0.5 -14.7586 83.83 0.5 -14.7586 (2005)). 0.5 -14.7586 partial-match 0.5 -14.7586 hy10. 0.5 -14.7586 fulldimensional 0.5 -14.7586 wheref 0.5 -14.7586 @csail.mit.edu 0.5 -14.7586 kbhall@google.com 0.5 -14.7586 postposition. 0.5 -14.7586 absolute), 0.5 -14.7586 roles). 0.5 -14.7586 (apd) 0.5 -14.7586 hmm-type-r, 0.5 -14.7586 (kumamoto, 0.5 -14.7586 phr, 0.5 -14.7586 (villemonte 0.5 -14.7586 audible 0.5 -14.7586 asdiscussed 0.5 -14.7586 word-sets 0.5 -14.7586 suitable. 0.5 -14.7586 precise. 0.5 -14.7586 35.73 0.5 -14.7586 av1+bv2 0.5 -14.7586 accents, 0.5 -14.7586 choice) 0.5 -14.7586 opposition. 0.5 -14.7586 (toselli 0.5 -14.7586 intimidating 0.5 -14.7586 72%, 0.5 -14.7586 (tbl) 0.5 -14.7586 6:1 0.5 -14.7586 atribute 0.5 -14.7586 ∗. 0.5 -14.7586 edit-based 0.5 -14.7586 cprime, 0.5 -14.7586 scren. 0.5 -14.7586 (2.9) 0.5 -14.7586 intelligence”. 0.5 -14.7586 65.93 0.5 -14.7586 vectori21 0.5 -14.7586 opponent, 0.5 -14.7586 .jo. 0.5 -14.7586 (rosenberg, 0.5 -14.7586 masses. 0.5 -14.7586 discursive 0.5 -14.7586 english–english 0.5 -14.7586 studentship. 0.5 -14.7586 75.59 0.5 -14.7586 tri+ 0.5 -14.7586 esser, 0.5 -14.7586 prepositions; 0.5 -14.7586 ptg1(usquigglerights) 0.5 -14.7586 erors, 0.5 -14.7586 dramaticaly 0.5 -14.7586 1.979 0.5 -14.7586 dependents) 0.5 -14.7586 negation). 0.5 -14.7586 p=3 0.5 -14.7586 wash. 0.5 -14.7586 absolute) 0.5 -14.7586 hispanics. 0.5 -14.7586 begins. 0.5 -14.7586 vorstand. 0.5 -14.7586 maxprob 0.5 -14.7586 march. 0.5 -14.7586 http://crfpp.sourceforge.net/ 0.5 -14.7586 concrete. 0.5 -14.7586 ctb, 0.5 -14.7586 fl-score 0.5 -14.7586 fine-tuning 0.5 -14.7586 ‘oracle’ 0.5 -14.7586 hunt’, 0.5 -14.7586 cp. 0.5 -14.7586 (tts). 0.5 -14.7586 6-12). 0.5 -14.7586 0.5 -14.7586 acceptability, 0.5 -14.7586 non-trivial. 0.222222 -14.7588 eisenstein et 0.2 -14.7589 for reasons 0.4 -14.759 modified 0.137255 -14.759 than half 0.4 -14.7603 semigroup 0.142857 -14.7603 li and 0.5 -14.7603 40.14 0.5 -14.7603 favourability, 0.5 -14.7603 nunes, 0.5 -14.7603 decoding? 0.5 -14.7603 (5.5) 0.5 -14.7603 table8. 0.5 -14.7603 collisions. 0.5 -14.7603 split). 0.5 -14.7603 tense”. 0.5 -14.7603 (§6.3). 0.5 -14.7603 alignment+ind. 0.5 -14.7603 “label”. 0.5 -14.7603 beamwidth 0.5 -14.7603 4.3), 0.5 -14.7603 color-coded 0.5 -14.7603 zhechev 0.5 -14.7603 sales. 0.5 -14.7603 α(v), 0.5 -14.7603 c(3,k), 0.5 -14.7603 mono-lingual 0.5 -14.7603 3740 0.5 -14.7603 seting, 0.5 -14.7603 s55s54 0.5 -14.7603 words(i) 0.5 -14.7603 (attributes, 0.5 -14.7603 non-literate. 0.5 -14.7603 (wagstaff 0.5 -14.7603 placeholders, 0.5 -14.7603 commutative: 0.5 -14.7603 speculation). 0.5 -14.7603 minutes). 0.5 -14.7603 gold-standard, 0.5 -14.7603 77.13 0.5 -14.7603 transaction. 0.5 -14.7603 hhmm. 0.5 -14.7603 46.74 0.5 -14.7603 smet 0.5 -14.7603 drel 0.5 -14.7603 (hayes, 0.5 -14.7603 343 0.5 -14.7603 t!t02z 0.5 -14.7603 0.763 0.5 -14.7603 (petrovi´c 0.5 -14.7603 win3, 0.5 -14.7603 [drawing]. 0.5 -14.7603 populationcenter 0.5 -14.7603 hamlet. 0.5 -14.7603 internally. 0.5 -14.7603 (nullnullnull@null) 0.5 -14.7603 rwth. 0.5 -14.7603 translationquality. 0.5 -14.7603 59.74 0.5 -14.7603 19%). 0.5 -14.7603 sharable 0.5 -14.7603 toujan 0.5 -14.7603 transactions). 0.5 -14.7603 rftagger. 0.5 -14.7603 (tepperman 0.5 -14.7603 dominance-preserving 0.5 -14.7603 pm. 0.5 -14.7603 potter, 0.5 -14.7603 non-comercial 0.5 -14.7603 =/= 0.5 -14.7603 24.88 0.5 -14.7603 23.04 0.5 -14.7603 ,lets 0.5 -14.7603 farrell, 0.5 -14.7603 picrowd 0.5 -14.7603 there?). 0.5 -14.7603 researchcyc 0.5 -14.7603 tense). 0.5 -14.7603 sequence). 0.5 -14.7603 (composed 0.5 -14.7603 smyth 0.5 -14.7603 del´eger 0.5 -14.7603 framework). 0.5 -14.7603 dendrogram. 0.5 -14.7603 figure1. 0.5 -14.7603 1963). 0.5 -14.7603 south. 0.5 -14.7603 politicians 0.5 -14.7603 l(sy)i 0.5 -14.7603 77.72 0.5 -14.7603 tence. 0.5 -14.7603 afternoon. 0.5 -14.7603 75.84 0.5 -14.7603 lapata. 0.5 -14.7603 rats. 0.5 -14.7603 (91.2) 0.5 -14.7603 phr. 0.5 -14.7603 extendible 0.5 -14.7603 β∗←regressionlscript2(xprime,yt2,λc) 0.5 -14.7603 λy.λx.loc(y,x) 0.5 -14.7603 (butterworth, 0.5 -14.7603 giraffe 0.5 -14.7603 (wer). 0.5 -14.7603 fsyn(.). 0.5 -14.7603 n· 0.5 -14.7603 deletion”. 0.5 -14.7603 q6 0.5 -14.7603 throut%. 0.5 -14.7603 mono. 0.5 -14.7603 multisentence 0.5 -14.7603 russell 0.5 -14.7603 wth 0.5 -14.7603 each-side 0.5 -14.7603 kpbl1. 0.5 -14.7603 convincing. 0.5 -14.7603 (bohnet, 0.5 -14.7619 story: 0.5 -14.7619 andthe 0.5 -14.7619 (12.5%). 0.5 -14.7619 all? 0.5 -14.7619 76.8. 0.5 -14.7619 sept 0.5 -14.7619 (milidi´u 0.5 -14.7619 k&c10 0.5 -14.7619 (hdp), 0.5 -14.7619 eparl+nc+1092+news 0.5 -14.7619 r− 0.5 -14.7619 quantile, 0.5 -14.7619 party” 0.5 -14.7619 hedges, 0.5 -14.7619 [pp09]: 0.5 -14.7619 93.62 0.5 -14.7619 web-browser. 0.5 -14.7619 losielle, 0.5 -14.7619 content), 0.5 -14.7619 31.39 0.5 -14.7619 0..3 0.5 -14.7619 schiefert. 0.5 -14.7619 7our 0.5 -14.7619 chest. 0.5 -14.7619 conversation) 0.5 -14.7619 grandparent), 0.5 -14.7619 seventeenth 0.5 -14.7619 validationdata. 0.5 -14.7619 91.67 0.5 -14.7619 (gill 0.5 -14.7619 flies. 0.5 -14.7619 prescription. 0.5 -14.7619 pasages, 0.5 -14.7619 strmatch npron 0.5 -14.7619 fr-en). 0.5 -14.7619 idiosyncratic. 0.5 -14.7619 74.84 0.5 -14.7619 2pr=(p 0.5 -14.7619 (?x1 0.5 -14.7619 basu 0.5 -14.7619 modules) 0.5 -14.7619 laptop. 0.5 -14.7619 onset. 0.5 -14.7619 morphemes). 0.5 -14.7619 measurably 0.5 -14.7619 ftb-dev 0.5 -14.7619 combinatorially 0.5 -14.7619 equipment. 0.5 -14.7619 asesment 0.5 -14.7619 roberts. 0.5 -14.7619 (cmnt2) 0.5 -14.7619 al.,2007). 0.5 -14.7619 interruption. 0.5 -14.7619 basis: 0.5 -14.7619 old-fashioned 0.5 -14.7619 hgfc, 0.5 -14.7619 site’s 0.5 -14.7619 manageable. 0.5 -14.7619 sharply. 0.5 -14.7619 somebody. 0.5 -14.7619 52.77 0.5 -14.7619 dpcand 0.5 -14.7619 dye 0.5 -14.7619 (boston 0.5 -14.7619 γeprime 0.5 -14.7619 ofv. 0.5 -14.7619 emotions—the 0.5 -14.7619 1500. 0.5 -14.7619 (jayaraman 0.5 -14.7619 afordances 0.5 -14.7619 ri,j, 0.5 -14.7619 rates; 0.5 -14.7619 enthymemes. 0.5 -14.7619 29.86 0.5 -14.7619 conceptresolver. 0.5 -14.7619 hurt. 0.5 -14.7619 arg2), 0.5 -14.7619 kneser-ney. 0.5 -14.7619 state). 0.5 -14.7619 5286 0.5 -14.7619 68.93 0.5 -14.7619 f1measure. 0.5 -14.7619 gallery, 0.5 -14.7619 (36 0.5 -14.7619 setting-3 0.5 -14.7619 sally 0.5 -14.7619 jaws. 0.5 -14.7619 hashtag, 0.5 -14.7619 article? 0.5 -14.7619 clips. 0.5 -14.7619 non-nested. 0.5 -14.7619 senti 0.5 -14.7619 abvd. 0.5 -14.7619 2002)). 0.5 -14.7619 (husain 0.5 -14.7619 organized. 0.5 -14.7619 1/ν. 0.5 -14.7619 0)] 0.5 -14.7619 google) 0.5 -14.7619 aux..vbn+vbd-prp 0.5 -14.7619 clinicians. 0.333333 -14.7625 s∈s 0.333333 -14.7625 resources (e.g. 0.4 -14.763 d≥1 0.210526 -14.7635 contributions are 0.5 -14.7636 ranking). 0.5 -14.7636 69.87 0.5 -14.7636 nevertheles 0.5 -14.7636 implementation). 0.5 -14.7636 71.25 0.5 -14.7636 (encoded 0.5 -14.7636 (gfs) 0.5 -14.7636 abstain 0.5 -14.7636 vunerable 0.5 -14.7636 black) 0.5 -14.7636 65.8% 0.5 -14.7636 socratic-style 0.5 -14.7636 alen’s 0.5 -14.7636 1261 0.5 -14.7636  0.5 -14.7636 eat; 0.5 -14.7636 self-references. 0.5 -14.7636 (imt) 0.5 -14.7636 ρn(·) 0.5 -14.7636 (%). 0.5 -14.7636 anaphoraantecedent 0.5 -14.7636 manyto-one 0.5 -14.7636 principled, 0.5 -14.7636 big. 0.5 -14.7636 1,...,l. 0.5 -14.7636 aggressively, 0.5 -14.7636 (karp, 0.5 -14.7636 torralba 0.5 -14.7636 drugs. 0.5 -14.7636 appropriateness, 0.5 -14.7636 (4.d), 0.5 -14.7636 mpbleu 0.5 -14.7636 monosyllabic 0.5 -14.7636 4150 0.5 -14.7636 initializer). 0.5 -14.7636 high; 0.5 -14.7636 pbrel 0.5 -14.7636 61.00 0.5 -14.7636 pneumonia. 0.5 -14.7636 71.1% 0.5 -14.7636 conference), 0.5 -14.7636 48.93 0.5 -14.7636 herbivore 0.5 -14.7636 ct(w). 0.5 -14.7636 selection” 0.5 -14.7636 longerdistance 0.5 -14.7636 journals. 0.5 -14.7636 46.00 0.5 -14.7636 q7 0.5 -14.7636 t.1 0.5 -14.7636 jokes. 0.5 -14.7636 singleword 0.5 -14.7636 61.62 0.5 -14.7636 result). 0.5 -14.7636 69.93 0.5 -14.7636 (lane 0.5 -14.7636 (respectively, 0.5 -14.7636 (vinh 0.5 -14.7636 71.20 0.5 -14.7636 topicbased 0.5 -14.7636 (token 0.5 -14.7636 erroneous. 0.5 -14.7636 rarely. 0.5 -14.7636 sh). 0.5 -14.7636 1..sd: 0.5 -14.7636 12.50% 0.5 -14.7636 pi∈e∗, 0.5 -14.7636 ozaidan@cs.jhu.edu 0.5 -14.7636 .681 0.5 -14.7636 topicalization. 0.5 -14.7636 −0.31triangleinv 0.5 -14.7636 ,...,fm 0.5 -14.7636 mt1. 0.5 -14.7636 wmpfbleu 0.5 -14.7636 transtype 0.5 -14.7636 pieced 0.5 -14.7636 1898 0.333333 -14.7637 butterworth 0.0852713 -14.7639 models were 0.4 -14.7644 role; 0.4 -14.7644 (noun, 0.0866142 -14.7646 a finite 0.107143 -14.7648 threshold. 0.333333 -14.7648 (50) 0.5 -14.7653 advs 0.5 -14.7653 forsyth 0.5 -14.7653 capmacro 0.5 -14.7653 familiarity. 0.5 -14.7653 wcpm 0.5 -14.7653 bye 0.5 -14.7653 10227 0.5 -14.7653 42.73 0.5 -14.7653 discounted. 0.5 -14.7653 appliances, 0.5 -14.7653 well-chosen 0.5 -14.7653 rprime. 0.5 -14.7653 ]= 0.5 -14.7653 (ignore, 0.5 -14.7653 (perl 0.5 -14.7653 43.83 0.5 -14.7653 82.96 0.5 -14.7653 abel.] 0.5 -14.7653 65.84 0.5 -14.7653 (tokenization, 0.5 -14.7653 69–77, 0.5 -14.7653 63.89 0.5 -14.7653 begr¨ußen. 0.5 -14.7653 enactment 0.5 -14.7653 data.10 0.5 -14.7653  flour 0.5 -14.7653 crisp 0.5 -14.7653 81.83 0.5 -14.7653 $150,000, 0.5 -14.7653 (0.89 0.5 -14.7653 shezaf 0.5 -14.7653 22–31, 0.5 -14.7653 honeymoon? 0.5 -14.7653 ter(w. 0.5 -14.7653 var err/mle 0.5 -14.7653 wpbleu 0.5 -14.7653 fold). 0.5 -14.7653 halt. 0.5 -14.7653 valencian 0.5 -14.7653 ace2004 0.5 -14.7653 (cuay´ahuitl 0.5 -14.7653 (0.14 0.5 -14.7653 childval(n1,n1′) 0.5 -14.7653 1.2) 0.5 -14.7653 check” 0.5 -14.7653 anova. 0.5 -14.7653 titles). 0.5 -14.7653 (morand, 0.5 -14.7653 attribute(s) 0.5 -14.7653 dor. 0.5 -14.7653 lirics 0.5 -14.7653 56.89 0.4 -14.7657 yusuke 0.333333 -14.7659 wen 0.5 -14.767 2.6% 0.5 -14.767 1985. 0.5 -14.767 avramidis 0.5 -14.767 dpphr. 0.5 -14.767 0.520 0.5 -14.767 “[subject] 0.5 -14.767 news) 0.5 -14.767 “counting” 0.5 -14.767 corasick, 0.5 -14.767 actionable. 0.5 -14.767 (crfs). 0.5 -14.767 low-level, 0.5 -14.767 61.93 0.5 -14.767 lay. 0.5 -14.767 87.15 0.5 -14.767 (hillardet 0.5 -14.767 almsaedat 0.5 -14.767 approximately. 0.5 -14.767 non-erroneous 0.5 -14.767 piano. 0.5 -14.767 (loeff 0.5 -14.767 brains. 0.5 -14.767 apartment. 0.5 -14.767 foxes 0.5 -14.767 (siddharthan, 0.5 -14.767 (zhou, 0.5 -14.767 adults. 0.5 -14.767 aimsweb 0.5 -14.767 project’s 0.5 -14.767 non-mwes. 0.5 -14.767 influenza. 0.5 -14.767 grammaticalinference. 0.5 -14.767 hedgehogs 0.5 -14.767 l(d,p) 0.5 -14.767 lexical-conceptual 0.5 -14.767 86.4% 0.5 -14.767 o(r,n,d,s). 0.5 -14.767 configurable, 0.5 -14.767 thurmair, 0.5 -14.767 13a, 0.5 -14.767 52. 0.5 -14.767 reconsider. 0.5 -14.767 52–60, 0.5 -14.767 35.99 0.5 -14.767 mdp 0.5 -14.767 72.72 0.5 -14.767 snow. 0.5 -14.767 karrowbothv(gprime1,gprime2), 0.5 -14.767 68.65 0.5 -14.767 haven 0.5 -14.767 type[sportsbike]) 0.5 -14.767 27.14 0.5 -14.767 (pirinen 0.4 -14.7671 0.955 0.210526 -14.7677 bought 0.333333 -14.7682 subsumer 0.4 -14.7684 xe 0.4 -14.7684 sameer 0.14 -14.7686 overcome this 0.5 -14.7687 injure, 0.5 -14.7687 nber, 0.5 -14.7687 true-case 0.5 -14.7687 96.1% 0.5 -14.7687 (szarvas 0.5 -14.7687 chats, 0.5 -14.7687 points.) 0.5 -14.7687 (hajiˇc, 0.5 -14.7687 synsetnodes 0.5 -14.7687 (cramer 0.5 -14.7687 87.0% 0.5 -14.7687 sij,aij,tij,rij 0.5 -14.7687 win4. 0.5 -14.7687 svr. 0.5 -14.7687 logp(x). 0.5 -14.7687 43.92 0.5 -14.7687 lost). 0.5 -14.7687 reduce) 0.5 -14.7687 machinery. 0.5 -14.7687 ‘verb’. 0.5 -14.7687 ‘tour’ 0.5 -14.7687 distractors) 0.5 -14.7687 66.4% 0.5 -14.7687 0.91. 0.5 -14.7687 49.61 0.5 -14.7687 1554 0.5 -14.7687 s∈n 0.5 -14.7687 (mel’ˇcuk, 0.5 -14.7687 (manevitz 0.5 -14.7687 whence 0.5 -14.7687 andk 0.5 -14.7687 0.005. 0.5 -14.7687 barnullba 0.5 -14.7687 180–189, 0.5 -14.7687 vation 0.5 -14.7687 2-21. 0.5 -14.7687 nations). 0.5 -14.7687 (milstein 0.5 -14.7687 bioinformatics. 0.5 -14.7687 svms) 0.5 -14.7687 cambodia, 0.5 -14.7687 streamlined 0.185185 -14.7692 significant (p 0.333333 -14.7693 (log 0.333333 -14.7693 universit`a 0.333333 -14.7693 1.54 0.333333 -14.7704 kaminski 0.112676 -14.7704 idea is to 0.5 -14.7704 evaluator. 0.5 -14.7704 66.54 0.5 -14.7704 /∈a. 0.5 -14.7704 wmf 0.5 -14.7704 temporal/causal 0.5 -14.7704 4200 0.5 -14.7704 27.93 0.5 -14.7704 takaaki 0.5 -14.7704 surveilance. 0.5 -14.7704 (lbr) 0.5 -14.7704 ground). 0.5 -14.7704 geographical, 0.5 -14.7704 modifiability 0.5 -14.7704 {〈z1,...,zn〉 0.5 -14.7704 satisfy. 0.5 -14.7704 smoothing” 0.5 -14.7704 gaming, 0.5 -14.7704 2.71 0.5 -14.7704 2.17 0.5 -14.7704 toppathsd(s)}. 0.5 -14.7704 it’. 0.5 -14.7704 true) 0.5 -14.7704 acero, 0.5 -14.7704 taxonomies. 0.5 -14.7704 orthographies. 0.5 -14.7704 capita 0.5 -14.7704 .662 0.5 -14.7704 deduction. 0.5 -14.7704 gap”. 0.5 -14.7704 pos-tgt 0.5 -14.7704 wmbleu 0.5 -14.7704 27.73 0.5 -14.7704 zagibalov 0.5 -14.7704 0.460 0.5 -14.7704 closely-matched 0.5 -14.7704 centroidbased 0.5 -14.7704 father. 0.5 -14.7704 lesser”. 0.5 -14.7704 (knight, 0.5 -14.7704 44, 0.5 -14.7704 96–104, 0.5 -14.7704 21). 0.333333 -14.7716 angel 0.5 -14.7721 agt−1 0.5 -14.7721 0.740. 0.5 -14.7721 setting; 0.5 -14.7721 (stickel, 0.5 -14.7721 feng. 0.5 -14.7721 hartu (‘he 0.5 -14.7721 68.79 0.5 -14.7721 athletes. 0.5 -14.7721 13-17 0.5 -14.7721 83.84 0.5 -14.7721 74.73 0.5 -14.7721 35.00 0.5 -14.7721 supplied. 0.5 -14.7721 “current 0.5 -14.7721 53–57, 0.5 -14.7721 pairs1. 0.5 -14.7721 š., 0.5 -14.7721 markets) 0.5 -14.7721 opinion/emotion 0.5 -14.7721 letters). 0.5 -14.7721 cu-tecto. 0.5 -14.7721 correspond. 0.5 -14.7721 φk, 0.5 -14.7721 -0.8353556 0.5 -14.7721 dream” 0.5 -14.7721 universit¨at stuttgart 0.5 -14.7721 utterance), 0.5 -14.7721 herbst, 0.5 -14.7721 27.83 0.5 -14.7721 conll. 0.5 -14.7721 (v,e). 0.5 -14.7721 93.0% 0.5 -14.7721 fellow. 0.5 -14.7721 correlation), 0.5 -14.7721 word-shape 0.5 -14.7721 65.67 0.5 -14.7721 1960; 0.5 -14.7721 wpf 0.5 -14.7721 mstparser’s 0.5 -14.7721 1.55% 0.4 -14.7725 donate 0.333333 -14.7727 scott, 0.333333 -14.7727 wikipedia data. 0.0943396 -14.773 negative. 0.0216138 -14.7735 this is 0.5 -14.7738 0.42. 0.5 -14.7738 81%, 0.5 -14.7738 mw). 0.5 -14.7738 λip 0.5 -14.7738 80.03++ 0.5 -14.7738 239–247, 0.5 -14.7738 authors' 0.5 -14.7738 wsj10, 0.5 -14.7738 40.49 0.5 -14.7738 comparable; 0.5 -14.7738 correct). 0.5 -14.7738 subclassof 0.5 -14.7738 .,i 0.5 -14.7738 oversimplification 0.5 -14.7738 2002b). 0.5 -14.7738 70.57 0.5 -14.7738 márquez 0.5 -14.7738 leak. 0.5 -14.7738 word-pair.” 0.5 -14.7738 (crf-d). 0.5 -14.7738 93.91 0.5 -14.7738 r 0.5 -14.7738 86.44 0.5 -14.7738 but. 0.5 -14.7738 two). 0.5 -14.7738 algorithm2. 0.5 -14.7738 deafness 0.5 -14.7738 minds. 0.5 -14.7738 barrachina 0.5 -14.7738 doing. 0.5 -14.7738 watch. 0.5 -14.7738 p(w). 0.5 -14.7738 1(·) 0.5 -14.7738 poss, 0.5 -14.7738 (kohli, 0.5 -14.7738 “cause” 0.5 -14.7738 (xi,xj). 0.210526 -14.7741 three annotators 0.5 -14.7755 d=log(1δ). 0.5 -14.7755 (do˘gan 0.5 -14.7755 equivalent: 0.5 -14.7755 ,...,a 0.5 -14.7755 (.33) 0.5 -14.7755 (liakata 0.5 -14.7755 51.89 0.5 -14.7755 t:±2 0.5 -14.7755 40.0% 0.5 -14.7755 zealand. 0.5 -14.7755 (1992)). 0.5 -14.7755 1756). 0.5 -14.7755 story). 0.5 -14.7755 krippendorff 0.5 -14.7755 mlbest 0.5 -14.7755 (0.8) 0.5 -14.7755 naist-jdic. 0.5 -14.7755 hope, 0.5 -14.7755 sentences.10 0.5 -14.7755 4,458 0.5 -14.7755 chap. 0.5 -14.7755 simplify, 0.0630252 -14.7766 one. 0.333333 -14.7772 0.256 0.5 -14.7772 59.93 0.5 -14.7772 83–91, 0.5 -14.7772 34.10 0.5 -14.7772 siblingsof 0.5 -14.7772 marrakech, morocco, 0.5 -14.7772 (.94) 0.5 -14.7772 remotely 0.5 -14.7772 47.84 0.5 -14.7772 abortion, 0.5 -14.7772 pf(wm1 0.5 -14.7772 43.07 0.5 -14.7772 getting. 0.5 -14.7772 presences 0.5 -14.7772 r 0.5 -14.7772 weinstein, 0.5 -14.7772 disguise 0.5 -14.7772 irover 0.5 -14.7772 shopping” 0.5 -14.7772 0.869 0.5 -14.7772 oceanic. 0.5 -14.7772 exponential, 0.5 -14.7772 90.0% 0.5 -14.7772 91.48 0.5 -14.7772 10000. 0.5 -14.7772 &reference-sequence. 0.5 -14.7772 reordering) 0.5 -14.7772 running’. 0.5 -14.7772 changes). 0.5 -14.7772 (hlt/emnlp), 0.5 -14.7772 ibw, 0.5 -14.7772 hall. 0.5 -14.7772 turkers) 0.0529412 -14.7773 divided 0.4 -14.7779 bei 0.5 -14.7789 (195). 0.5 -14.7789 85.1% 0.5 -14.7789 q-elab(a,b). 0.5 -14.7789 σ22 0.5 -14.7789 output), 0.5 -14.7789 0.789 0.5 -14.7789 5grams 0.5 -14.7789 29.14 0.5 -14.7789 cnt(w,t,ri) 0.5 -14.7789 heads) 0.5 -14.7789 30.99 0.5 -14.7789 80.23++ 0.5 -14.7789 bike?”. 0.5 -14.7789 researcher’s 0.5 -14.7789 29.26 0.5 -14.7789 dor 0.5 -14.7789 2.9. 0.5 -14.7789 describedby 0.5 -14.7789 31.00 0.5 -14.7789 (gw) 0.5 -14.7789 auxiliaries. 0.5 -14.7789 entailment/contradiction 0.5 -14.7789 pluscomp. 0.5 -14.7789 (louis 0.5 -14.7789 0.102 0.5 -14.7789 200,000-word 0.5 -14.7789 1230 0.5 -14.7789 readability). 0.5 -14.7789 eprk 0.5 -14.7789 (3.45%) 0.5 -14.7789 viewers 0.4 -14.7793 216594 0.4 -14.7793 97.4 0.333333 -14.7806 bendersky et 0.5 -14.7806 replays 0.5 -14.7806 k(·,·) 0.5 -14.7806 reasoners 0.5 -14.7806 regenerating 0.5 -14.7806 (1.5) 0.5 -14.7806 +5-gram/plsa 0.5 -14.7806 43.59 0.5 -14.7806 singers, 0.5 -14.7806 (1999)). 0.5 -14.7806 finite-state controller, 0.5 -14.7806 45.15 0.5 -14.7806 ≤n 0.5 -14.7806 denominator. 0.5 -14.7806 (tm). 0.5 -14.7806 caring 0.5 -14.7806 ndcg@1 0.5 -14.7806 (vaidya 0.5 -14.7806 betrag 0.5 -14.7806 74.07star 0.5 -14.7806 descriptions). 0.5 -14.7806 scissors 0.4 -14.7807 haifa 0.4 -14.7807 caused, 0.15 -14.781 council 0.166667 -14.7818 1.0 1.0 0.4 -14.782 form), 0.4 -14.782 wing 0.3 -14.7821 frequency distribution 0.5 -14.7823 blank. 0.5 -14.7823 76.4% 0.5 -14.7823 culturally 0.5 -14.7823 49.22 0.5 -14.7823 normed 0.5 -14.7823 subexpressions 0.5 -14.7823 sameannotator. 0.5 -14.7823 〈0,α,δ1,...,δj 0.5 -14.7823 morphf 0.5 -14.7823 83.38 0.4 -14.7834 tt, 0.5 -14.784 +0.75 0.5 -14.784 yg 0.5 -14.784 3.20 0.5 -14.784 ’10) 0.5 -14.784 dict-monofull 0.5 -14.784 55.1% 0.5 -14.784 possession. 0.5 -14.784 cnt. 0.5 -14.784 tokens’. 0.5 -14.784 (et) 0.5 -14.784 str. 0.5 -14.784 nullvj 0.5 -14.784 (sr) 0.5 -14.784 38.88 0.5 -14.784 provided). 0.5 -14.784 35.32 0.5 -14.784 iis-0968489. 0.5 -14.784 ward’s 0.5 -14.784 99.9% 0.5 -14.784 ⃗xj(2)qi 0.5 -14.784 40.67 0.272727 -14.7841 reestimate 0.333333 -14.7841 895 0.333333 -14.7841 geo:cityunionsqgeo:state) 0.333333 -14.7841 majid 0.4 -14.7848 pinto 0.333333 -14.7852 0.288 0.333333 -14.7852 5800 0.5 -14.7858 wrote, 0.5 -14.7858 poveda, mihai 0.5 -14.7858 n-grams), 0.5 -14.7858 adjunction). 0.5 -14.7858 mwes; 0.5 -14.7858 (convergent 0.5 -14.7858 (lexdiff) 0.5 -14.7858 nserc 0.5 -14.7858 37.48 0.5 -14.7858 discretion 0.5 -14.7858 (gr), 0.5 -14.7858 (schulte 0.4 -14.7862 n(n 0.4 -14.7862 with” 0.4 -14.7875 (anderson, 0.5 -14.7875 moses5 0.5 -14.7875 25.89 0.5 -14.7875 correspond’. 0.5 -14.7875 5,5and 7,7regular 0.5 -14.7875 (0.19) 0.5 -14.7875 wnuk, 0.5 -14.7875 ilp+bf+subj 0.5 -14.7875 macro-averaging. 0.5 -14.7875 3(a), 0.5 -14.7875 wordf 0.5 -14.7875 imp, exp〉)))) 0.5 -14.7875 .689 0.5 -14.7875 α+β 0.5 -14.7875 tntn 0.235294 -14.7879 which performs 0.333333 -14.7887 role” 0.4 -14.7889 xfer-orig 0.5 -14.7892 word/error 0.5 -14.7892 4.91 0.5 -14.7892 2521 (3.55%) 0.5 -14.7892 ucph-simple.de 0.5 -14.7892 “walked,” 0.5 -14.7892 nondeterministic, 0.5 -14.7892 incorrect”. 0.5 -14.7892 (toscano 0.5 -14.7892 epk 0.5 -14.7892 −1.5 0.5 -14.7892 stress: 0.5 -14.7892 1,257 0.5 -14.7892 theft 0.5 -14.7892 husain, 0.333333 -14.7898 y(xi) 0.192308 -14.7902 cavity 0.4 -14.7903 topd(w) 0.222222 -14.7906 schoenmackers 0.5 -14.7909 s32s66s111s117s110s100s101s100 s32s85s110s98s111s117s110s100s101s100 0.5 -14.7909 more/less 0.5 -14.7909 day” 0.5 -14.7909 hmm). 0.5 -14.7909 http://www.cs.pitt.edu/mpqa 0.5 -14.7909 π34 0.5 -14.7909 concurs, 0.5 -14.7909 !!! 0.5 -14.7909 walton’s scheme-set, 0.166667 -14.791 we also thank 0.4 -14.7917 86.25 0.4 -14.7917 75.3 0.333333 -14.7921 'do' 0.5 -14.7927 66.29 0.5 -14.7927 guide’s 0.5 -14.7927 wij. 0.5 -14.7927 dqg'dq.ohlq 0.5 -14.7927 (cw). 0.5 -14.7927 0.3). 0.5 -14.7927 (t). 0.5 -14.7927 91.89 0.5 -14.7927 mt). 0.5 -14.7927 93.72 0.5 -14.7927 bests 0.5 -14.7927 approachin 0.5 -14.7927 51.5%, 0.5 -14.7927 (complement, friends) 0.5 -14.7927 colocation, 0.5 -14.7927 gmf2, 0.235294 -14.7928 2008a; 0.4 -14.7931 toronto, 0.4 -14.7931 université 0.3 -14.7934 particular. 0.5 -14.7944 ribeiro-neto, 0.5 -14.7944 providesa 0.5 -14.7944 uncertaininput/surprisal 0.5 -14.7944 purity (pu), 0.5 -14.7944 mat)). 0.5 -14.7944 1525 0.5 -14.7944 c(t,e) 0.5 -14.7944 pending. 0.5 -14.7944 msr’ 0.5 -14.7944 loc. 0.5 -14.7944 savova 0.5 -14.7944 87.24 0.5 -14.7944 smoothing (zhai 0.4 -14.7945 crowdflower, 0.122807 -14.7945 tweets are 0.4 -14.7958 hartu (‘to 0.4 -14.7958 #the 0.5 -14.7961 60.83 0.5 -14.7961 0.666 0.5 -14.7961 p.o.box 0.5 -14.7961 v2.0). 0.5 -14.7961 resultsand 0.5 -14.7961 window), 0.5 -14.7961 630 0.5 -14.7961 (db) 0.5 -14.7961 gmf 0.5 -14.7961 qlf, 0.5 -14.7961 “there” 0.222222 -14.7968 population. 0.333333 -14.7968 site(entity), 0.333333 -14.7968 piwek 0.4 -14.7972 ˆσ 0.4 -14.7972 2.43 0.5 -14.7979 52.91 0.5 -14.7979 u5: 0.5 -14.7979 delx 0.5 -14.7979 4-5). 0.5 -14.7979 lincoln. 0.5 -14.7979 akm 0.5 -14.7979 songs. 0.5 -14.7979 canes 0.5 -14.7979 run-2 0.5 -14.7979 hats 0.5 -14.7979 run-1 0.5 -14.7979 gorjanc, 0.5 -14.7979 times: 0.4 -14.7986 jingbo 0.4 -14.7986 k7t 0.3 -14.7987 p5r5 0.333333 -14.7991 nerve 0.5 -14.7996 (he/she 0.5 -14.7996 gazetteers), 0.5 -14.7996 small.) 0.5 -14.7996 says. 0.5 -14.7996 summationdisplay b∈σ 0.5 -14.7996 (dda) 0.5 -14.7996 86.93 0.5 -14.7996 30%). 0.5 -14.7996 leusch 0.5 -14.7996 (en-dir.) 0.108108 -14.8001 however, they 0.0278005 -14.8009 compare 0.222222 -14.8014 josef 0.4 -14.8014 2.74 0.4 -14.8014 holger 0.333333 -14.8014 [predicate 0.5 -14.8014 setting-1 0.5 -14.8014 tactile 0.5 -14.8014 initialy 0.5 -14.8014 oil. 0.5 -14.8014 1629 0.5 -14.8014 (nass 0.5 -14.8014 prior), 0.5 -14.8014 cohen. 0.5 -14.8014 0.158 0.5 -14.8014 album) 0.5 -14.8014 40.72 0.5 -14.8014 pregnant 0.4 -14.8028 σ(k,lscript) 0.4 -14.8028 proper-noun 0.5 -14.8031 segment(c,θ) 0.5 -14.8031 91.72 0.5 -14.8031 mayfield, 0.5 -14.8031 ‘u’ 0.5 -14.8031 960 0.5 -14.8031 3.9, 0.5 -14.8031 wanted. 0.5 -14.8031 overlap; 0.5 -14.8031 1174 0.5 -14.8031 sidered 0.5 -14.8031 countr(c) 0.5 -14.8031 9each 0.5 -14.8031 mer, 0.5 -14.8031 (odds) 0.5 -14.8031 48.05 0.5 -14.8031 93.66 0.5 -14.8031 englishfrench 0.5 -14.8031 65.96 0.5 -14.8031 excluded). 0.5 -14.8031 pos-grams 0.5 -14.8031 spotted, 0.333333 -14.8037 65.47 0.333333 -14.8037 t/c 0.333333 -14.8037 independence, 0.4 -14.8042 3.74 0.0497382 -14.8044 a list 0.272727 -14.8049 standard) 0.333333 -14.8049 y∈g(x) 0.333333 -14.8049 sofie 0.5 -14.8049 conceptualise 0.5 -14.8049 “made 0.5 -14.8049 85.02 0.5 -14.8049 45.91 0.5 -14.8049 ,σ 0.5 -14.8049 97.44 0.5 -14.8049 np#2. 0.5 -14.8049 helped, 0.5 -14.8049 〈s, 0.5 -14.8049 hyp. 0.5 -14.8049 858 0.5 -14.8049 disambiguation,” 0.5 -14.8049 chen2010all 0.4 -14.8056 θdoc 0.106667 -14.806 gives an 0.333333 -14.8061 rover 0.5 -14.8066 frazier, 0.5 -14.8066 78.8% 0.5 -14.8066 34.74 0.5 -14.8066 ambiguous), 0.5 -14.8066 journalisten 0.5 -14.8066 52.14 0.5 -14.8066 wins, 0.5 -14.8066 extractionitself. 0.5 -14.8066 95.14 0.5 -14.8066 f(xi,xj) 0.5 -14.8066 0.851 0.5 -14.8066 (aylett 0.5 -14.8066 puzicha, 0.5 -14.8066 for). 0.5 -14.8066 7.47 0.4 -14.807 c” 0.333333 -14.8072 none), 0.4 -14.8084 (katz, 0.4 -14.8084 context-identifier 0.5 -14.8084 pwnv 0.5 -14.8084 (lf/uf), 0.5 -14.8084 75%. 0.5 -14.8084 years? 0.5 -14.8084 generatedsentences 0.5 -14.8084 (lr/ur). 0.5 -14.8084 29.03 0.5 -14.8084 technology tallaght 0.5 -14.8084 btg-hmd, 0.5 -14.8084 protein_catabolism 0.5 -14.8084 34.17 0.5 -14.8084 0.67). 0.5 -14.8084 soj 0.5 -14.8084 j∈i− 0.333333 -14.8096 łq 0.4 -14.8099 liwc+bigrams+svm 0.4 -14.8099 texans 0.4 -14.8099 +mono 0.132075 -14.81 ter bleu 0.5 -14.8102 444 0.5 -14.8102 43.31 0.5 -14.8102 (di) 0.5 -14.8102 (felbaum, 0.5 -14.8102 virtual therapist, 0.5 -14.8102 zhazj, 0.166667 -14.8108 involved. 0.5 -14.8119 wangm, 0.5 -14.8119 testset, 0.5 -14.8119 (mackinlay 0.5 -14.8119 37.67 0.5 -14.8119 92.07 0.5 -14.8119 85.76 0.5 -14.8119 cristo”. 0.5 -14.8119 verbnet1 0.5 -14.8119 chunks; 0.5 -14.8119 m00 0.5 -14.8119 dt(dt 0.5 -14.8119 (metzler 0.5 -14.8119 (singular, 0.5 -14.8119 to”). 0.0378549 -14.8121 we show 0.4 -14.8127 unsup, 0.4 -14.8127 yor, 0.210526 -14.8128 sections 3 0.333333 -14.8131 giovanni 0.333333 -14.8131 are averaged. 0.5 -14.8137 macwhiney, 0.5 -14.8137 46.55 0.5 -14.8137 29.63% 0.5 -14.8137 1484 0.5 -14.8137 “bass” 0.5 -14.8137 3100 0.5 -14.8137 appointments 0.5 -14.8137 52.67 0.5 -14.8137 0.481 0.5 -14.8137 coverage), 0.5 -14.8137 syntax: 0.333333 -14.8143 aki 0.333333 -14.8143 engages 0.0797101 -14.8149 second, we 0.4 -14.8155 p(w1 0.4 -14.8155 nel 0.4 -14.8155 π(b) 0.5 -14.8155 30.79 0.5 -14.8155 section3. 0.5 -14.8155 (mle). 0.5 -14.8155 definition), 0.5 -14.8155 0.519 0.5 -14.8155 chronicling 0.5 -14.8155 83.8% 0.5 -14.8155 vlpf 0.5 -14.8155 80.12 0.5 -14.8155 wsjnull 0.5 -14.8155 wordnet), 0.5 -14.8155 in– 0.5 -14.8155 (gonz´alez-rubio 0.5 -14.8155 65.66 0.5 -14.8155 1459 0.5 -14.8155 37.7% 0.5 -14.8155 1983)) 0.4 -14.8169 -m 0.4 -14.8169 0.40.6 0.4 -14.8169 light(x) 0.5 -14.8172 (forsbom, 0.5 -14.8172 bg882>e dg882>e 0.5 -14.8172 nullo,onullnull 0.5 -14.8172 761 0.5 -14.8172 81.90 0.5 -14.8172 92.4% 0.5 -14.8172 inequalities) 0.5 -14.8172 unlikely, 0.5 -14.8172 co-hyponymy, meronymy, 0.5 -14.8172 90th 0.5 -14.8172 -0.80 0.5 -14.8172 swap-rates 0.5 -14.8172 93.10 0.5 -14.8172 l2,1 0.210526 -14.8173 4.2 experimental 0.333333 -14.8179 f,e 0.4 -14.8184 (expressing 0.4 -14.8184 levenshteindistance 0.4 -14.8184 organization) 0.5 -14.819 r(z)i 0.5 -14.819 coloring 0.5 -14.819 15.3% 0.5 -14.819 simplification), 0.5 -14.819 3,727 0.5 -14.819 (order 0.5 -14.819 wps 0.5 -14.819 ibm-bleu 0.5 -14.819 combig.pl 0.5 -14.819 64.78 0.5 -14.819 29.36 0.5 -14.819 cue-scope 0.5 -14.819 “private 0.4 -14.8198 semi-learnt 0.333333 -14.8202 carmen 0.333333 -14.8202 “killed” 0.222222 -14.8203 service. 0.5 -14.8208 denotational 0.5 -14.8208 100; 0.5 -14.8208 0.540 0.5 -14.8208 -0.17 0.5 -14.8208 56.88 0.5 -14.8208 cccp 0.5 -14.8208 2.65 0.5 -14.8208 (pre.), 0.5 -14.8208 24.67 0.5 -14.8208 first-class citizens 0.5 -14.8208 pinkal 0.5 -14.8208 _多 _少 0.5 -14.8208 {wi} 0.5 -14.8208 (velikovich 0.4 -14.8212 wolverhampton 0.0884956 -14.822 rules from 0.5 -14.8226 topic” 0.5 -14.8226 occur; 0.5 -14.8226 perfect” 0.5 -14.8226 1160 0.5 -14.8226 93.53 0.25 -14.8244 adp 0.5 -14.8244 0.915 0.5 -14.8244 “false positives” 0.5 -14.8244 40.32 0.5 -14.8244 overwrote 0.5 -14.8244 newstest 0.5 -14.8244 1567 0.5 -14.8244 "head 0.5 -14.8244 70.47 0.5 -14.8244 zqt 0.5 -14.8244 (arbitrary-image) 0.5 -14.8244 inferred, 0.5 -14.8244 49.66 0.5 -14.8244 lists dpi, 0.5 -14.8244 84.4% 0.5 -14.8244 dependency-native 0.5 -14.8244 bpcs; 0.5 -14.8244 sites” 0.5 -14.8244 36.83 0.5 -14.8244 41.55 0.5 -14.8244 kr¨anze 0.4 -14.8255 1,224 0.4 -14.8255 traf2 0.106667 -14.826 also have 0.5 -14.8262 norwegian, 0.5 -14.8262 sch 0.5 -14.8262 (synonyms, 0.5 -14.8262 29.0% 0.5 -14.8262 “eighth 0.5 -14.8262 (kuich 0.5 -14.8262 subfamilies 0.5 -14.8262 nege 0.5 -14.8262 26.19 0.5 -14.8262 86.84 0.5 -14.8262 ∀null((null(null) ∧nullnullnullnullnullnullnullnullnullnullnullnullnull 0.5 -14.8262 yoshimura, 0.5 -14.8262 click-words 0.5 -14.8262 1precision 0.5 -14.8262 acceptor 0.5 -14.8262 41.4% 0.132075 -14.8274 bold. 0.142857 -14.8277 create an 0.5 -14.8279 50.14 0.5 -14.8279 13.20 0.5 -14.8279 published) 0.5 -14.8279 chunker-based 0.5 -14.8279 θh, 0.5 -14.8279 (non-digit, 0.5 -14.8279 sentence-meaning 0.5 -14.8279 52,746 0.5 -14.8279 pbta. 0.5 -14.8279 2009-02-10: 0.5 -14.8279 null-labeled 0.5 -14.8279 brady, 0.5 -14.8279 paolillo, 0.3 -14.8284 (du 0.4 -14.8284 “i’m 0.4 -14.8284 yj, 0.185185 -14.8286 (zhu 0.333333 -14.8286 automation, 0.5 -14.8297 accelerated 0.5 -14.8297 recscore+bias: 0.5 -14.8297 11.98 0.5 -14.8297 ‘complex’ 0.5 -14.8297 maid 0.5 -14.8297 per-annotator 0.5 -14.8297 (giampiccolo 0.5 -14.8297 60.06 0.5 -14.8297 9it 0.5 -14.8297 dispersal 0.5 -14.8297 en→de 0.5 -14.8297 0.176 0.5 -14.8297 1075 0.5 -14.8297 92.8% 0.5 -14.8297 91.65 0.5 -14.8297 theme; 0.5 -14.8297 pleasant) 0.5 -14.8297 something; 0.2 -14.8298 cases when 0.0532915 -14.8309 than one 0.5 -14.8315 k. saravanan, 0.5 -14.8315 3while 0.5 -14.8315 internetslang 0.5 -14.8315 24.6% 0.5 -14.8315 3.3.2. 0.5 -14.8315 0.3) 0.5 -14.8315 21.36 0.5 -14.8315 6ltr 0.5 -14.8315 hospitals. 0.5 -14.8315 trivially, 0.5 -14.8315 sauri 0.5 -14.8315 134,001 0.5 -14.8315 962 0.5 -14.8315 42.01 0.5 -14.8315 (0.8 0.5 -14.8315 (rwk). 0.5 -14.8315 0.392 0.5 -14.8315 93.37 0.5 -14.8315 rq1 0.5 -14.8315 499 0.5 -14.8315 dyads. 0.5 -14.8315 2.57 0.5 -14.8315 1,540 0.0581395 -14.832 null null 0.4 -14.8327 q∗t(f) 0.5 -14.8333 59.56 0.5 -14.8333 injected, 0.5 -14.8333 watchstander 0.5 -14.8333 5-gram/plsa 0.5 -14.8333 zooming 0.5 -14.8333 thejth 0.5 -14.8333 0.0116 0.5 -14.8333 pools, 0.5 -14.8333 9,961,795 0.5 -14.8333 optional; 0.5 -14.8333 idiosyncracies 0.5 -14.8333 (precision, 0.5 -14.8333 (5c) 0.5 -14.8333 fuhr, 0.5 -14.8333 (rychl´y 0.5 -14.8333 3,589 0.4 -14.8341 mx 0.133333 -14.8346 the reference. 0.333333 -14.8346 fabio 0.5 -14.8352 lamb, 0.5 -14.8352 http://web.science.mq.edu.au/ 0.5 -14.8352 +ann 0.5 -14.8352 began, 0.5 -14.8352 19-year 0.5 -14.8352 violated; 0.5 -14.8352 (8,000 0.5 -14.8352 disable 0.5 -14.8352 unwind 0.5 -14.8352 ’s’ 0.5 -14.8352 goaltest 0.5 -14.8352 0.891 0.5 -14.8352 ‘treasure 0.5 -14.8352 40.83 0.4 -14.8356 “’s” 0.333333 -14.8358 non-propagation 0.3 -14.8359 cjj 0.5 -14.837 91.21 0.5 -14.837 −→θ 0.5 -14.837 (matching 0.5 -14.837 computer” 0.5 -14.837 l∞,1 0.5 -14.837 3school 0.5 -14.837 p(target sentm1 0.5 -14.837 affiliations 0.5 -14.837 s− 0.4 -14.8385 1,400 0.5 -14.8388 place; 0.5 -14.8388 silliman. 0.5 -14.8388 carroll. 0.5 -14.8388 94.55 0.5 -14.8388 6,129 0.5 -14.8388 entites 0.5 -14.8388 test-52 0.5 -14.8388 66.18 0.5 -14.8388 1-500 0.5 -14.8388 exampleλ-hybrid 0.5 -14.8388 experimentson 0.5 -14.8388 92.86 0.5 -14.8388 iha 0.5 -14.8388 eventuality, 0.5 -14.8388 egnppa 0.272727 -14.839 gme 0.139535 -14.839 studies in 0.0918367 -14.8397 an evaluation 0.4 -14.8399 6.2.1 0.4 -14.8399 ei, 0.4 -14.8399 noisy-seed 0.333333 -14.8406 86.69 0.5 -14.8406 flag, 0.5 -14.8406 25.96 0.5 -14.8406 (xia, 0.5 -14.8406 (mrf) 0.5 -14.8406 94.51 0.5 -14.8406 .039 0.5 -14.8406 environment; 0.5 -14.8406 foi 0.5 -14.8406 1993)), 0.5 -14.8406 duan07 0.5 -14.8406 (nullnullfsnullnullnulllabnullnullnullnull) 0.5 -14.8406 npceditor, 0.3 -14.8414 aplied to 0.210526 -14.8416 subtree. 0.333333 -14.8418 affiliation 0.5 -14.8424 0.911 0.5 -14.8424 #12_months 0.5 -14.8424 37.93 28.97 0.5 -14.8424 pasr(u) 0.5 -14.8424 484 0.5 -14.8424 basedmethodcombinedwithseveraleffective featuresis 0.5 -14.8424 wuensch" 0.5 -14.8424 dalin’s 0.5 -14.8424 g,f,s 0.5 -14.8424 undetected 0.5 -14.8424 purposes; 0.5 -14.8424 entailmentstyle 0.5 -14.8424 (required); 0.5 -14.8424 50.63 0.5 -14.8424 federico sangati 0.5 -14.8424 african-american 0.5 -14.8424 670–675, 0.5 -14.8424 decp 0.5 -14.8424 rie 0.333333 -14.8431 alef 0.0827068 -14.8437 to rank 0.5 -14.8442 (1999)), 0.5 -14.8442 well) 0.5 -14.8442 verb+particle 0.5 -14.8442 77.20 0.5 -14.8442 3.98 0.5 -14.8442 birthplace 0.5 -14.8442 syntactically) 0.5 -14.8442 superset, 0.5 -14.8442 production-grade 0.5 -14.8442 mice. 0.5 -14.8442 3600 0.25 -14.8443 jennifer 0.333333 -14.8443 “after 0.333333 -14.8443 85.9 0.172414 -14.8451 to accomplish 0.139535 -14.8451 using standard 0.333333 -14.8455 ugly 0.5 -14.8461 1.12 0.5 -14.8461 folds) 0.5 -14.8461 chen2010bi 0.5 -14.8461 mext 0.5 -14.8461 in-links 0.5 -14.8461 normative 0.5 -14.8461 worksheet 0.5 -14.8461 doc, 0.5 -14.8461 45.18 0.5 -14.8461 1032 0.5 -14.8461 fatigued 0.5 -14.8461 723 0.230769 -14.8471 its simplicity, 0.4 -14.8472 excellatach 0.25 -14.8479 fpr 0.230769 -14.8479 second-best 0.5 -14.8479 pc3 0.5 -14.8479 danforth, 0.5 -14.8479 dominic 0.5 -14.8479 altered, 0.5 -14.8479 s-d-tfidf 0.5 -14.8479 clean-up 0.5 -14.8479 84.57 0.5 -14.8479 pseen 0.5 -14.8479 occurring) 0.5 -14.8479 iniesta 0.5 -14.8479 (dodds 0.3 -14.848 threw 0.4 -14.8487 malouf 0.5 -14.8497 lscript1/lscriptp-regularized 0.5 -14.8497 pattern’s 0.5 -14.8497 0.297 0.5 -14.8497 boprime 0.5 -14.8497 83.49 0.5 -14.8497 39.14 0.5 -14.8497 dlji 0.5 -14.8497 intentional, 0.5 -14.8497 (bendersky 0.5 -14.8497 what phone number 0.5 -14.8497 58.1% 0.5 -14.8497 maryland. 0.5 -14.8497 fain, 0.5 -14.8497 ({on},{may,june}) 0.4 -14.8501 0.128 0.4 -14.8501 ‘bilal 0.333333 -14.8504 h1: 0.5 -14.8516 knee 0.5 -14.8516 cruys, 0.5 -14.8516 27.16 0.5 -14.8516 ‘supervised’ 0.5 -14.8516 post-procesing 0.5 -14.8516 sad). 0.5 -14.8516 (hit); 0.5 -14.8516 “secondary 0.5 -14.8516 (183 0.5 -14.8516 54.71 0.5 -14.8516 43.13 0.5 -14.8516 ”almost 0.5 -14.8516 fully, 0.0523077 -14.8518 deal 0.272727 -14.852 ore 0.4 -14.8531 jr., 0.4 -14.8531 buckets: 0.5 -14.8534 55.62 0.5 -14.8534 64.71 0.5 -14.8534 bpk=25 0.5 -14.8534 stuck, 0.5 -14.8534 -context 0.5 -14.8534 whenusing goldstandardmentionboundaries,weachieve 0.5 -14.8534 “test 0.5 -14.8534 (corpus) 0.5 -14.8534 61.35 0.5 -14.8534 -1.03 0.5 -14.8534 25.94 0.5 -14.8534 “add-one” 0.5 -14.8534 41.85 0.5 -14.8534 77.9% 0.3 -14.8546 n=2 0.4 -14.8546 0.276 0.5 -14.8552 saxon 0.5 -14.8552 frailty 0.5 -14.8552 2,147 0.5 -14.8552 misra 0.5 -14.8552 +hv) 0.5 -14.8552 widely, 0.5 -14.8552 0.227 0.5 -14.8552 #5 0.5 -14.8552 eleventh 0.5 -14.8552 occurence 0.5 -14.8552 “hard 0.5 -14.8552 1c-40l 0.25 -14.8553 fronted 0.4 -14.856 ££day 0.4 -14.856 fem 0.4 -14.856 “blue 0.230769 -14.8564 thread. 0.5 -14.8571 beled 0.5 -14.8571 “upstairs” 0.5 -14.8571 -0.55 0.5 -14.8571 14.7% 0.5 -14.8571 5,375 0.5 -14.8571 57.74 0.5 -14.8571 zitouni 0.5 -14.8571 antibiotics 0.5 -14.8571 (bodenstab 0.5 -14.8571 92.49 0.4 -14.8575 1department 0.5 -14.8589 87.1% 0.5 -14.8589 green), 0.5 -14.8589 67.18 0.5 -14.8589 packaging 0.5 -14.8589 edge-counting 0.5 -14.8589 42.3% 0.5 -14.8589 500), 0.5 -14.8589 85.93 0.5 -14.8589 markp 0.5 -14.8589 mile 0.3 -14.859 (chambers 0.3 -14.859 loc, 0.4 -14.859 (fleiss, 0.4 -14.859 3rd-person 0.230769 -14.859 szpektor et 0.190476 -14.86 what do 0.4 -14.8605 89.53 0.4 -14.8605 bulls 0.5 -14.8608 72.13 0.5 -14.8608 (t1) 0.5 -14.8608 (4.c), 0.5 -14.8608 “lect” 0.5 -14.8608 1990): 0.5 -14.8608 2.2.5 0.5 -14.8608 ’negative’ 0.5 -14.8608 3.73 0.118644 -14.861 respectively). 0.4 -14.862 snake 0.4 -14.862 collin 0.5 -14.8627 12.77% 0.5 -14.8627 dict-hmmp-basic 0.5 -14.8627 rulesets 0.5 -14.8627 thankful 0.5 -14.8627 1307 0.5 -14.8627 jan. 0.5 -14.8627 summationtextprior 0.5 -14.8627 token-level: 0.5 -14.8627 categorize, 0.5 -14.8627 57.66% 0.5 -14.8627 lf, 0.5 -14.8627 attr span 0.5 -14.8627 72.02 0.5 -14.8627 34.62 0.5 -14.8627 configurability 0.333333 -14.8639 ionic 0.5 -14.8645 (larson, 0.5 -14.8645 structuring: 0.5 -14.8645 (devlin 0.5 -14.8645 littered 0.5 -14.8645 (lower) 0.5 -14.8645 (205), 0.5 -14.8645 near-miss 0.5 -14.8645 0.02), 0.5 -14.8645 65.04 0.5 -14.8645 s[wq]/(s[dcl]/np) 0.5 -14.8645 kocos.pl 0.5 -14.8645 proportionality 0.5 -14.8645 within-domain 0.5 -14.8645 a0-5 0.5 -14.8645 most-false-xb 0.5 -14.8645 0.167 0.333333 -14.8652 d'melo, 0.333333 -14.8652 e-step, 0.05 -14.8655 provide a 0.230769 -14.8658 differ. 0.4 -14.8664 ∗⇒ 0.5 -14.8664 (jonquet 0.5 -14.8664 0.485 0.5 -14.8664 (mte) 0.5 -14.8664 persuaded 0.5 -14.8664 §3.1), 0.5 -14.8664 ncet (ncet-08-0006). 0.5 -14.8664 40,674 0.5 -14.8664 50-headword, 10-cluster 0.5 -14.8664 3723 0.5 -14.8664 follow-up, 0.5 -14.8664 misunderstood, 0.5 -14.8664 (analogously 0.5 -14.8664 47.80 0.333333 -14.8677 0.723 0.333333 -14.8677 (and) 0.4 -14.8679 top-scoring 0.5 -14.8682 maintenance, 0.5 -14.8682 thursday, 0.5 -14.8682 machinery, 0.5 -14.8682 presented: 0.5 -14.8682 2-best 0.5 -14.8682 ha-2: 0.5 -14.8682 jaist, 0.5 -14.8682 85.37 0.5 -14.8682 4.05 0.5 -14.8682 migration, 0.5 -14.8682 (roark, 0.5 -14.8682 “gappy 0.5 -14.8682 ounis, 0.5 -14.8701 (0.80 0.5 -14.8701 diagnosed, 0.5 -14.8701 basis); 0.5 -14.8701 scales: 0.5 -14.8701 89.13 0.5 -14.8701 day’s 0.5 -14.8701 (46%), 0.5 -14.8701 playlist 0.5 -14.8701 split: 0.5 -14.8701 s0l 0.130435 -14.8712 accounted 0.139535 -14.872 2010a). 0.5 -14.872 -fr 0.5 -14.872 88.63 93.52 0.5 -14.872 (watkinson 0.5 -14.872 faust+1n 0.5 -14.872 berria 0.5 -14.872 4.97 0.5 -14.872 tilt, 0.5 -14.872 (string 0.5 -14.872 f∈d 0.5 -14.872 tomaz.erjavec@ijs.si 0.5 -14.872 blurs 0.5 -14.872 28.91 0.5 -14.872 vaguely 0.5 -14.872 paninian 0.5 -14.872 42.03 0.5 -14.872 st〉→〈[vp 0.5 -14.872 84.38 0.5 -14.872 pain” (code 0.5 -14.872 officials, 0.5 -14.872 nodes/tree 0.5 -14.872 15.2% 0.272727 -14.8722 wmt11. 0.4 -14.8739 p(x,y)p(x)p(y) 0.4 -14.8739 z12 0.25 -14.8739 (balahur 0.5 -14.8739 0.613 0.5 -14.8739 studied; 0.5 -14.8739 urgent 0.5 -14.8739 mt05, mt06, 0.5 -14.8739 0.339 0.5 -14.8739 chomskian 0.5 -14.8739 “turkey” 0.5 -14.8739 interpretations: 0.5 -14.8739 ppattachment 0.5 -14.8739 gr¨un 0.5 -14.8739 32,000 0.5 -14.8739 called, 0.5 -14.8739 0.203 0.5 -14.8739 featured, 0.5 -14.8739 i-bart 0.0758621 -14.8748 features were 0.333333 -14.8752 adja 0.118644 -14.8757 june, 0.5 -14.8757 annotator-algorithm 0.5 -14.8757 (wd) 0.5 -14.8757 expedient 0.5 -14.8757 amsterdam: 0.5 -14.8757 1892, 0.5 -14.8757 w1,...,wk 0.5 -14.8757 short-time 0.5 -14.8757 ∈bon−1(wi−11 0.5 -14.8757 pictorially 0.5 -14.8757 13.00 0.5 -14.8757 all-features 0.5 -14.8757 sweeping 0.5 -14.8757 38.62 0.5 -14.8757 φ1,...,φk 0.25 -14.8758 tease 0.146341 -14.8765 are almost 0.4 -14.8769 ―ق ‖ 0.0677083 -14.8773 resulted 0.5 -14.8776 +15 0.5 -14.8776 comp.arg1 0.5 -14.8776 sj¨obergh 0.5 -14.8776 1/5 0.5 -14.8776 statistical) 0.5 -14.8776 accurac 0.5 -14.8776 grammars) 0.5 -14.8776 unigrams), 0.5 -14.8776 ‘glance.v’ 0.5 -14.8776 swsdmjl 0.5 -14.8776 cell) 0.5 -14.8776 gefoltert 0.5 -14.8776 scope-underspecified 0.5 -14.8776 verschwunden 0.5 -14.8776 (probabilistic) 0.5 -14.8776 sub/ok 0.4 -14.8784 speculations 0.4 -14.8784 #train 0.5 -14.8795 unvisited 0.5 -14.8795 ipr 0.5 -14.8795 (søgaard 0.5 -14.8795 v* 0.5 -14.8795 61.48 0.5 -14.8795 visualises 0.5 -14.8795 (tokenization 0.5 -14.8795 lessened 0.5 -14.8795 nicola.cancedda@xrce.xerox.com 0.5 -14.8795 40.70 0.5 -14.8795 bmt+aux+mterater 0.5 -14.8795 capitol 0.5 -14.8795 tagger; 0.5 -14.8795 (blumkulka 0.5 -14.8795 enter-s2 0.5 -14.8795 singular; 0.5 -14.8795 7); 0.5 -14.8795 indicative, 0.5 -14.8795 wethen 0.5 -14.8795 †indicates 0.5 -14.8795 (qu 0.5 -14.8795 trivial: 0.5 -14.8795 database; 0.5 -14.8795 4null2 0.5 -14.8795 rekishi 0.4 -14.88 2,4 0.333333 -14.8802 cd(s) 0.333333 -14.8814 premise, 0.5 -14.8814 49,000 0.5 -14.8814 21.5% 0.5 -14.8814 p(y) 0.5 -14.8814 concept-by-concept 0.5 -14.8814 captured; 0.5 -14.8814 ‘listen.v’, 0.5 -14.8814 evidences2. 0.5 -14.8814 ˆs2 0.5 -14.8814 (oh 0.5 -14.8814 commented, 0.5 -14.8814 mzhang, 0.5 -14.8814 fna 0.5 -14.8814 12.8% 0.5 -14.8814 (lp 0.5 -14.8814 breaks) 0.5 -14.8814 (meth) 0.5 -14.8814 71.3% 0.5 -14.8814 (9)) 0.5 -14.8814 rights: 0.5 -14.8814 newssyscombtune 0.25 -14.8815 target-to-source 0.0891089 -14.8817 has focused 0.4 -14.883 landauer, 0.5 -14.8833 df(wi) 0.5 -14.8833 costa-juss`a 0.5 -14.8833 zdenˇek 0.5 -14.8833 lemma-only 0.5 -14.8833 premodifiers, 0.5 -14.8833 formalizations 0.5 -14.8833 8.24 0.5 -14.8833 r% 0.5 -14.8833 randy 0.5 -14.8833 (0.10 0.5 -14.8833 78.4% 0.5 -14.8833 30.42 0.5 -14.8833 grandparent, 0.5 -14.8833 kernels; 0.5 -14.8833 6.1m 0.5 -14.8833 pz 0.5 -14.8833 esprc 0.5 -14.8833 approachesin 0.12766 -14.8834 reasons, 0.333333 -14.884 josé 0.142857 -14.8842 recall precision 0.333333 -14.8852 hana, 0.5 -14.8852 flickr 0.5 -14.8852 m. teresa 0.5 -14.8852 2,966,274 0.5 -14.8852 newscorp 0.5 -14.8852 stop-words, 0.5 -14.8852 hebrew), 0.5 -14.8852 steepest 0.5 -14.8852 90.69 0.5 -14.8852 (62%), 0.5 -14.8852 functor, 0.5 -14.8852 31.5% 0.5 -14.8852 0.65, 0.5 -14.8852 xmwe) 0.5 -14.8852 graeme. 0.5 -14.8852 (booth 0.5 -14.8852 1,906 0.5 -14.8852 beagle 0.5 -14.8852 dependently 0.5 -14.8852 micciulla, 0.5 -14.8852 (vendler, 0.5 -14.8852 16.3% 0.5 -14.8852 63.92 0.5 -14.8852 l(θ,α) 0.5 -14.8852 hli}@i2r.a-star.edu.sg 0.5 -14.8852 edge-cover 0.5 -14.8852 makhoul. 0.044186 -14.8854 kind 0.4 -14.886 entity; 0.333333 -14.8865 erlbaum 0.5 -14.8871 9in 0.5 -14.8871 human: 0.5 -14.8871 iqr 0.5 -14.8871 (9%) 0.5 -14.8871 existential, 0.5 -14.8871 bnf 0.5 -14.8871 “xylitol” 0.5 -14.8871 mikhail, 0.5 -14.8871 concesion 0.5 -14.8871 netbooks, 0.5 -14.8871 fpos 0.5 -14.8871 ∗∗” 0.5 -14.8871 pirates 0.5 -14.8871 −2 10−1 0.5 -14.8871 non-p 0.5 -14.889 lacking, 0.5 -14.889 brown: 0.5 -14.889 82.9% 0.5 -14.889 top-five 0.5 -14.889 40.41 0.5 -14.889 pavlik 0.5 -14.889 ochiai 0.5 -14.889 buffer”, 0.5 -14.889 super-classes 0.5 -14.889 5-letter 0.5 -14.889 1600, 0.5 -14.889 48.68 0.5 -14.889 (e,f,a) 0.5 -14.889 17.9 0.5 -14.889 +ts 0.5 -14.889 38.92 0.5 -14.889 almuhareb-poesio 0.5 -14.889 stretched 0.5 -14.889 (417) 0.5 -14.889 74.60 0.5 -14.889 41.96 0.5 -14.889 7,384 0.333333 -14.8903 32.68 0.4 -14.8906 err 0.5 -14.8909 dear 0.5 -14.8909 overbearing 0.5 -14.8909 substitution-only 0.5 -14.8909 655 0.5 -14.8909 (dl 0.5 -14.8909 5×104 0.5 -14.8909 pa-pcfg+features 0.5 -14.8909 :h,hprime: 0.5 -14.8909 zhekova 0.5 -14.8909 loss functionl(e,eprime) 0.5 -14.8909 antibody 0.5 -14.8909 guang 0.5 -14.8909 pos-1 0.5 -14.8909 capitalism 0.5 -14.8909 so+gi 0.5 -14.8909 “compromise” 0.5 -14.8909 that-classification 0.25 -14.891 tomorrow 0.0516129 -14.8917 depends 0.4 -14.8921 0.605 0.5 -14.8928 budgets, 0.5 -14.8928 lakers, 0.5 -14.8928 longman, 0.5 -14.8928 1.6k 0.5 -14.8928 553 0.5 -14.8928 non-questions, 0.5 -14.8928 sdev, 0.5 -14.8928 goiserian 0.5 -14.8928 entailment: 0.5 -14.8928 paumier, 0.5 -14.8928 48.82 0.5 -14.8928 88.94 0.5 -14.8928 monolingual-based 0.5 -14.8928 king saud 0.5 -14.8928 conductance, 0.166667 -14.8929 approach uses 0.333333 -14.8929 diversify 0.0595745 -14.8936 for future 0.230769 -14.8937 compositional. 0.333333 -14.8942 (magnini 0.0748299 -14.8943 restrict 0.181818 -14.8946 sebastiani, 0.5 -14.8948 exactly; 0.5 -14.8948 sasha 0.5 -14.8948 χ2, 0.5 -14.8948 kipper-schuler, 0.5 -14.8948 equal-length 0.5 -14.8948 uq 0.5 -14.8948 1246 0.5 -14.8948 upv/ehu 0.5 -14.8948 scl+500bio 0.5 -14.8948 tufte 0.5 -14.8948 {[−4, 0.5 -14.8948 prlm-train 0.5 -14.8948 4.39 0.5 -14.8948 ne), 0.5 -14.8948 clasified 0.5 -14.8948 nt) 0.5 -14.8948 26.57 0.5 -14.8948 heiden” 0.5 -14.8948 mohanty 0.5 -14.8948 20.4 0.5 -14.8948 ptminer 0.5 -14.8948 “cotton.” 0.5 -14.8948 incomparable): 0.5 -14.8948 type" 0.4 -14.8952 chmm 0.4 -14.8967 cheese, eggs, 0.4 -14.8967 boring 0.333333 -14.8967 jigs 0.333333 -14.8967 tout 0.5 -14.8967 reading-time 0.5 -14.8967 “touch 0.5 -14.8967 outwith 0.5 -14.8967 “class 0.5 -14.8967 alrand 0.5 -14.8967 (unique) 0.5 -14.8967 [−3, 0.5 -14.8967 exp, 0.5 -14.8967 term), 0.5 -14.8967 sinclair 0.5 -14.8967 visuo-spatial 0.5 -14.8967 flat) 0.5 -14.8967 iraq, 0.5 -14.8967 %bleu 0.5 -14.8967 kaplan, 0.5 -14.8967 level” 0.5 -14.8967 noised 0.5 -14.8967 p<.001) 0.5 -14.8967 hybridexts 0.5 -14.8967 gamble 0.5 -14.8967 1language 0.5 -14.8967 1357 0.5 -14.8967 tag-cloud 0.333333 -14.898 genomes 0.181818 -14.8981 shifters 0.0882353 -14.8983 treating 0.5 -14.8986 27.10 0.5 -14.8986 judgement, 0.5 -14.8986 0.0163 0.5 -14.8986 element; 0.5 -14.8986 (lkb) 0.5 -14.8986 01, 0.5 -14.8986 whcr 0.5 -14.8986 38.5% 0.5 -14.8986 multiplication (×) 0.5 -14.8986 24,000 0.5 -14.8986 dataset1 0.5 -14.8986 treebank), 0.5 -14.8986 (195) 0.5 -14.8986 29.98 0.5 -14.8986 “eric 0.5 -14.8986 (actions), 0.5 -14.8986 distributor 0.5 -14.8986 augmented comunicator´s 0.5 -14.8986 provider, 0.4 -14.8998 marie 0.5 -14.9005 unsel, 0.5 -14.9005 fold; 0.5 -14.9005 (mae) 0.5 -14.9005 rebellion 0.5 -14.9005 [my] 0.5 -14.9005 55.63 0.5 -14.9005 unrest 0.5 -14.9005 (learned 0.5 -14.9005 frege’s 0.5 -14.9005 55.7% 0.5 -14.9005 kotlerman 0.5 -14.9005 sloven 0.5 -14.9005 leaders, 0.5 -14.9005 sha-1 0.5 -14.9005 locale 0.5 -14.9005 (xie 0.5 -14.9005 (none 0.333333 -14.9006 −log 0.272727 -14.9013 y(x) 0.4 -14.9014 nullnullnullnull. 0.4 -14.9014 (articles 0.4 -14.9014 30.03 0.333333 -14.9019 c[s:i] 0.5 -14.9025 alas, 0.5 -14.9025 (objects, 0.5 -14.9025 56.53 0.5 -14.9025 hpsg-based 0.5 -14.9025 imaginable 0.5 -14.9025 introductions 0.5 -14.9025 exp1: 0.5 -14.9025 25.14 0.5 -14.9025 iteration) 0.5 -14.9025 iso-timeml, 0.5 -14.9025 0.304 0.5 -14.9025 89.6% 0.5 -14.9025 33.76 0.5 -14.9025 anil thakur. 0.5 -14.9025 (knlm), 0.5 -14.9025 premiere 0.5 -14.9025 87.54 0.5 -14.9025 (cann 0.5 -14.9025 coordinates) 0.5 -14.9025 605,906 0.5 -14.9025 “spoken 0.4 -14.9029 0.661 0.272727 -14.9034 “human 0.190476 -14.9041 joint conference 0.5 -14.9044 25-word 0.5 -14.9044 (accuracy), 0.5 -14.9044 rapport, 0.5 -14.9044 toll 0.5 -14.9044 ’private’ 0.5 -14.9044 41.90 0.5 -14.9044 64.37 0.5 -14.9044 0.923 0.5 -14.9044 71.50 0.5 -14.9044 latvia 0.5 -14.9044 “+unsup” 0.5 -14.9044 substree 0.5 -14.9044 “methylation” 0.5 -14.9044 85.7(93.3) 0.5 -14.9044 framefiles 0.5 -14.9044 president-elect 0.4 -14.9045 (obozinski 0.4 -14.9045 0.454 0.0331565 -14.9051 the final 0.25 -14.9054 table 10 0.4 -14.906 (men) 0.0865385 -14.9063 we hypothesize 0.5 -14.9063 48.64 0.5 -14.9063 iepa 0.5 -14.9063 0], 0.5 -14.9063 γ′ 0.5 -14.9063 under, 0.5 -14.9063 submodular: 0.5 -14.9063 (key-list) 0.5 -14.9063 lock 0.5 -14.9063 change), 0.5 -14.9063 silence) 0.5 -14.9063 (r/p/f) 0.5 -14.9063 (gmm), 0.5 -14.9063 pal, 0.5 -14.9063 .01 -fb 0.5 -14.9063 1.97 0.5 -14.9063 1705 0.5 -14.9063 “save” 0.5 -14.9063 jargon, 0.5 -14.9063 keyphrases: 0.173913 -14.9075 raghavan et 0.2 -14.9075 bionlp’11 0.25 -14.9083 wj, 0.5 -14.9083 0.33, 0.5 -14.9083 endeavour, 0.5 -14.9083 (prepositions, 0.5 -14.9083 ens 0.5 -14.9083 8note 0.5 -14.9083 (meng, 0.5 -14.9083 cs-svm, 0.5 -14.9083 9our 0.5 -14.9083 ncm 0.5 -14.9083 (sumita 0.5 -14.9083 14.58 0.5 -14.9083 preor 0.5 -14.9083 abstracts) 0.5 -14.9083 lifts 0.5 -14.9083 intelligible, 0.5 -14.9083 (edus) 0.5 -14.9083 main, 0.5 -14.9083 chunking: 0.5 -14.9083 harmonium 0.5 -14.9083 fundamentally, 0.5 -14.9083 corine 0.5 -14.9083 3.8% 0.5 -14.9083 villing, 0.5 -14.9083 43.8% 0.5 -14.9083 grandmother. 0.5 -14.9083 70.68 0.5 -14.9083 wsk+csk 0.0561798 -14.9085 level. 0.285714 -14.9089 taskease 0.214286 -14.9102 model training. 0.5 -14.9102 unlabeled), 0.5 -14.9102 cataloguing 0.5 -14.9102 (tsl) 0.5 -14.9102 2003).8 0.5 -14.9102 deco 0.5 -14.9102 stephen, 0.5 -14.9102 barbieri, 0.5 -14.9102 uptraining 0.5 -14.9102 62.10 0.5 -14.9102 53.94 0.5 -14.9102 nullwnull 0.5 -14.9102 47.72 0.5 -14.9102 (ggp) 0.5 -14.9102 l1regularization 0.5 -14.9102 n1,2 0.5 -14.9102 (feruci 0.5 -14.9102 0.758 0.5 -14.9102 lenses 0.5 -14.9102 4,026 0.5 -14.9102 co-click 0.5 -14.9102 “document” 0.5 -14.9102 59.21 0.4 -14.9107 synchronous, 0.333333 -14.9109 k∏ 0.210526 -14.9114 programming (ilp) 0.272727 -14.9119 [b,c] 0.4 -14.9122 product-producer 0.5 -14.9122 computationaly 0.5 -14.9122 63.69 0.5 -14.9122 (expert 0.5 -14.9122 1466 0.5 -14.9122 (pages 0.5 -14.9122 “back 0.5 -14.9122 42.40 0.5 -14.9122 sic 0.5 -14.9122 featuresand 0.5 -14.9122 1140 0.5 -14.9122 unigrams) 0.5 -14.9122 (peterson, 0.5 -14.9122 “left 0.5 -14.9122 lctwl 0.5 -14.9122 52.94 0.5 -14.9122 storytelling 0.5 -14.9122 50.17 0.5 -14.9122 semantics-oriented 0.5 -14.9122 ‘simple’ 0.5 -14.9122 intra-document 0.5 -14.9122 20-newsgroupsaccuracy 0.5 -14.9122 medelyan, 0.272727 -14.913 b+-tree 0.5 -14.9141 implement: 0.5 -14.9141 s1,...,sn 0.5 -14.9141 eyewitness 0.5 -14.9141 rs2 0.5 -14.9141 legal, 0.5 -14.9141 ontology) 0.5 -14.9141 counter-intuitive, 0.5 -14.9141 addison-wesley, 0.5 -14.9141 crosslinguistic 0.5 -14.9141 headtag 0.5 -14.9141 (5-gram 0.5 -14.9141 0.3577 0.5 -14.9141 22.8% 0.5 -14.9141 meronomy 0.5 -14.9141 ‘+dynamic’ 0.5 -14.9141 4.53 0.5 -14.9141 limsi/cnrs 0.5 -14.9141 1324 0.5 -14.9141 modeland 0.097561 -14.9142 each topic 0.0332889 -14.9151 set, 0.0613208 -14.9157 this work was 0.333333 -14.9161 #word 0.5 -14.9161 (seed 0.5 -14.9161 gold-original 0.5 -14.9161 (kaljurand 0.5 -14.9161 0.2615 0.5 -14.9161 1.41% 0.5 -14.9161 (galanis 0.5 -14.9161 np-chunked 0.5 -14.9161 (ranta, 0.5 -14.9161 75.2% 0.5 -14.9161 modifier; 0.5 -14.9161 qualifying 0.5 -14.9161 pixel 0.5 -14.9161 (abeill´e, 0.5 -14.9161 action”) 0.5 -14.9161 on-trunk 0.5 -14.9161 (agg) 0.4 -14.9169 4-5 0.4 -14.9169 tft 0.214286 -14.9169 (maamouri 0.333333 -14.9174 duan, 0.285714 -14.9179 excellent, 0.285714 -14.9179 ‘and 0.5 -14.9181 accr 0.5 -14.9181 1327 0.5 -14.9181 din 0.5 -14.9181 hallucinated 0.5 -14.9181 social-media 0.5 -14.9181 sequel 0.5 -14.9181 strongest, 0.5 -14.9181 zeichen 0.5 -14.9181 hours) 0.5 -14.9181 colorado, 0.5 -14.9181 chin 0.5 -14.9181 framefile 0.5 -14.9181 0.541 0.5 -14.9181 (§3.2) 0.5 -14.9181 p/ 0.5 -14.9181 now) 0.0236167 -14.9184 as well 0.4 -14.9185 these combinations 0.15625 -14.9186 and strube, 0.5 -14.92 schegloff 0.5 -14.92 15this 0.5 -14.92 mainew-simpleew 0.5 -14.92 has-property, 0.5 -14.92 ready, 0.5 -14.92 out-ofcontext 0.5 -14.92 sublexical 0.5 -14.92 [380] 0.5 -14.92 bayesnet 0.5 -14.92 15.19 0.5 -14.92 levin-style 0.5 -14.92 (feghali, 0.5 -14.92 confidence-rated 0.5 -14.92 (contrast 0.5 -14.92 [the] 0.5 -14.92 69.46 0.5 -14.92 international herald 0.5 -14.92 seeds; 0.5 -14.92 ££geographic-area 0.5 -14.92 preposition-specific 0.4 -14.9201 paske 0.230769 -14.9206 5 7 0 . 0.4 -14.9217 pfc 0.4 -14.9217 rebase 0.5 -14.922 � 0.5 -14.922 ths 0.5 -14.922 news-test2010 0.5 -14.922 gallows 0.5 -14.922 61.22 0.5 -14.922 16.1k 0.5 -14.922 query-query 0.5 -14.922 dispatching 0.5 -14.922 not: 0.5 -14.922 foaf 0.5 -14.922 apartments 0.5 -14.922 1608 0.5 -14.922 72.4% 0.5 -14.922 maxent-based 0.5 -14.922 60second 0.5 -14.922 (frawley, 0.5 -14.922 magnets 0.5 -14.922 (duc), 0.5 -14.922 4.71 0.5 -14.922 cand(x) 0.5 -14.922 0.078 0.5 -14.922 p+ 0.5 -14.922 (probabilistic 0.5 -14.922 chronicle 0.5 -14.922 mean; 0.5 -14.922 (mbrsc) 0.5 -14.922 {altaf,vince}@hlt.utdallas.edu 0.5 -14.922 xæ“k 0.5 -14.922 pvalues 0.5 -14.922 left-endoc, 0.5 -14.922 (v,n1,p,n2) 0.25 -14.923 d. manning. 0.0561798 -14.9234 space. 0.333333 -14.924 (banea et 0.5 -14.924 avgkl 0.5 -14.924 wikipedians 0.5 -14.924 63.40 0.5 -14.924 unambiguous: 0.5 -14.924 expended 0.5 -14.924 (s[wq]/pp)/n 0.5 -14.924 wij) 0.5 -14.924 (paraphrase) 0.5 -14.924 (15.b) 0.5 -14.924 547) 0.5 -14.924 anette 0.5 -14.924 dist(i−j) 0.5 -14.924 /b/-to-/d/ 0.5 -14.924 79.57 0.5 -14.924 look: 0.5 -14.924 1428 0.5 -14.924 (h), 0.5 -14.924 18.30 0.5 -14.924 disease/symptom 0.5 -14.924 upt 0.5 -14.924 0.412 0.5 -14.924 synthesized tweet: 0.5 -14.924 colege 0.5 -14.924 unwieldy 0.5 -14.924 (53%) 0.5 -14.924 type-1 0.5 -14.924 unix, 0.5 -14.924 posbefore 0.5 -14.924 non-equivalence 0.5 -14.924 (weese 0.5 -14.924 court, 0.5 -14.924 reveals, 0.0769231 -14.9246 all features 0.4 -14.9248 (international 0.333333 -14.9253 nadejda mihaylova 0.112903 -14.9253 are learned 0.285714 -14.9258 warranted 0.5 -14.926 12.78% 0.5 -14.926 kills 0.5 -14.926 subdomain, 0.5 -14.926 oosten 0.5 -14.926 1553 0.5 -14.926 (included 0.5 -14.926 37.60 0.5 -14.926 kidnapping 0.5 -14.926 frozenness 0.5 -14.926 n< 0.5 -14.926 measurability 0.5 -14.926 stadthagen-gonzalez 0.5 -14.926 http://nlp.stanford.edu/software/segmenter.shtml 0.5 -14.926 registers, 0.5 -14.926 vectort 0.5 -14.926 utepict 0.5 -14.926 polarity; 0.4 -14.9264 “hat” 0.097561 -14.9265 parsing model 0.038817 -14.9278 2006). 0.5 -14.9279 hlt-coe 0.5 -14.9279 archivist 0.5 -14.9279 recorded, 0.5 -14.9279 models,on 0.5 -14.9279 bloody, 0.5 -14.9279 vbp, 0.5 -14.9279 “virtual” 0.5 -14.9279 lucene5 0.5 -14.9279 28k 0.5 -14.9279 n(n−1) 0.5 -14.9279 collocating 0.5 -14.9279 36.04 0.5 -14.9279 ove 0.5 -14.9279 power) 0.5 -14.9279 195; 0.5 -14.9279 seu[their] 0.5 -14.9279 6×108 0.5 -14.9279 min-yen 0.5 -14.9279 “explanation” 0.5 -14.9279 margolis 0.5 -14.9279 vds-7: 0.5 -14.9279 organizing, 0.5 -14.9279 970 0.5 -14.9279 imagability, 0.5 -14.9279 regrouping 0.285714 -14.928 [1,4] 0.285714 -14.928 plain-text 0.4 -14.928 thirteenth 0.151515 -14.9287 ∈ v 0.333333 -14.9293 judges) 0.333333 -14.9293 e-step: 0.333333 -14.9293 u.n. 0.4 -14.9296 87.07 0.5 -14.9299 scenario designers 0.5 -14.9299 (machine, 0.5 -14.9299 investigatory 0.5 -14.9299 4.18 0.5 -14.9299 400) 0.5 -14.9299 dominey, 0.5 -14.9299 -2.30 0.5 -14.9299 crabb´e, 0.5 -14.9299 machine-learning-based 0.5 -14.9299 dg882>k 0.5 -14.9299 cell-specific 0.5 -14.9299 verbalize 0.5 -14.9299 i[ti↔tj] 0.5 -14.9299 ptrs 0.5 -14.9299 sponsors 0.5 -14.9299 83.41 0.5 -14.9299 prediciton 0.5 -14.9299 apparent, 0.5 -14.9299 enlarges 0.5 -14.9299 (as) 0.5 -14.9299 +prop 0.5 -14.9299 difficult-to-read 0.5 -14.9299 81.02 0.5 -14.9299 curator 0.5 -14.9299 apconsolidate(s) 0.5 -14.9299 frankenstein 0.5 -14.9299 pilots 0.5 -14.9299 bg882>k 0.5 -14.9299 overestimated 0.5 -14.9299 participantsand 0.5 -14.9299 sonorants, 0.5 -14.9299 arena 0.4 -14.9312 45.5% 0.5 -14.9319 recognizer). 0.5 -14.9319 worst-case runtime. 0.5 -14.9319 25.80 0.5 -14.9319 fortuitous 0.5 -14.9319 issued, 0.5 -14.9319 csr, 0.5 -14.9319 yazdani 0.5 -14.9319 [1,n] 0.5 -14.9319 nonhead 0.5 -14.9319 uncached 0.5 -14.9319 paradise, 0.5 -14.9319 coreferenceresolution,we 0.5 -14.9319 flour, 0.5 -14.9319 ratner 0.5 -14.9319 barely, 0.5 -14.9319 uniform) 0.5 -14.9319 6,800 0.5 -14.9319 ls(hw,g) 0.5 -14.9319 ’phone 0.5 -14.9319 ‘give’ 0.5 -14.9319 differential, 0.5 -14.9319 tackle: 0.5 -14.9319 quoted, 0.5 -14.9319 flames 0.5 -14.9319 linktypebasedpre-cluster 0.5 -14.9319 experimentswith 0.5 -14.9319 phonologies 0.5 -14.9319 51.15 0.5 -14.9319 “@p 0.5 -14.9319 2.87 0.5 -14.9319 prepositionsin 0.5 -14.9319 (tomita 0.5 -14.9319 mbdp−1 ngs−u 0.5 -14.9319 yleaf 0.5 -14.9319 boxplot 0.5 -14.9319 14.29% 0.5 -14.9319 (sna) 0.5 -14.9319 cros-language 0.5 -14.9319 tagger) 0.5 -14.9319 29.55 0.5 -14.9319 (syntax 0.5 -14.9319 bmt+aux 0.5 -14.9319 “feature” 0.5 -14.9319 (barker 0.1 -14.9324 so, we 0.285714 -14.9326 preservers 0.285714 -14.9326 day) 0.0927835 -14.9328 be translated 0.4 -14.9328 presidents, 0.333333 -14.9333 doer 0.333333 -14.9333 ‘chains’ 0.5 -14.9339 standards) 0.5 -14.9339 46.66 0.5 -14.9339 16.09 0.5 -14.9339 disimilar 0.5 -14.9339 non-oracle 0.5 -14.9339 eq.5 0.5 -14.9339 co., 0.5 -14.9339 mpc: 0.5 -14.9339 scg 0.5 -14.9339 conll2010 0.5 -14.9339 kounye 0.5 -14.9339 threshold), 0.5 -14.9339 eparl+nc+1092+ir 0.5 -14.9339 verb-object-subject 0.5 -14.9339 2g 0.5 -14.9339 827 0.5 -14.9339 [sic] 0.5 -14.9339 span-based 0.5 -14.9339 16.93% 0.5 -14.9339 adj/noun2 0.5 -14.9339 believer 0.5 -14.9339 1711 0.5 -14.9339 1996-1999 0.5 -14.9339 contender 0.0882353 -14.9339 all sentences 0.4 -14.9344 tj, 0.333333 -14.9346 pedro 0.214286 -14.9348 countries. 0.0865385 -14.9353 wall 0.5 -14.9359 ariya 0.5 -14.9359 lampert 0.5 -14.9359 matrixspace 0.5 -14.9359 dq, 0.5 -14.9359 (miler 0.5 -14.9359 (osgood 0.5 -14.9359 #data 0.5 -14.9359 “just 0.5 -14.9359 17.9%, 0.5 -14.9359 (global) 0.5 -14.9359 lapse 0.5 -14.9359 emulates 0.5 -14.9359 dev: 0.5 -14.9359 ‘accept’ 0.5 -14.9359 vectorv 0.5 -14.9359 prints 0.5 -14.9359 4457 0.5 -14.9359 ld(t) 0.5 -14.9359 siskind, 0.5 -14.9359 nakov@comp.nus.edu.sg 0.5 -14.9359 magnate 0.5 -14.9359 0.312 0.5 -14.9359 famously 0.5 -14.9359 07], 0.5 -14.9359 (p). 0.5 -14.9359 distribution” 0.5 -14.9359 none_of_the_above 0.285714 -14.936 bertoldi, 0.5 -14.9379 50.26 0.5 -14.9379 1,113 0.5 -14.9379 percentile, 0.5 -14.9379 1148 0.5 -14.9379 soften 0.5 -14.9379 ti(x) 0.5 -14.9379 (redlawsk, 0.5 -14.9379 wikinyt-5·106 0.5 -14.9379 “labels” 0.5 -14.9379 1966 0.5 -14.9379 highprecision 0.5 -14.9379 wysiwyg 0.5 -14.9379 arg2-loc 0.5 -14.9379 (morphological) 0.5 -14.9379 1,438 0.5 -14.9379 977,069,902 0.5 -14.9379 prosody: 0.5 -14.9379 >ojv-mt3 0.5 -14.9851 addreses 0.5 -14.9851 1,716 0.5 -14.9851 31.6% 0.5 -14.9851 2.6.1 0.5 -14.9851 ssa) 0.5 -14.9851 joan 0.5 -14.9851 + 0.1*(-0.75, 0.5 -14.9851 rion 0.5 -14.9851 21.80 0.5 -14.9851 vocalic 0.5 -14.9851 tail: 0.0691824 -14.9852 which may 0.5 -14.9872 67.2% 0.5 -14.9872 0.952 0.5 -14.9872 david, 0.5 -14.9872 compressions: 0.5 -14.9872 dame 0.5 -14.9872 scoreasr(w) 0.5 -14.9872 n(0,1), 0.5 -14.9872 “eat”, 0.5 -14.9872 34.36 0.5 -14.9872 rinaldi 0.5 -14.9872 79.05 0.5 -14.9872 53.60 0.5 -14.9872 25.37 0.5 -14.9872 firstlevel 0.5 -14.9872 x(p(pii)) 0.5 -14.9872 prlg, 0.5 -14.9872 mert-tuned 0.5 -14.9872 veale, 0.5 -14.9872 stem) 0.5 -14.9872 27.12 0.5 -14.9872 general: 0.5 -14.9872 en2fr 0.5 -14.9872 n(n girl) 0.5 -14.9872 54,496 0.5 -14.9872 similarly). 0.5 -14.9872 o.jwj2/ 0.5 -14.9872 32.55 0.5 -14.9872 (“labeled”) 0.5 -14.9872 blatant 0.285714 -14.9876 zou 0.1 -14.9877 2011 shared 0.190476 -14.9879 likelihood estimation 0.5 -14.9893 mesh25 0.5 -14.9893 33.66 0.5 -14.9893 heidorn, 0.5 -14.9893 hierarchial 0.5 -14.9893 enlargement 0.5 -14.9893 94.20 0.5 -14.9893 qsh,t 0.5 -14.9893 tokan 0.5 -14.9893 66.56 0.5 -14.9893 penalizing, 0.5 -14.9893 structure-sharing 0.5 -14.9893 540 0.5 -14.9893 ablating 0.5 -14.9893 laplace) 0.5 -14.9893 bob, 0.5 -14.9893 (nuggets) 0.5 -14.9893 defending 0.5 -14.9893 (250 0.5 -14.9893 burdened 0.5 -14.9893 11.4% 0.5 -14.9893 anthology, 0.5 -14.9893 hit), 0.5 -14.9893 (friedman, 0.5 -14.9893 nod 0.5 -14.9893 concerts 0.5 -14.9893 824 0.5 -14.9893 popel, 0.5 -14.9893 acrossthe 0.5 -14.9893 roads 0.5 -14.9893 (words), 0.5 -14.9914 corpus-extracted 0.5 -14.9914 predictive, 0.5 -14.9914 -0.28 0.5 -14.9914 up; 0.5 -14.9914 “reject” 0.5 -14.9914 sys17 0.5 -14.9914 dag, 0.5 -14.9914 89.3%/94.0% 0.5 -14.9914 np/nn 0.5 -14.9914 gmf2 0.5 -14.9914 1,4 0.5 -14.9914 (events, 0.5 -14.9914 conll-2005 0.5 -14.9914 n-gram wni 0.5 -14.9914 ,” 0.5 -14.9914 wasting 0.5 -14.9914 happens: 0.5 -14.9914 (poveda 0.5 -14.9914 1.61 0.5 -14.9914 snow’s 0.5 -14.9914 motorcycle 0.5 -14.9914 (quirk, 0.5 -14.9914 reformed 0.5 -14.9914 efficiently: 0.147059 -14.9916 15th 0.056 -14.9925 ensure 0.113208 -14.9929 it only 0.5 -14.9935 d∈d(t) 0.5 -14.9935  0.5 -14.9935 cubojar 0.5 -14.9935 valuable, 0.5 -14.9935 1,274 0.5 -14.9935 undecidable 0.5 -14.9935 prolog, 0.5 -14.9935 (ceafe+muc+b3)/3 0.5 -14.9935 mover’s 0.5 -14.9935 76, 0.5 -14.9935 73.3% 0.5 -14.9935 mentioned), 0.5 -14.9935 ‘noise’ 0.5 -14.9935 07 0.5 -14.9935 70.69 0.5 -14.9935 akaike 0.5 -14.9935 27.50 0.5 -14.9935 domain/genre 0.5 -14.9935 anlscript1 0.5 -14.9935 stronger, 0.5 -14.9935 serbian, 0.5 -14.9935 8although 0.5 -14.9935 propagation; 0.5 -14.9935 nagy, 0.5 -14.9935 46.8% 0.5 -14.9935 annotator1-gold 0.5 -14.9935 smileys, 0.285714 -14.9936 -measure 0.5 -14.9957 performace 0.5 -14.9957 gazetteer, 0.5 -14.9957 barrage 0.5 -14.9957 82.85 0.5 -14.9957 “variation detection” 0.5 -14.9957 pv 0.5 -14.9957 (surface 0.5 -14.9957 “binding” 0.5 -14.9957 (uptk) 0.5 -14.9957 85.38 0.5 -14.9957 2.35 0.5 -14.9957 reverser 0.5 -14.9957 pitchrange 0.5 -14.9957 57.1% 0.5 -14.9957 3−5 0.5 -14.9957 singleor 0.5 -14.9957 〈xprime,ξprime〉matches 0.5 -14.9957 “problem” 0.5 -14.9957 sander 0.5 -14.9957 1214 0.5 -14.9957 elkan 0.5 -14.9957 extract: 0.5 -14.9957 aces 0.5 -14.9957 monoboot 0.5 -14.9957 0.077 0.5 -14.9957 hypernymic 0.333333 -14.9964 +0.07 0.285714 -14.9973 exacerbated 0.333333 -14.9978 culties 0.5 -14.9978 unconscious 0.5 -14.9978 cdac 0.5 -14.9978 80.03 0.5 -14.9978 (62.5% 0.5 -14.9978 0,9 0.5 -14.9978 semantex 0.5 -14.9978 (rst), 0.5 -14.9978 bi-phrases, 0.5 -14.9978 john] 0.5 -14.9978 pprime) 0.5 -14.9978 nn*” 0.5 -14.9978 bengston 0.5 -14.9978 46.46 0.5 -14.9978 project3 0.5 -14.9978 asynchronous, 0.5 -14.9978 (bc 0.5 -14.9978 e* 0.5 -14.9978 (pps), 0.5 -14.9978 set2 0.5 -14.9978 67.66 0.5 -14.9978 encompassed 0.5 -14.9978 recei 0.5 -14.9978 true-casing 0.5 -14.9978 (yngve, 0.5 -14.9978 commissive, 0.5 -14.9978 44.97 0.5 -14.9978 streams: 0.5 -14.9978 evaluated: 0.5 -14.9978 year.sea 0.173913 -14.998 compound. 0.181818 -14.9989 sum, 0.230769 -14.9991 (grosz 0.2 -14.9995 tratz 0.5 -14.9999 isnot 0.5 -14.9999 namespace 0.5 -14.9999 removeinputfilesafterprocessing 0.5 -14.9999 coreference-task 0.5 -14.9999 79.52 0.5 -14.9999 at(α,λ) 0.5 -14.9999 hasfeature 0.5 -14.9999 1556 0.5 -14.9999 0.956 0.5 -14.9999 customizable 0.5 -14.9999 well), 0.5 -14.9999 aquaint, 0.5 -14.9999 dish 0.5 -14.9999 79.41 0.5 -14.9999 2140 0.5 -14.9999 “flight” 0.5 -14.9999 pronouns), 0.5 -14.9999 implicit) 0.5 -14.9999 3-point 0.5 -14.9999 tacked 0.5 -14.9999 pages), 0.5 -14.9999 s0r, 0.5 -14.9999 kivran-swaine 0.5 -14.9999 7using 0.5 -14.9999 as-is 0.5 -14.9999 pekar 0.0609137 -15.0014 department of computer 0.5 -15.002 1e 0.5 -15.002 1076 0.5 -15.002 tant 0.5 -15.002 (ciw) 0.5 -15.002 l0-norm 0.5 -15.002 /p 0.5 -15.002 (rank 0.5 -15.002 sign) 0.5 -15.002 →a 0.5 -15.002 banks, 0.5 -15.002 (cucerzan07) 0.5 -15.002 accurate) 0.5 -15.002 hurt) 0.5 -15.002 information.3 0.5 -15.002 75.39 0.5 -15.002 stringw 0.5 -15.002 (should 0.5 -15.002 phrase-tables, 0.5 -15.002 brush 0.5 -15.002 isolation: 0.5 -15.002 92.3% 0.5 -15.002 mmnr−1, 0.5 -15.002 ,σ, 0.5 -15.002 ,l,w 0.5 -15.002 space dn(f) 0.5 -15.002 eye, 0.5 -15.002 (null5) 0.5 -15.002 α(·) 0.5 -15.002 6.85 0.333333 -15.0021 etc.; 0.333333 -15.0035 < .001. 0.25 -15.0042 assistant 0.5 -15.0042 748 0.5 -15.0042 trans3 0.5 -15.0042 badlao” 0.5 -15.0042 forrsooth-based 0.5 -15.0042 springhouse, 0.5 -15.0042 6.5x 0.5 -15.0042 enlisted 0.5 -15.0042 [3, 0.5 -15.0042 “lady 0.5 -15.0042 contrived 0.5 -15.0042 26.74 0.5 -15.0042 1930, 0.5 -15.0042 headword-lexicalization 0.5 -15.0042 (mayfield 0.5 -15.0042 browsed 0.5 -15.0042 localcost(·) 0.5 -15.0042 2.91% 0.5 -15.0042 27.4% 0.5 -15.0042 webps-2 0.5 -15.0042 hughes 0.5 -15.0042 39.57 0.5 -15.0042 nlpr_kbp 0.5 -15.0042 d0– 0.5 -15.0042 johansson, 0.5 -15.0042 69.44 0.5 -15.0042 clarify, 0.5 -15.0042 dnt 0.5 -15.0042 x”) 0.5 -15.0042 93.77 0.5 -15.0042 skeptical 0.5 -15.0042 66.60 0.5 -15.0042 re-ordered, 0.5 -15.0042 (conventionally 0.5 -15.0042 pfl 0.5 -15.0042 (match) 0.5 -15.0042 dimension: 0.5 -15.0042 cross-page 0.5 -15.0042 mwu-based 0.5 -15.0042 α=0 0.5 -15.0042 0.819 0.5 -15.0063 (zarrieß 0.5 -15.0063 198) 0.5 -15.0063 non-loading 0.5 -15.0063 26.79 0.5 -15.0063 fastr, 0.5 -15.0063 32.27 0.5 -15.0063 art: 0.5 -15.0063 humble 0.5 -15.0063 topic-balanced gender 0.5 -15.0063 31.06 0.5 -15.0063 voil´ee 0.5 -15.0063 1079 0.5 -15.0063 sv-lvc, 0.5 -15.0063 maturational 0.5 -15.0063 34,000 0.5 -15.0063 topping 0.5 -15.0063 02455845-a 0.5 -15.0063 “installation” 0.5 -15.0063 software.2 0.5 -15.0063 21m 0.5 -15.0063 redesigned 0.5 -15.0063 lr(θ,α,β) 0.5 -15.0063 q-data, 0.5 -15.0063 wisdomlmde 0.5 -15.0063 (con) 0.5 -15.0063 sub/ins/ok 0.5 -15.0063 kei 0.5 -15.0063 0.12%, 0.5 -15.0063 discrepancies, 0.5 -15.0063 testers, 0.5 -15.0063 “groupie” 0.5 -15.0063 reinforcement, 0.5 -15.0063 rap, 0.5 -15.0063 non-temporal 0.5 -15.0063 “i’d 0.5 -15.0063 1,243 0.5 -15.0063 zile 0.0228188 -15.0064 shown in 0.230769 -15.007 communicator 0.230769 -15.008 segmented translation 0.5 -15.0085 lnullnullnull 0.5 -15.0085 (position 0.5 -15.0085 gps 0.5 -15.0085 sql, 0.5 -15.0085 ist? 0.5 -15.0085 15.55 0.5 -15.0085 42.25 0.5 -15.0085 second-degree 0.5 -15.0085 lazer, 0.5 -15.0085 base-line 0.5 -15.0085 64.60 0.5 -15.0085 emulation 0.5 -15.0085 θt, 0.5 -15.0085 uas) 0.5 -15.0085 norman 0.5 -15.0085 for(l 0.5 -15.0085 0.966 0.5 -15.0085 corpus6 0.5 -15.0085 16.62 0.5 -15.0085 (approximating 0.5 -15.0085 x-axes 0.5 -15.0085 1308 0.5 -15.0085 gap” 0.5 -15.0085 39.2% 0.5 -15.0085 modest, 0.5 -15.0085 d+n 0.5 -15.0085 46.43 0.5 -15.0085 longum 0.5 -15.0085 37.98 0.5 -15.0085 precondition, 0.5 -15.0106 (de) 0.5 -15.0106 swearing 0.5 -15.0106 ceremony 0.5 -15.0106 plitgs 0.5 -15.0106 fantastic 0.5 -15.0106 32.47 0.5 -15.0106 art piece(x2):18 0.5 -15.0106 steers 0.5 -15.0106 releasing, 0.5 -15.0106 misinterpreting 0.5 -15.0106 27.33 0.5 -15.0106 periodic 0.5 -15.0106 (pet) 0.5 -15.0106 (-1, 0.5 -15.0106 epi,...,epj 0.5 -15.0106 mathematics, 0.5 -15.0106 “once 0.5 -15.0106 ˆs1..d1..t 0.5 -15.0106 chi, 0.5 -15.0106 rooted, 0.5 -15.0106 bits: 0.5 -15.0106 0.584 0.5 -15.0106 68.87 0.5 -15.0106 ify 0.5 -15.0106 intersentence 0.5 -15.0106 668 0.5 -15.0106 marginalise 0.5 -15.0106 (1+2) 0.333333 -15.0107 〉, 0.333333 -15.0107 philippines 0.5 -15.0128 ii,r 0.5 -15.0128 revoked, 0.5 -15.0128 azrieli 0.5 -15.0128 recognition) 0.5 -15.0128 dingcheng 0.5 -15.0128 regressions, 0.5 -15.0128 (1986), 0.5 -15.0128 nießen 0.5 -15.0128 (monotone, 0.5 -15.0128 intro 0.5 -15.0128 ya, 0.5 -15.0128 pragmatic-cause 0.5 -15.0128 lshtc 0.5 -15.0128 gini 0.5 -15.0128 biotechnology, 0.5 -15.0128 vvfin 0.5 -15.0128 nakanishi 0.5 -15.0128 susex, 0.5 -15.0128 qjt 0.5 -15.0128 2par, 0.5 -15.0128 velardi, 0.5 -15.0128 +0.88 0.5 -15.0128 ar-en 0.5 -15.0128 bi-, 0.5 -15.0128 1.27 0.5 -15.0128 heavy-np 0.5 -15.0128 kiss 0.5 -15.0128 low- 0.5 -15.0128 ai+ac 0.5 -15.0128 situationally 0.5 -15.0128 one-basic-type-per-discourse 0.5 -15.0128 (schone 0.5 -15.0128 attentions 0.5 -15.0128 reliable) 0.5 -15.0128 ii), 0.5 -15.0128 added/removed 0.285714 -15.0144 ros´e 0.285714 -15.0144 , 0.5 -15.0236 0.377 0.5 -15.0236 (elaborated 0.5 -15.0236 1370 0.5 -15.0236 sake, 0.5 -15.0236 converters 0.5 -15.0236 82.16 0.5 -15.0236 bibliography, 0.5 -15.0236 (unger 0.5 -15.0236 &pmipa,aa 0.5 -15.0236 actin 0.5 -15.0236 jian, 0.5 -15.0236 referents, 0.5 -15.0236 hatwidefx, 0.5 -15.0236 1(a): 0.5 -15.0236 1.3, 0.5 -15.0236 mergers, 0.5 -15.0236 automataand 0.5 -15.0236 record: 0.5 -15.0236 mcgrath 0.5 -15.0236 1613 0.333333 -15.0237 +morph 0.333333 -15.0237 nose 0.285714 -15.0256 51.9 0.5 -15.0258 loking 0.5 -15.0258 larger) 0.5 -15.0258 64.76 0.5 -15.0258 austronesian: 0.5 -15.0258 axes, 0.5 -15.0258 riedel, 0.5 -15.0258 bottom-up, 0.5 -15.0258 vandalize 0.5 -15.0258 worksfor 0.5 -15.0258 farms 0.5 -15.0258 reading-style. 0.5 -15.0258 (xk,yk) 0.5 -15.0258 (w1,w2,n,l,m,r), 0.5 -15.0258 radioactive 0.5 -15.0258 sub/ins/del/ok 0.5 -15.0258 hmmp-basic 0.5 -15.0258 fox, 0.5 -15.0258 t,wh 0.5 -15.0258 [kc10] 0.5 -15.0258 superlative-oes 0.5 -15.0258 kronlid, 0.5 -15.0258 (vocabulary 0.5 -15.0258 0.507 0.5 -15.0258 1.04 0.00756029 -15.027 can 0.2 -15.0271 ann arbor, 0.5 -15.028 37.09 0.5 -15.028 1934 0.5 -15.028 “phrases” 0.5 -15.028 lda-8s 0.5 -15.028 29.4% 0.5 -15.028 [slot2,slot4], 0.5 -15.028 i−dn 0.5 -15.028 “thank 0.5 -15.028 (minoux, 0.5 -15.028 hn, 0.5 -15.028 “3par” 0.5 -15.028 76.13 0.5 -15.028 longest, 0.5 -15.028 english-arabic, 0.5 -15.028 64.2% 0.5 -15.028 database: 0.5 -15.028 aceptable 0.5 -15.028 strategy’s 0.5 -15.028 6(b) 0.5 -15.028 npcs 0.5 -15.028 maslow´s 0.5 -15.028 hulden 0.5 -15.028 gprimeprefix 0.5 -15.028 nowson, 0.5 -15.028 evaluee: 0.5 -15.028 afiliation 0.5 -15.028 pasr(u), 0.5 -15.028 〈σ,n,τ,n0,ω〉 0.5 -15.028 l1) 0.5 -15.028 he” 0.5 -15.028 6.5) 0.5 -15.028 untyped, 0.5 -15.028 accessindia 0.5 -15.028 (mknlm), 0.5 -15.028 nullh, 0.5 -15.028 dtm 0.5 -15.028 (q, 0.285714 -15.0281 donald 0.0789474 -15.0288 course, 0.131579 -15.0291 al., 1998). 0.0588235 -15.0293 rules are 0.285714 -15.0294 achieved comparable 0.333333 -15.0295 “easier” 0.5 -15.0302 demanda 0.5 -15.0302 micheal 0.5 -15.0302 amino acids, 0.5 -15.0302 0.5 -15.0302 93.64 0.5 -15.0302 repeated bisections 0.5 -15.0302 atomic, 0.5 -15.0302 (cluto 0.5 -15.0302 abkani, 0.5 -15.0302 offer”, 0.5 -15.0302 +2.0 0.5 -15.0302 subtypes), 0.5 -15.0302 p/r/f 0.5 -15.0302 crouch, 0.5 -15.0302 reevaluation 0.5 -15.0302 0.150 0.5 -15.0302 ep09 0.5 -15.0302 figure10 0.5 -15.0302 contradictions, 0.5 -15.0302 “model” 0.5 -15.0302 granting 0.5 -15.0302 fully-learnt 0.5 -15.0302 p≤.001; 0.5 -15.0302 nucleus” 0.5 -15.0302 know” 0.12 -15.0307 a trigram 0.333333 -15.031 cikm 0.333333 -15.031 -fb 0.25 -15.0314 maryland college 0.0662651 -15.0318 organized as 0.107692 -15.0324 suspect 0.5 -15.0324 baseline3 0.5 -15.0324 ago” 0.5 -15.0324 design), 0.5 -15.0324 +fb, 0.5 -15.0324 φb 0.5 -15.0324 /le2(x 0.5 -15.0324 27.06 0.5 -15.0324 [ma08] 0.5 -15.0324 pr(t) 0.5 -15.0324 (parallel) 0.5 -15.0324 wmt, 0.5 -15.0324 predicate(s) 0.5 -15.0324 (mechanical 0.5 -15.0324 36.71 0.5 -15.0324 coordination; 0.5 -15.0324 compensatory 0.5 -15.0324 isalias 0.5 -15.0324 foil, 0.5 -15.0324 0.214 0.5 -15.0324 38.28 0.5 -15.0324 [* 0.5 -15.0324 morpheme-based 0.5 -15.0324 {(i,j)} 0.5 -15.0324 graphpartitioning 0.5 -15.0324 argm-vlv 0.5 -15.0324 dependencies) 0.5 -15.0324 pute 0.5 -15.0324 meta-evaluated 0.5 -15.0324 kappa-agreement 0.5 -15.0324 zhao-rul 0.5 -15.0324 ‘path’ 0.5 -15.0324 (zollo, 0.5 -15.0324 rethinking 0.5 -15.0324 žirafe 0.5 -15.0324 durations, 0.5 -15.0324 539 0.5 -15.0324 4.2.5 0.5 -15.0324 funnel 0.5 -15.0324 hz), 0.5 -15.0324 (n,σ,p,s), 0.16 -15.0344 and christopher 0.5 -15.0346 morton 0.5 -15.0346 node(s) 0.5 -15.0346 118–124, 0.5 -15.0346 843 0.5 -15.0346 jeremiah, 0.5 -15.0346 rdn×dn 0.5 -15.0346 raw-text 0.5 -15.0346 open: 0.5 -15.0346 0.780 0.5 -15.0346 gbdt 0.5 -15.0346 accomodates 0.5 -15.0346 bmmm 0.5 -15.0346 score(a→b) 0.5 -15.0346 all,we 0.5 -15.0346 emphasise 0.5 -15.0346 infant-directed 0.5 -15.0346 watrin 0.5 -15.0346 ’i’, 0.5 -15.0346 41.3% 0.5 -15.0346 88.71 0.5 -15.0346 ground: 0.5 -15.0346 upset 0.5 -15.0346 lyons, 0.5 -15.0346 (keselj 0.5 -15.0346 calibrating 0.5 -15.0346 lazyff 0.5 -15.0346 perfomance 0.5 -15.0346 corection, 0.5 -15.0346 5uk 0.5 -15.0346 semantic-role 0.5 -15.0346 smith] 0.5 -15.0346 52.3% 0.5 -15.0346 -omaprox1-ts 0.5 -15.0346 “attacker” 0.5 -15.0346 58.23 0.5 -15.0346 no-arc→ 0.5 -15.0346 bullet, 0.5 -15.0346 0.987 0.333333 -15.0354 85.19 0.091954 -15.0356 describing the 0.5 -15.0368 aplications, 0.5 -15.0368 0.157 0.5 -15.0368 (au,t) 0.5 -15.0368 radiocarbon 0.5 -15.0368 (n-best 0.5 -15.0368 (bias) 0.5 -15.0368 methods) 0.5 -15.0368 pects 0.5 -15.0368 heriberto 0.5 -15.0368 34.05% 0.5 -15.0368 p(e1)p(e2) p(r(e2)l(e1))p(r(e2))p(l(e1)) 0.5 -15.0368 (occurring 0.5 -15.0368 shadowed 0.5 -15.0368 demarcate 0.5 -15.0368 variables), 0.5 -15.0368 subzone, 0.5 -15.0368 0.210 0.5 -15.0368 derogatory 0.5 -15.0368 2centre 0.5 -15.0368 [♣] 0.5 -15.0368 object-component 0.5 -15.0368 common) 0.5 -15.0368 6.6m 0.5 -15.0368 dop, 0.5 -15.0368 keep, 0.5 -15.0368 augmented-filtered 0.5 -15.0368 urn, 0.5 -15.0368 precise) 0.5 -15.0368 lowest) 0.5 -15.0368 +1.40 0.5 -15.0368 0.215 0.5 -15.0368 wsj0020.16 0.5 -15.0368 (t,cf) 0.5 -15.0368 1735 0.5 -15.0368 930 0.5 -15.0368 zip.01 0.5 -15.0368 p@50 0.5 -15.0368 cat) 0.5 -15.0368 patb-tokenized 0.5 -15.0368 at-mentions, 0.5 -15.0368 2p, 0.5 -15.0368 (ldc 0.5 -15.0368 dictionary5 0.5 -15.0368 collating 0.5 -15.0368 staphorsius, 0.181818 -15.0368 will often 0.333333 -15.0383 agent(x, 0.5 -15.039 headverb 0.5 -15.039 ch en 0.5 -15.039 harison, 0.5 -15.039 finest 0.5 -15.039 w-js 0.5 -15.039 (30% 0.5 -15.039 (reduced) 0.5 -15.039 coling/acl, 0.5 -15.039 si,j,w 0.5 -15.039 192, 0.5 -15.039 coindices 0.5 -15.039 snippet, 0.5 -15.039 explosives 0.5 -15.039 digs 0.5 -15.039 unconnectedwords 0.5 -15.039 iwslt'08 0.5 -15.039 (largely 0.5 -15.039 aso, 0.5 -15.039 exclamation, 0.5 -15.039 w(u,v), 0.5 -15.039 mergeable 0.5 -15.039 fritz 0.5 -15.039 “curing 0.5 -15.039 subcomponent 0.5 -15.039 60%) 0.5 -15.039 p(tag) 0.5 -15.039 p=1; 0.5 -15.039 bad: 0.5 -15.039 biometrics 0.5 -15.039 high-pass butterworth 0.5 -15.039 cortese 0.5 -15.039 preverbal 0.5 -15.039 (train), 0.5 -15.039 §02-21 0.2 -15.0395 (artiles 0.0128148 -15.0399 been 0.0945946 -15.0412 reasons. 0.5 -15.0413 spawned 0.5 -15.0413 professor, 0.5 -15.0413 parser9 0.5 -15.0413 ideographic 0.5 -15.0413 hst 0.5 -15.0413 chu-liu-edmonds 0.5 -15.0413 wat 0.5 -15.0413 hale’s 0.5 -15.0413 elaboration: 0.5 -15.0413 mpp, 0.5 -15.0413 curvature 0.5 -15.0413 acl-2010, 0.5 -15.0413 elghamry 0.5 -15.0413 desrt, 0.5 -15.0413 kubat 0.5 -15.0413 pomdp-based 0.5 -15.0413 67.76 0.5 -15.0413 mallows 0.5 -15.0413 stylistically 0.5 -15.0413 dichotomy, 0.5 -15.0413 2a), 0.5 -15.0413 ksmart(x) 0.5 -15.0413 “networking” 0.5 -15.0413 39.60 0.5 -15.0413 v+v 0.5 -15.0413 slap, 0.5 -15.0413 sentiment-detection 0.5 -15.0413 semi-automatically) 0.5 -15.0413 oflazer 0.5 -15.0413 nullnullnull(nullnull) 0.5 -15.0413 equalities 0.5 -15.0413 73.80 0.5 -15.0413 5.1m 0.5 -15.0413 (h1,1,h2,2) 0.5 -15.0413 35.19 0.5 -15.0413 simsem 0.5 -15.0413 0.032 0.5 -15.0413 92.30 0.5 -15.0413 en2de 0.25 -15.0424  meat. 0.142857 -15.0432 who were 0.25 -15.0435 pereira. 0.5 -15.0435 7.79 0.5 -15.0435 network’s 0.5 -15.0435 ferret, 0.5 -15.0435 61.02 0.5 -15.0435 jumper, 0.5 -15.0435 (35 0.5 -15.0435 87.32 0.5 -15.0435 (gonz´alez-brenes 0.5 -15.0435 f(tu,in) 0.5 -15.0435 25.44 0.5 -15.0435 theorized 0.5 -15.0435 manuscript, 0.5 -15.0435 0.438 0.5 -15.0435 labelable” 0.5 -15.0435 thek 0.5 -15.0435 (lsh), 0.5 -15.0435 exactness, 0.5 -15.0435 reversed: 0.5 -15.0435 tolstoy 0.5 -15.0435 whenit 0.5 -15.0435 947 0.5 -15.0435 →+ 0.5 -15.0435 vform 0.5 -15.0435 mention-ranking 0.5 -15.0435 levenstein 0.5 -15.0435 non-redundant, 0.5 -15.0435 probability), 0.5 -15.0435 (ddi) 0.2 -15.0439 prague, 0.333333 -15.0443 “will 0.0115691 -15.0448 is a 0.070922 -15.0453 for example, if 0.5 -15.0457 (xt,yt) 0.5 -15.0457 0* 0.5 -15.0457 multi-resolution 0.5 -15.0457 proaches 0.5 -15.0457 2lm 0.5 -15.0457 xliii 0.5 -15.0457 (threshold 0.5 -15.0457 a} 0.5 -15.0457 86.07 0.5 -15.0457 text-trained 0.5 -15.0457 22.91 0.5 -15.0457 671 0.5 -15.0457 (tp), 0.5 -15.0457 overal, 0.5 -15.0457 psychophysics 0.5 -15.0457 dedicating 0.5 -15.0457 1584 0.5 -15.0457 80.49 0.5 -15.0457 quality”, 0.5 -15.0457 attributional 0.5 -15.0457 56.46 0.5 -15.0457 (acc), 0.5 -15.0457 (schaeffer, 0.5 -15.0457 planers 0.5 -15.0457 max-fac(l) 0.5 -15.0457 mold 0.5 -15.0457 demir 0.5 -15.0457 rainbow 0.5 -15.0457 aded, 0.5 -15.0457 reconcilebased 0.5 -15.0457 tion; 0.5 -15.0457 203; 0.5 -15.0457 (arhar 0.5 -15.0457 13.67 0.5 -15.048 sequent 0.5 -15.048 gruzitis 0.5 -15.048 algorithm3 0.5 -15.048 -dep==t 0.5 -15.048 brooklyn 0.5 -15.048 (64.0%) 0.5 -15.048 57.4%, 0.5 -15.048 sumlm, 0.5 -15.048 (least 0.5 -15.048 time-homogeneous 0.5 -15.048 208; 0.5 -15.048 rough, 0.5 -15.048 underphosphorylated 0.5 -15.048 95.09 0.5 -15.048 joined, 0.5 -15.048 impressively 0.5 -15.048 5b > 0.5 -15.048 czechenglish 0.5 -15.048 (46 0.5 -15.048 citing, 0.5 -15.048 5%, 0.5 -15.048 24.93 0.5 -15.048 “enter” 0.5 -15.048 th, 0.5 -15.048 i(x,y) 0.5 -15.048 is” 0.5 -15.048 svm; 0.5 -15.048 calm 0.5 -15.048 towns 0.5 -15.048 january 12th, 0.5 -15.048 cought 0.5 -15.048 narative 0.5 -15.048 shortestpath 0.5 -15.048 fscore; 0.5 -15.048 (2,267 0.5 -15.048 unpredictive 0.5 -15.048 rargmaxi 0.5 -15.048 ≥0) 0.5 -15.048 (lexicalized) 0.181818 -15.049 paraphrase. 0.117647 -15.05 although there 0.5 -15.0502 girlfriend 0.5 -15.0502 dimensions), 0.5 -15.0502 cures, 0.5 -15.0502 profiting 0.5 -15.0502 neutral kaon 0.5 -15.0502 #5: 0.5 -15.0502 0.4788 0.5 -15.0502 literature-based 0.5 -15.0502 shermis 0.5 -15.0502 (usp) 0.5 -15.0502 tanl 0.5 -15.0502 flu) 0.5 -15.0502 comments) 0.5 -15.0502 8.2, 0.5 -15.0502 verb-second 0.5 -15.0502 uc(s) 0.5 -15.0502 60.65 0.5 -15.0502 36.56 0.5 -15.0502 (e+1 0.5 -15.0502 #iis-0447214, 0.5 -15.0502 (cat, 0.5 -15.0502 morpheme-pos 0.5 -15.0502 djob 0.5 -15.0502 gutiérrez 0.5 -15.0502 factorisations 0.5 -15.0502 45.22 0.5 -15.0502 bj, 0.5 -15.0502 2.1.5 0.5 -15.0502 second-highest 0.5 -15.0502 16.52 0.5 -15.0502 0.772 0.5 -15.0502 precedence, 0.0594059 -15.0505 to take 0.0490196 -15.0505 fact, 0.5 -15.0524 “average” 0.5 -15.0524 huge-count.pl 0.5 -15.0524 cognizer: 0.5 -15.0524 “latent 0.5 -15.0524 retire 0.5 -15.0524 15.52 0.5 -15.0524 in-neighbors, 0.5 -15.0524 hillclimbing 0.5 -15.0524 nullrnullnulll 0.5 -15.0524 88.85 0.5 -15.0524 subject-object 0.5 -15.0524 glyphs 0.5 -15.0524 compacting 0.5 -15.0524 proceding 0.5 -15.0524 hunston 0.5 -15.0524 k′1 0.5 -15.0524 q3: 0.5 -15.0524 829 0.5 -15.0524 combinedmsd 0.5 -15.0524 siting 0.5 -15.0524 infrastructural 0.5 -15.0524 l∪f 0.5 -15.0524 wastes 0.5 -15.0524 initializeθ 0.5 -15.0524 julie 0.5 -15.0524 netanyahu 0.5 -15.0524 µij 0.5 -15.0524 third-year 0.5 -15.0524 (controlling 0.5 -15.0524 cumulated 0.5 -15.0524 39.35 0.5 -15.0524 ctakes 0.5 -15.0524 #oscars” 0.5 -15.0524 corpus,2 0.5 -15.0524 nulls,a,d,fnull 0.0888889 -15.053 with probability 0.5 -15.0547 ˆb 0.5 -15.0547 +fc 0.5 -15.0547 naal 0.5 -15.0547 inspected, 0.5 -15.0547 shanahan, 0.5 -15.0547 killing, 0.5 -15.0547 1.91 0.5 -15.0547 1295 0.5 -15.0547 r′, 0.5 -15.0547 0.996 0.5 -15.0547 37.26 0.5 -15.0547 sentencepair 0.5 -15.0547 textmining 0.5 -15.0547 evicting 0.5 -15.0547 0.002, 0.5 -15.0547 give2, 0.5 -15.0547 •random 0.5 -15.0547 thr 0.5 -15.0547 63.42 0.5 -15.0547 nullkenyan president,lose,electionnull 0.5 -15.0547 sunscreen 0.5 -15.0547 thisallows 0.5 -15.0547 pccg 0.5 -15.0547 “^posphrase 0.5 -15.0547 crossroads 0.5 -15.0547 0.514 0.5 -15.0547 reproduces 0.5 -15.0547 (propbank, 0.5 -15.0547 parser5 0.5 -15.0547 (intelligent 0.5 -15.0547 guest 0.5 -15.0547 0.344. 0.5 -15.0547 [slot2,slot3], 0.5 -15.0547 77.42 0.5 -15.0547 “our” 0.5 -15.0547 91.1% 0.5 -15.0547 superior/subordinate 0.5 -15.0547 54.20 0.285714 -15.0548 yd 0.045584 -15.0554 the authors 0.285714 -15.0561 (yarowsky 0.5 -15.0569 tms 0.5 -15.0569 (tv 0.5 -15.0569 q∗t 0.5 -15.0569 ∆l(s,fi) 0.5 -15.0569 perplexed 0.5 -15.0569 figure 0.5 -15.0569 knob 0.5 -15.0569 talking, 0.5 -15.0569 “russia” 0.5 -15.0569 factors; 0.5 -15.0569 66.15 0.5 -15.0569 synthesis: 0.5 -15.0569 86.77 0.5 -15.0569 0.5 -15.0569 5* 0.5 -15.0569 sct 0.5 -15.0569 s0u 0.5 -15.0569 =, 0.5 -15.0569 enormously 0.5 -15.0569 (biadsy 0.5 -15.0569 ’@ alhkwmyplanckover2pi1 0.5 -15.0569 personality, 0.5 -15.0569 hubbie, 0.5 -15.0569 (coyotl-morales 0.5 -15.0569 47.52 0.5 -15.0569 73.97 0.5 -15.0569 46.69 0.5 -15.0569 declaratively 0.5 -15.0569 sources), 0.5 -15.0569 “burstiness” 0.5 -15.0569 adaptive, 0.5 -15.0569 ‘transgene’ 0.5 -15.0569 connectedparses 0.5 -15.0569 smt: 0.5 -15.0569 52.56 0.0689655 -15.0578 show that our 0.25 -15.0581 (en: 0.333333 -15.0592 6k 0.5 -15.0592 5,418 0.5 -15.0592 hgfcunconstrained 0.5 -15.0592 fables, 0.5 -15.0592 0.0106 0.5 -15.0592 steelers, 0.5 -15.0592 µ∞ 0.5 -15.0592 43.33 0.5 -15.0592 predictions: 0.5 -15.0592 64.68 0.5 -15.0592 effect 0.5 -15.0592 48.07 0.5 -15.0592 harbor 0.5 -15.0592 19.5% 0.5 -15.0592 planer 0.5 -15.0592 56.1% 0.5 -15.0592 817 0.5 -15.0592 90, 0.5 -15.0592 (child) 0.5 -15.0592 (kappa) 0.5 -15.0592 pleasantness, 0.5 -15.0592 testament 0.5 -15.0592 vulgarities 0.5 -15.0592 vp/npl 0.5 -15.0592 24.06 0.5 -15.0592 niv 0.5 -15.0592 duncan, 0.5 -15.0592 58.0% 0.5 -15.0592 1center 0.5 -15.0592 6similar 0.5 -15.0592 edict, 0.5 -15.0592 1471 0.5 -15.0592 s[i] 0.5 -15.0592 (svd). 0.5 -15.0592 2,256 0.5 -15.0592 (citation 0.5 -15.0592 smokers, 0.5 -15.0592 (dep) 0.5 -15.0592 hachey 0.5 -15.0615 51.50 0.5 -15.0615 (kaplan, 0.5 -15.0615 k’th 0.5 -15.0615 nadya=ne 0.5 -15.0615 freeing 0.5 -15.0615 formality), 0.5 -15.0615 914 0.5 -15.0615 transitivity: 0.5 -15.0615 (rus 0.5 -15.0615 precision-oriented 0.5 -15.0615 3.33 0.5 -15.0615 licensor 0.5 -15.0615 5.5%, 0.5 -15.0615 yago’s 0.5 -15.0615 89.95 0.5 -15.0615 45.17 0.5 -15.0615 dataset1, 0.5 -15.0615 adressed 0.5 -15.0615 inparticular, 0.5 -15.0615 kennedy, 0.5 -15.0615 averting 0.5 -15.0615 (<= 0.5 -15.0615 contest 0.5 -15.0615 (pennbioie, 0.5 -15.0615 good; 0.5 -15.0615 0.466 0.5 -15.0615 position-based 0.5 -15.0615 intra-speaker 0.5 -15.0615 ancillary 0.5 -15.0615 überweisung 0.5 -15.0615 “weighted 0.5 -15.0615 russianenglish 0.5 -15.0615 arl 0.5 -15.0637 syllabification 0.5 -15.0637 +conf 0.5 -15.0637 @djpaulyd 0.5 -15.0637 symbol), 0.5 -15.0637 34.13 0.5 -15.0637 0.81, 0.5 -15.0637 lexeme” 0.5 -15.0637 er}, 0.5 -15.0637 non-contiguity 0.5 -15.0637 conv(h[m];h) 0.5 -15.0637 mean’ 0.5 -15.0637 65.69 0.5 -15.0637 75.68 0.5 -15.0637 corpus11 0.5 -15.0637 367) 0.5 -15.0637 bf: 0.5 -15.0637 1b), 0.5 -15.0637 target-context 0.5 -15.0637 (hhmm) 0.5 -15.0637 verbaliser 0.5 -15.0637 cooper 0.5 -15.0637 collections: 0.5 -15.0637 40.34 0.5 -15.0637 (arithmetic) 0.5 -15.0637 thrax: 0.5 -15.0637 percentage, 0.5 -15.0637 “so”, 0.5 -15.0637 0.5 -15.0637 (3.3) 0.5 -15.0637 site” 0.5 -15.0637 ‘grounded’ 0.5 -15.0637 1840 0.5 -15.0637 >50 0.5 -15.0637 countable 0.5 -15.0637 (topic) 0.5 -15.0637 humanly 0.5 -15.0637 skiena, 0.5 -15.0637 “lattice” 0.5 -15.0637 〈i〉s 0.5 -15.0637 medt-1 0.113208 -15.0644 need not 0.25 -15.0649  of 0.5 -15.066 cz-en, 0.5 -15.066 1452 0.5 -15.066 0.162 0.5 -15.066 “context” 0.5 -15.066 ne: 0.5 -15.066 explicitly; 0.5 -15.066 110k 0.5 -15.066 nulli) 0.5 -15.066 ψ(p(pii)) 0.5 -15.066 testified 0.5 -15.066 (eprime,fprime) 0.5 -15.066 0.554 0.5 -15.066 parameterization, 0.5 -15.066 simplifythe 0.5 -15.066 1551 0.5 -15.066 −agrm&l, 0.5 -15.066 vignet: 0.5 -15.066 ‘person’ 0.5 -15.066 (and/or 0.5 -15.066 gerber, 0.5 -15.066 41.03 0.5 -15.066 1541 0.5 -15.066 0.360 0.5 -15.066 dimensions; 0.5 -15.066 foes, 0.5 -15.066 manually-generated 0.5 -15.066 6one 0.5 -15.066 40.60 0.5 -15.066 2010a); 0.5 -15.066 2091 0.5 -15.066 raw-clean 0.5 -15.066 54.1% 0.5 -15.066 would-be 0.5 -15.066 consistent) 0.5 -15.066 collision, 0.5 -15.066 59.63 0.5 -15.066 matching; 0.5 -15.066 interventions, 0.5 -15.066 kn-rnn-all 0.5 -15.066 included” 0.5 -15.066 cals 0.5 -15.066 participancy 0.5 -15.066 rooms, 0.5 -15.066 197k 0.5 -15.066 scope– 0.5 -15.066 m20 0.5 -15.066 56.11 0.5 -15.066 speaker-normalized 0.12 -15.0667 they would 0.333333 -15.0668 w 0.5 -15.0683 augmentedloss 0.5 -15.0683 scs 0.5 -15.0683 (attia 0.5 -15.0683 all-in-one 0.5 -15.0683 beverly 0.5 -15.0683 25.96% 0.5 -15.0683 parliament) 0.5 -15.0683 (1969) 0.5 -15.0683 z-mert, 0.5 -15.0683 discourse/pragmatic 0.5 -15.0683 multi-robot 0.5 -15.0683 g32 0.5 -15.0683 0.660 0.5 -15.0683 steedman’s 0.5 -15.0683 32.75 0.5 -15.0683 beneficial: 0.5 -15.0683 (leftmost 0.5 -15.0683 catchwords 0.5 -15.0683 (ewn) 0.5 -15.0683 freshmen 0.5 -15.0683 markvn 0.5 -15.0683 (expressions 0.5 -15.0683 praised 0.5 -15.0683 ’m, 0.5 -15.0683 57.85 0.5 -15.0683 wordnetaffect: 0.5 -15.0683 real, 0.5 -15.0706 #cl 0.5 -15.0706 boundary: 0.5 -15.0706 characteristically 0.5 -15.0706 differentevent 0.5 -15.0706 36, 0.5 -15.0706 (kokkinakis, 0.5 -15.0706 election” 0.5 -15.0706 affairs, 0.5 -15.0706 4,145 0.5 -15.0706 &ord-described-item 0.5 -15.0706 kira 0.5 -15.0706 waters 0.5 -15.0706 (argument, 0.5 -15.0706 (follow, 0.5 -15.0706 12.34 0.5 -15.0706 v´eronis 0.5 -15.0706 60.1% 0.5 -15.0706 positions) 0.5 -15.0706 successor, 0.5 -15.0706 rom(x,bos)∧f(x) 0.5 -15.0706 yonder 0.5 -15.0706 -.77 0.5 -15.0706 piers 0.5 -15.0706 all-word 0.5 -15.0706 privacy: 0.5 -15.0706 1993b; 0.5 -15.0706 flash 0.5 -15.0706 hacioglu, 0.5 -15.0706 4to 0.5 -15.0706 10,282,536 0.5 -15.0706 bowden 0.5 -15.0706 22.05 0.5 -15.0706 compression: 0.5 -15.0706 1223 0.5 -15.0706 “prec” 0.5 -15.0706 okinawa 0.2 -15.072 sampo 0.333333 -15.0729 nmflib 0.333333 -15.0729 pos-tagger. 0.5 -15.0729 13.84 0.5 -15.0729 71.63 0.5 -15.0729 40.02 0.5 -15.0729 〈other〉 0.5 -15.0729 96.22 0.5 -15.0729 england’s 0.5 -15.0729 sooner 0.5 -15.0729 old) 0.5 -15.0729 (documents, 0.5 -15.0729 renkema, 0.5 -15.0729 shawe-taylor 0.5 -15.0729 example’s 0.5 -15.0729 artifacts, 0.5 -15.0729 de1the 0.5 -15.0729 next; 0.5 -15.0729 overspecialized 0.5 -15.0729 (budgets 0.5 -15.0729 ceremonies 0.5 -15.0729 darling 0.5 -15.0729 nullv,wnull 0.5 -15.0729 (likely) 0.5 -15.0729 name/last 0.5 -15.0729 inevitably, 0.5 -15.0729 demarcated 0.5 -15.0729 0.548 0.5 -15.0729 fugett, 0.5 -15.0729 compositionally, 0.5 -15.0729 (rissanen, 0.5 -15.0729 singularnumber 0.5 -15.0729 6all 0.5 -15.0729 disfavor 0.5 -15.0729 stops, 0.5 -15.0729 content-window-based 0.5 -15.0729 histogram, 0.5 -15.0729 46.40 0.5 -15.0729 (hawking,he) 0.5 -15.0729 paradise-style 0.285714 -15.0743 yuval 0.333333 -15.0744 yale 0.5 -15.0752 england, 0.5 -15.0752 multiplication: 0.5 -15.0752 intensity) 0.5 -15.0752 56.65 0.5 -15.0752 25.25 0.5 -15.0752 0.79% 0.5 -15.0752 ctb2tp) 0.5 -15.0752 (**: 0.5 -15.0752 burke 0.5 -15.0752 9.09 0.5 -15.0752 phase) 0.5 -15.0752 unchanged; 0.5 -15.0752 84.02 0.5 -15.0752 99.60 0.5 -15.0752 arguments), 0.5 -15.0752 0.593 0.5 -15.0752 t2score 0.5 -15.0752 hwst 0.5 -15.0752 genus, 0.5 -15.0752 15.43 0.5 -15.0752 (english), 0.5 -15.0752 62.49 0.5 -15.0752 57.54 0.5 -15.0752 83.57 0.5 -15.0752 newsletter, 0.5 -15.0752 ect 0.5 -15.0752 human-edited 0.5 -15.0752 setting-2 0.5 -15.0752 87.10 0.5 -15.0752 a.01 0.5 -15.0752 1049 0.5 -15.0752 distance; 0.5 -15.0752 danulla 0.5 -15.0752 individuals’ 0.5 -15.0752 elide 0.5 -15.0752 (mit) 0.5 -15.0752 67.44 0.285714 -15.0756 2,3 0.285714 -15.0769 φik 0.5 -15.0775 unreliable, 0.5 -15.0775 (sub 0.5 -15.0775 locally-weighted 0.5 -15.0775 trough 0.5 -15.0775 4.87 0.5 -15.0775 519,359 0.5 -15.0775 synt, 0.5 -15.0775 lowercased; 0.5 -15.0775 39.46 0.5 -15.0775 et.al., 0.5 -15.0775 50.66 0.5 -15.0775 92.36 0.5 -15.0775 &: 0.5 -15.0775 28.42 0.5 -15.0775 undermining 0.5 -15.0775 lags, 0.5 -15.0775 τη 0.5 -15.0775 lattice-parsing 0.5 -15.0775 (ies), 0.5 -15.0775 (mountain 0.5 -15.0775 1445 0.5 -15.0775 l0-weight 0.5 -15.0775 wordsflags 0.5 -15.0775 communicated, 0.5 -15.0775 (2011b), 0.5 -15.0775 judging, 0.5 -15.0775 65.07 0.131579 -15.0776 (buchholz 0.333333 -15.079 agreement/disagreement 0.333333 -15.079 koc-combo 0.333333 -15.079 maite 0.5 -15.0798 open-source1 0.5 -15.0798 eparl+nc+1092 0.5 -15.0798 text-derived 0.5 -15.0798 dws, 0.5 -15.0798 spatial, 0.5 -15.0798 22.21 0.5 -15.0798 “bagof-words” 0.5 -15.0798 43.95 0.5 -15.0798 left-child 0.5 -15.0798 nls 0.5 -15.0798 (schoenemann, 0.5 -15.0798 swdp 0.5 -15.0798 1982, 0.5 -15.0798 bestout(c) 0.5 -15.0798 wiki-trained 0.5 -15.0798 (dada, 0.5 -15.0798 yoshida, 0.5 -15.0798 comand-line 0.5 -15.0798 bilinear, 0.5 -15.0798 approval, 0.5 -15.0798 away) 0.5 -15.0798 1using 0.5 -15.0798 46.7% 0.5 -15.0798 (agresti, 0.5 -15.0798 oil, 0.5 -15.0798 88.39 0.5 -15.0798 51.30 0.5 -15.0798 vectorφ 0.5 -15.0798 nullnullnullnullnullnullnullnullpnullvit(g)nullnullnull 0.5 -15.0798 potentials, 0.5 -15.0798 chance-corrected 0.5 -15.0798 “nonrelevant” 0.5 -15.0798 per-template 0.5 -15.0798 p(ci) 0.5 -15.0798 montague, 0.5 -15.0798 bionlp/nlpba 0.5 -15.0798 adjoints 0.0419753 -15.0806 leads 0.0985915 -15.0816 of them. 0.5 -15.0821 dangling 0.5 -15.0821 indexi 0.5 -15.0821 lower-scoring 0.5 -15.0821 ξ(very 0.5 -15.0821 rudolf 0.5 -15.0821 2-gram, 0.5 -15.0821 94.50 0.5 -15.0821 28.76 0.5 -15.0821 top-down/bottom-up 0.5 -15.0821 tc-based 0.5 -15.0821 thomson-reuters 0.5 -15.0821 albanian 0.5 -15.0821 sucre) 0.5 -15.0821 pass(pink1,pink11) 0.5 -15.0821 attendance 0.5 -15.0821 (okazaki, 0.5 -15.0821 reverb¬lex, 0.5 -15.0821 tut 0.5 -15.0821 context-theoretic 0.5 -15.0821 antonymy, 0.5 -15.0821 angles, 0.5 -15.0821 (10k-1m 0.5 -15.0821 px(e,o), 0.5 -15.0821 rising-trend 0.5 -15.0821 form2: 0.5 -15.0821 162m 0.5 -15.0821 (labels 0.5 -15.0821 library2 0.5 -15.0821 wars, 0.5 -15.0821 23.70 0.5 -15.0821 lm-fst, 0.5 -15.0821 wicentowski, 0.5 -15.0821 saiyam 0.5 -15.0821 agrement, 0.5 -15.0821 ‗useful‘ 0.5 -15.0821 en, 0.5 -15.0821 fa8750-09-c-0172 0.5 -15.0821 lrscore: 0.5 -15.0821 (xi,xj)∈y 0.5 -15.0821 669 0.5 -15.0821 (klakow, 0.1 -15.0835 providing a 0.5 -15.0844 4.99 0.5 -15.0844 somers, 0.5 -15.0844 np{answer}” 0.5 -15.0844 num:date 0.5 -15.0844 worried 0.5 -15.0844 tried: 0.5 -15.0844 losses, 0.5 -15.0844 φ; 0.5 -15.0844 creationof 0.5 -15.0844 go:0006810 0.5 -15.0844 (p-a) 0.5 -15.0844 precursors 0.5 -15.0844 d.c., 0.5 -15.0844 ςšqy4 0.5 -15.0844 steroids 0.5 -15.0844 crfsuite 0.5 -15.0844 (syn) 0.5 -15.0844 vectorψ 0.5 -15.0844 54.40 0.5 -15.0844 284–291, 0.5 -15.0844 inter-evaluator 0.5 -15.0844 21.20 0.5 -15.0844 word-latticebased 0.5 -15.0844 position,‖ 0.5 -15.0844 ors 0.5 -15.0844 standardmentionboundariesaregiven. 0.5 -15.0844 legalization 0.5 -15.0844 lenzi 0.5 -15.0844 imageability, 0.5 -15.0844 strack 0.5 -15.0844 78.57 0.5 -15.0844 0.508 0.5 -15.0844 (niederhoffer 0.5 -15.0844 (seconds 0.5 -15.0844 whethergold 0.5 -15.0844 realize, 0.5 -15.0844 dictionary-matching 0.5 -15.0844 collinsthompson 0.5 -15.0844 fiveyear 0.5 -15.0844 auto-punc 0.5 -15.0844 72.3% 0.5 -15.0844 5.6k 0.5 -15.0844 cancel, 0.5 -15.0844 39.95 0.5 -15.0844 edge’s endpoints. 0.5 -15.0844 32.92 0.5 -15.0844 opera 0.5 -15.0844 heiner 0.5 -15.0844 invalid, 0.5 -15.0844 (sauri 0.5 -15.0844 flowing 0.285714 -15.0848 neq 0.285714 -15.0848 non-mwe 0.333333 -15.0852 torture 0.2 -15.0867 be inferred. 0.5 -15.0867 hat, 0.5 -15.0867 (91%) 0.5 -15.0867 thahir 0.5 -15.0867 (o), 0.5 -15.0867 32.000 0.5 -15.0867 unrepresentative 0.5 -15.0867 27.26 0.5 -15.0867 6.4%, 0.5 -15.0867 circumference 0.5 -15.0867 modulate 0.5 -15.0867 91.2% 0.5 -15.0867 72.76 0.5 -15.0867  0.5 -15.0867 thereare 0.5 -15.0867 cips-sighan 0.5 -15.0867 strategy; 0.5 -15.0867 adjectivication 0.5 -15.0867 allocating 0.5 -15.0867 0.096 0.5 -15.0867 ilab 0.5 -15.0867 16.10 0.5 -15.0867 coaches 0.5 -15.0867 communist, 0.5 -15.0867 87.93 0.5 -15.0867 +0.66 0.5 -15.0867 61.65 0.5 -15.0867 reservations, 0.5 -15.0867 carmel 0.5 -15.0867 sofia 0.5 -15.0867 10.78 0.5 -15.0867 meyerhof, 0.5 -15.0867 japanese), 0.5 -15.0867 75, 0.5 -15.0867 sammons, 0.5 -15.0867 (alone) 0.5 -15.0867 “open 0.25 -15.0868 chapman, 0.333333 -15.0868 sn, 0.5 -15.0891 2and 0.5 -15.0891 39.49 0.5 -15.0891 87.78 0.5 -15.0891 reestimating 0.5 -15.0891 (marry 0.5 -15.0891 the), 0.5 -15.0891 university) 0.5 -15.0891 φ(q,d) 0.5 -15.0891 datetime 0.5 -15.0891 42.2% 0.5 -15.0891 3.2.2, 0.5 -15.0891 weekdays 0.5 -15.0891 2.89 0.5 -15.0891 ‘precedesequal’ 0.5 -15.0891 41.10 0.5 -15.0891 ologreg, 0.5 -15.0891 word-association 0.5 -15.0891 rating) 0.5 -15.0891 ∆stk 0.5 -15.0891 varies: 0.5 -15.0891 re-normalized 0.5 -15.0891 91.10 0.5 -15.0891 rameterized 0.5 -15.0891 bundle 0.5 -15.0891 41.6% 0.5 -15.0891 55.31 0.5 -15.0891 leave-one-occupation-out 0.5 -15.0891 '-*./"0 0.5 -15.0891 fields) 0.5 -15.0891 86.90 0.5 -15.0891 scotland 0.5 -15.0891 negatives; 0.5 -15.0891 rnn-lm 0.5 -15.0891 (namely: 0.5 -15.0891 40.55 0.5 -15.0891 nebraska 0.333333 -15.0899 c1000 0.333333 -15.0914 beforehand 0.333333 -15.0914 l2, 0.5 -15.0914 67.03 0.5 -15.0914 55.77 0.5 -15.0914 closure) 0.5 -15.0914 klotz 0.5 -15.0914 toefl® 0.5 -15.0914 “truth” 0.5 -15.0914 sector, 0.5 -15.0914 (brent 0.5 -15.0914 beckman 0.5 -15.0914 iis-1018314 0.5 -15.0914 kanayama, 0.5 -15.0914 1005 0.5 -15.0914 92.32 0.5 -15.0914 abbreviatory 0.5 -15.0914 x1...xn 0.5 -15.0914 phone-based 0.5 -15.0914 br: 0.5 -15.0914 spyware 0.5 -15.0914 64.59 0.5 -15.0914 3.12 0.5 -15.0914 torpedoes 0.5 -15.0914 expletive, 0.5 -15.0914 793 0.5 -15.0914 46.20 0.5 -15.0914 9.75 0.5 -15.0914 weka, 0.5 -15.0914 28.68 0.5 -15.0914 kiran gvr 0.5 -15.0914 nhgri 0.5 -15.0914 ponent 0.2 -15.0923 gim´enez 0.142857 -15.0934 bicycles 0.5 -15.0937 an−1i−1 0.5 -15.0937 j∈i+ 0.5 -15.0937 variances, 0.5 -15.0937 72.28 0.5 -15.0937 ˆphybr 0.5 -15.0937 (f: 0.5 -15.0937 19.90 0.5 -15.0937 succesfull 0.5 -15.0937 desabamento[landslide] 0.5 -15.0937 expected agreementby 0.5 -15.0937 (links 0.5 -15.0937 revisioniterator 0.5 -15.0937 (aer 0.5 -15.0937 24,319 0.5 -15.0937 36.50 0.5 -15.0937 fc 0.5 -15.0937 required; 0.5 -15.0937 ambiguities) 0.5 -15.0937 broder’s 0.5 -15.0937 72.9% 0.5 -15.0937 wilkes-gibbs, 0.5 -15.0937 30.55 0.5 -15.0937 “italian” 0.5 -15.0937 90.41 0.5 -15.0937 (10, 0.5 -15.0937 (1984) 0.5 -15.0937 mert: 0.5 -15.0937 oovs; 0.5 -15.0937 non-decomposable 0.5 -15.0937 89.36 0.5 -15.0937 taln 0.5 -15.0937 mated 0.5 -15.0937 set): 0.5 -15.0937 (rules) 0.5 -15.0937 e−βh 0.5 -15.0937 30.47 0.5 -15.0937 flew 0.5 -15.0937 ut+1 0.5 -15.0937 39.92 0.5 -15.0937 9,250 0.230769 -15.0939 option, 0.25 -15.0949 [ilp] 0.214286 -15.0961 (daelemans 0.5 -15.0961 31.76 0.5 -15.0961 23.28 0.5 -15.0961 “visual 0.5 -15.0961 (swan 0.5 -15.0961 53.40 0.5 -15.0961 35.20 0.5 -15.0961 experts), 0.5 -15.0961 3.6x 0.5 -15.0961 technical, 0.5 -15.0961 teeth 0.5 -15.0961 2007)(model 0.5 -15.0961 dataset), 0.5 -15.0961 26.62 0.5 -15.0961 56.03 0.5 -15.0961 fluid, 0.5 -15.0961 “identity 0.5 -15.0961 prakken 0.5 -15.0961 in–depth 0.5 -15.0961 e-2 0.5 -15.0961 (cocke 0.5 -15.0961 provide, 0.5 -15.0961 ergun 0.5 -15.0961 reimplementations 0.5 -15.0961 highlighted; 0.5 -15.0961 85.17 0.5 -15.0961 -0.9 0.5 -15.0961 07.8 0.5 -15.0961 screened 0.5 -15.0961 duh, 0.5 -15.0961 ternary, 0.5 -15.0961 noteworthy, 0.333333 -15.0977 sprime1 0.5 -15.0984 +wnφn(k,m,lscript) 0.5 -15.0984 0.108 0.5 -15.0984 non-unary 0.5 -15.0984 gamon (2004) 0.5 -15.0984 bye, 0.5 -15.0984 sectionwe 0.5 -15.0984 (49) 0.5 -15.0984 (longer 0.5 -15.0984 tothe 0.5 -15.0984 2,601 0.5 -15.0984 pθ 0.5 -15.0984 87.80 0.5 -15.0984 47.92 0.5 -15.0984 k7p 0.5 -15.0984 matrixw 0.5 -15.0984 6results 0.5 -15.0984 inexpensive, 0.5 -15.0984 (fossum 0.5 -15.0984 (decision 0.5 -15.0984 (begin 0.5 -15.0984 dependable 0.5 -15.0984 50.22 0.5 -15.0984 upward-leading 0.5 -15.0984 ordinarily, 0.5 -15.0984 oracle’s 0.5 -15.0984 87.01 0.5 -15.0984 86.65 0.5 -15.0984 text—a 0.5 -15.0984 “announced 0.5 -15.0984 73.58 0.5 -15.0984 8010 0.5 -15.0984 56.12 0.5 -15.0984 ˚ι 0.5 -15.0984 jo˜ao 0.5 -15.0984 (ponte 0.5 -15.0984 alanfjar 0.5 -15.0984 (98 0.230769 -15.1004 things, 0.5 -15.1008 void, 0.5 -15.1008 (uniform 0.5 -15.1008 query-oriented fashion. 0.5 -15.1008 operate, 0.5 -15.1008 “restart 0.5 -15.1008 fossum 0.5 -15.1008 23-27 0.5 -15.1008 calcuated 0.5 -15.1008 non-regular, 0.5 -15.1008 (uk) 0.5 -15.1008 corpus1, 0.5 -15.1008 1950’s 0.5 -15.1008 test2010 0.5 -15.1008 dimension) 0.5 -15.1008 0.074 0.5 -15.1008 metrics) 0.5 -15.1008 ddk=1 0.5 -15.1008 23.43 0.5 -15.1008 when) 0.5 -15.1008 0.967 0.5 -15.1008 fourclass 0.5 -15.1008 site-specific 0.5 -15.1008 (lópez-cózar 0.5 -15.1008 61.55% 0.5 -15.1008 −tagdict 0.5 -15.1008 (e1,f2) 0.5 -15.1008 mentionedin 0.5 -15.1008 spotlight 0.5 -15.1008 kiat 0.5 -15.1008 in-order 0.5 -15.1008 laird, 0.5 -15.1008 (shinzato 0.5 -15.1008 19.57 0.5 -15.1008 16.99 0.5 -15.1008 ysi−1 0.5 -15.1008 nullnull(nullnullnullnullnull) 0.5 -15.1008 1-ter 0.5 -15.1008 (poster 0.5 -15.1008 glpk 0.5 -15.1008 f1-scores, 0.5 -15.1008 nintendo 0.333333 -15.1024 input’s 0.333333 -15.1024 (sect. 0.5 -15.1031 userusage, 0.5 -15.1031 genotype 0.5 -15.1031 diabetes, 0.5 -15.1031 detector; 0.5 -15.1031 (a-te, 0.5 -15.1031 3although 0.5 -15.1031 “primary 0.5 -15.1031 hoag 0.5 -15.1031 pipe 0.5 -15.1031 studying, 0.5 -15.1031 adjudicator, 0.5 -15.1031 factual, 0.5 -15.1031 (mc 0.5 -15.1031 step2: 0.5 -15.1031 by) 0.5 -15.1031 (gooi 0.5 -15.1031 ablations: 0.5 -15.1031 w+s(m) 0.5 -15.1031 pi), 0.5 -15.1031 811 0.5 -15.1031 posafter 0.5 -15.1031 confound 0.5 -15.1031 helpful: 0.5 -15.1031 rosenberg 0.5 -15.1031 56.2 28.1 0.1875 -15.1032 saved 0.25 -15.1044 loud 0.333333 -15.1055 verkehrszeichen 0.5 -15.1055 dar-se 0.5 -15.1055 author) 0.5 -15.1055 argument), 0.5 -15.1055 fleiss 0.5 -15.1055 laly. 0.5 -15.1055 92.60 0.5 -15.1055 hsieh, 0.5 -15.1055 (1995); 0.5 -15.1055 top-left 0.5 -15.1055 61–69, 0.5 -15.1055 61.23 0.5 -15.1055 1958; 0.5 -15.1055 feruci 0.5 -15.1055 (48 0.5 -15.1055 29.29 0.5 -15.1055 938 0.5 -15.1055 mt: 0.5 -15.1055 arguments’ 0.5 -15.1055 sussna 0.5 -15.1055 complication, 0.5 -15.1055 ko, 0.5 -15.1055 1430 0.5 -15.1055 riesa 0.5 -15.1055 44.27 0.5 -15.1055 dspathform 0.5 -15.1055 700k 0.5 -15.1055 survival-based 0.5 -15.1055 86.17 0.5 -15.1055 lex-mi 0.5 -15.1055 83.40 0.5 -15.1055 dataoriented 0.5 -15.1055 3.16 0.5 -15.1055 (d, 0.5 -15.1055 37.15 0.5 -15.1055 58.8% 0.5 -15.1055 object(s) 0.285714 -15.1063 〈w1,wprime1〉 0.25 -15.1067 (e 0.25 -15.1067 by establishing 0.333333 -15.1071 moran, 0.0842105 -15.1078 sequence. 0.25 -15.1079 shrinkage 0.5 -15.1079 adj2 0.5 -15.1079 tl, 0.5 -15.1079 -omaprox2-ts 0.5 -15.1079 x→x 0.5 -15.1079 absolutives 0.5 -15.1079 (yates, 0.5 -15.1079 17.69 0.5 -15.1079 ted.briscoe@cl.cam.ac.uk 0.5 -15.1079 (minority 0.5 -15.1079 ‘travel’ 0.5 -15.1079 article‟s 0.5 -15.1079 (barhillel, 0.5 -15.1079 ation 0.5 -15.1079 (90 0.5 -15.1079 25.38 0.5 -15.1079 4.79 0.5 -15.1079 (unobserved) 0.5 -15.1079 clasifies 0.5 -15.1079 readingstyle 0.5 -15.1079 (was 0.5 -15.1079 3.35 0.5 -15.1079 n/c 0.5 -15.1079 “什恶魔” 0.5 -15.1079 4782 0.5 -15.1079 (bar 0.5 -15.1079 73.22 0.5 -15.1079 approves 0.5 -15.1079 insufficient, 0.5 -15.1079 (dienes 0.5 -15.1079 [1992] 0.5 -15.1079 crept 0.5 -15.1079 (…) 0.5 -15.1079 (string) 0.5 -15.1079 oand 0.5 -15.1079 insist, 0.5 -15.1079 dudes 0.5 -15.1079 wehrli 0.5 -15.1079 “closing” 0.5 -15.1079 species: 0.5 -15.1079 (p,q)〉. 0.5 -15.1079 fk(nullwnull) 0.5 -15.1079 v6.0 0.5 -15.1079 connoting 0.5 -15.1103 nas 0.5 -15.1103 dewey 0.5 -15.1103 ˆckl 0.5 -15.1103 1970); 0.5 -15.1103 contenders 0.5 -15.1103 likelihood; 0.5 -15.1103 31.21 0.5 -15.1103 equiv 0.5 -15.1103 brown-token-r 0.5 -15.1103 “attribute” 0.5 -15.1103 ica 0.5 -15.1103 resolve, 0.5 -15.1103 r(x,z) 0.5 -15.1103 affect: 0.5 -15.1103 24.42 0.5 -15.1103 q(st,at,d) 0.5 -15.1103 scrambled, 0.5 -15.1103 t( 0.5 -15.1103 webcam 0.5 -15.1103 177.1 0.5 -15.1103 link’s 0.5 -15.1103 guideline, 0.5 -15.1103 k),which 0.5 -15.1103 19.28 0.5 -15.1103 reviews; 0.5 -15.1103 anda 0.5 -15.1103 qprime 0.5 -15.1103 lbp 0.5 -15.1103 discriminating; 0.5 -15.1103 in-neighbors 0.5 -15.1126 rte3, 0.5 -15.1126 tering 0.5 -15.1126 lt3, 0.5 -15.1126 retrospect, 0.5 -15.1126 auth(s) 0.5 -15.1126 94.75 0.5 -15.1126 apart” 0.5 -15.1126 “borrow” 0.5 -15.1126 51.09 0.5 -15.1126 (sparck 0.5 -15.1126 crf), 0.5 -15.1126 wk-4f+wn-4f, 0.5 -15.1126 64.0% 0.5 -15.1126 rhyming: 0.5 -15.1126 synonymy) 0.5 -15.1126 (kidnap, 0.5 -15.1126 oov), 0.5 -15.1126 21.53 0.5 -15.1126 file), 0.5 -15.1126 40.88 0.5 -15.1126 foil 0.5 -15.1126 make2, 0.5 -15.1126 0.184 0.5 -15.1126 93.76 0.5 -15.1126 (tsai 0.5 -15.1126 j4.8 0.5 -15.1126 32.21 0.5 -15.1126 maturities 0.5 -15.1126 iteration: 0.5 -15.1126 accurate; 0.5 -15.1126 short-list 0.5 -15.1126 11.82 0.5 -15.1126 data1 0.5 -15.1126 engine: 0.5 -15.1126 nouveau-rouge: 0.5 -15.1126 (wallach, 0.5 -15.1126 much”, 0.5 -15.1126 n-gram/plsa, 0.5 -15.1126 (sur) 0.5 -15.1126 delegated 0.5 -15.1126 (gst), 0.5 -15.1126 i% 0.5 -15.1126 keyboards 0.0454545 -15.1139 document. 0.2 -15.1141 “give 0.285714 -15.1144 dn(f) 0.5 -15.115 532 0.5 -15.115 cf). 0.5 -15.115 w1,...,wm 0.5 -15.115 medpost 0.5 -15.115 colombia 0.5 -15.115 forward/backward 0.5 -15.115 non-optimal 0.5 -15.115 39.18 0.5 -15.115 happy, 0.5 -15.115 (litman, 0.5 -15.115 harvard, 0.5 -15.115 here”mention 0.5 -15.115 bilou 0.5 -15.115 26.43 0.5 -15.115 (theune 0.5 -15.115 q’s 0.5 -15.115 55.41 0.5 -15.115 meta-learning 0.5 -15.115 mentiondefinitionwe 0.5 -15.115 cooper, 0.5 -15.115 synonyms; 0.5 -15.115 alignmentmoves 0.5 -15.115 46.65 0.5 -15.115 4some 0.5 -15.115 rows), 0.5 -15.115 interlingua, 0.5 -15.115 carl 0.5 -15.115 theory; 0.5 -15.115 8.39 0.5 -15.115 λp 0.5 -15.115 collaborators, 0.5 -15.115 prp+prp1 0.5 -15.115 shorthands, 0.5 -15.115 systemspecific 0.5 -15.115 sub-networks 0.5 -15.115 +1.0 0.5 -15.115 favourably 0.5 -15.115 _-hashtags 0.5 -15.115 σ∈s 0.5 -15.115 thereon 0.5 -15.115 ‘salim 0.5 -15.115 virgin 0.176471 -15.1154 stereotype 0.230769 -15.1158 on). 0.00370657 -15.116 we 0.5 -15.1174 surnames, 0.5 -15.1174 schlesinger 0.5 -15.1174 wiki100l, 0.5 -15.1174 (19 0.5 -15.1174 505k 0.5 -15.1174 originaly 0.5 -15.1174 degradation, 0.5 -15.1174 pool: 0.5 -15.1174 mod" 0.5 -15.1174 yin-wen 0.5 -15.1174 two-layer 0.5 -15.1174 98.76 0.5 -15.1174 p(t1,t2) 0.5 -15.1174 renin 0.5 -15.1174 data-sparsity 0.5 -15.1174 wmt’s 0.5 -15.1174 37.55 0.5 -15.1174 (0.5), 0.5 -15.1174 0.105 0.5 -15.1174 26.42 0.5 -15.1174 (disaster 0.5 -15.1174 disciplinary 0.5 -15.1174 37.27 0.5 -15.1174 assignment’s 0.5 -15.1174 68.38 0.5 -15.1174 tsl, 0.5 -15.1174 (cutler 0.5 -15.1174 58.75 0.5 -15.1174 html) 0.5 -15.1174 marrafa 0.25 -15.1175 disability 0.25 -15.1187 vlachos 0.5 -15.1198 1573 0.5 -15.1198 61.29 0.5 -15.1198 6,774 0.5 -15.1198 2005a), 0.5 -15.1198 w−1w0w1 0.5 -15.1198 ahead, 0.5 -15.1198 1607 0.5 -15.1198 acl’08 0.5 -15.1198 (dup) 0.5 -15.1198 almost, 0.5 -15.1198 1812 0.5 -15.1198 62.77 0.5 -15.1198 dalgleish 0.5 -15.1198 4.14 0.5 -15.1198 79.26 0.5 -15.1198 22.60 0.5 -15.1198 w+s(i) 0.5 -15.1198 67.70 0.5 -15.1198 (rgb) 0.5 -15.1198 (temporal) 0.5 -15.1198 api: 0.5 -15.1198 (into 0.5 -15.1198 spice 0.5 -15.1198 “eparl+nc+crawled2” 0.5 -15.1198 asher, 0.5 -15.1198 post-experiment 0.5 -15.1198 eparl+nc+1092+news+ir 0.5 -15.1198 (21 0.5 -15.1198 74.15 0.5 -15.1198 969 0.5 -15.1198 23.84 0.5 -15.1198 use1 0.5 -15.1198 off1 0.5 -15.1198 s0l, 0.5 -15.1198 4.27 0.333333 -15.1199 sense-disambiguated 0.0322581 -15.1202 workshop 0.230769 -15.1202 ˇzabokrtsk´y, 0.0921053 -15.1209 (marcus et 0.230769 -15.1213 deviation. 0.230769 -15.1213 publishers. 0.5 -15.1222 mmci 0.5 -15.1222 two-month 0.5 -15.1222 bioinfer, 0.5 -15.1222 broken, 0.5 -15.1222 “say 0.5 -15.1222 3v 0.5 -15.1222 extraversion 0.5 -15.1222 6-tuple 0.5 -15.1222 92.41 0.5 -15.1222 (optimizing 0.5 -15.1222 6.27 0.5 -15.1222 energetic 0.5 -15.1222 hayashi 0.5 -15.1222 sentiment), 0.5 -15.1222 classinstance 0.5 -15.1222 degradation: 0.5 -15.1222 x!y 0.5 -15.1222 sys15 0.5 -15.1222 n(w,d) 0.5 -15.1222 network” 0.5 -15.1222 samba 0.5 -15.1222 lexicographer, 0.5 -15.1222 household 0.5 -15.1222 “reasonable” 0.5 -15.1222 period) 0.5 -15.1222 1.4%, 0.5 -15.1222 29.41 0.5 -15.1222 relieved 0.5 -15.1222 adapted) 0.5 -15.1222 kelly, 0.5 -15.1222 matsoukas, 0.5 -15.1222 nedict 0.5 -15.1222 smh, 0.5 -15.1222 16-20 0.5 -15.1222 mst-1 0.285714 -15.1227 (vincze 0.0566038 -15.1236 order. 0.5 -15.1246 groundedlearningtasks 0.5 -15.1246 model.1 0.5 -15.1246 i(b) 0.5 -15.1246 bio, 0.5 -15.1246 81.88 0.5 -15.1246 soviet, 0.5 -15.1246 setting) 0.5 -15.1246 63.35 0.5 -15.1246 £@ 0.5 -15.1246 modular: 0.5 -15.1246 1457 0.5 -15.1246 its, 0.5 -15.1246 polaritybased 0.5 -15.1246 grammix 0.5 -15.1246 individually), 0.5 -15.1246 gmf2(y) 0.5 -15.1246 ngs 0.5 -15.1246 17.29% 0.5 -15.1246 erkan 0.5 -15.1246 thenth 0.5 -15.1246 fod, 0.5 -15.1246 40.42 0.5 -15.1246 (“apple”, company) 0.5 -15.1246 singularities 0.5 -15.1246 207) 0.5 -15.1246 47.51 0.5 -15.1246 (collective 0.5 -15.1246 (7, 0.5 -15.1246 speech-related 0.5 -15.1246 virulence 0.5 -15.1246 (l, 0.5 -15.1246 +null+aalign 0.5 -15.1246 interns 0.0337838 -15.1257 2008; 0.5 -15.1271 eyetracking 0.5 -15.1271 lease, 0.5 -15.1271 63.5% 0.5 -15.1271 cabezas 0.5 -15.1271 chomsky, 0.5 -15.1271 width: 0.5 -15.1271 affective, 0.5 -15.1271 (kukich, 0.5 -15.1271 smiled, 0.5 -15.1271 embodiments 0.5 -15.1271 auto-parsed 0.5 -15.1271 sullivan 0.5 -15.1271 (coordination 0.5 -15.1271 sine 0.5 -15.1271 (1999, 0.5 -15.1271 eat), 0.5 -15.1271 φt(z,e,a) 0.5 -15.1271 50.69 0.5 -15.1271 (1959) 0.5 -15.1271 left  0.5 -15.1271 copulas, 0.5 -15.1271 73.38 0.5 -15.1271 swearing, 0.5 -15.1271 predicates; 0.5 -15.1271 (+0.04 0.5 -15.1271 #nodes 0.5 -15.1271 brill’s 0.5 -15.1271 43.51 0.5 -15.1271 t+ 0.5 -15.1271 91.44 0.5 -15.1271 rity 0.5 -15.1271 hotel’s 0.5 -15.1271 0.258 0.5 -15.1271 pivot, 0.5 -15.1271 65.82 0.5 -15.1271 68.50 0.5 -15.1271 tensed 0.5 -15.1271 “technologies” 0.5 -15.1271 #pi# 0.5 -15.1271 “parsing” 0.5 -15.1271 long-time 0.5 -15.1271 vowel, 0.5 -15.1271 prevhypo 0.5 -15.1271 indiscriminately, 0.5 -15.1271 (icon), 0.5 -15.1271 differs, 0.5 -15.1271 sympathy 0.5 -15.1271 carbon-dating 0.5 -15.1271 topic-model 0.5 -15.1271 h), 0.0983607 -15.1278 contrary 0.285714 -15.1282 windowdiff 0.25 -15.1295 entailment, emnlp 0.333333 -15.1295 popularized 0.5 -15.1295 three-class 0.5 -15.1295 dagger 0.5 -15.1295 fertilities, 0.5 -15.1295 subrequest” 0.5 -15.1295 (http://www.im2.ch). 0.5 -15.1295 3,000,000 0.5 -15.1295 15176613 0.5 -15.1295 da-english 0.5 -15.1295 (kanerva, 0.5 -15.1295 90.57 0.5 -15.1295 events’ 0.5 -15.1295 (krishnan 0.5 -15.1295 the” 0.5 -15.1295 (roughly, 0.5 -15.1295 1169 0.5 -15.1295 eparl+nc+1092, 0.5 -15.1295 (kern 0.5 -15.1295 47.7% 0.5 -15.1295 definitions) 0.5 -15.1295 (totaling 0.5 -15.1295 dynavox 0.5 -15.1295 assessor-grounding 0.5 -15.1295 iteratively: 0.5 -15.1295 0.943 0.5 -15.1295 pedigrees 0.5 -15.1295 agrees, 0.5 -15.1295 meter 0.5 -15.1295 nigam 0.5 -15.1295 15179801 0.5 -15.1295 (justeson 0.5 -15.1295 (mulkar 0.5 -15.1295 tgrep2 0.5 -15.1295 cmda 0.5 -15.1295 stigs 0.5 -15.1295 “#wheniwasakid” 0.5 -15.1295 shadows 0.5 -15.1295 happiness, 0.5 -15.1319 0123456789 0.5 -15.1319 yay 0.5 -15.1319 (liao 0.5 -15.1319 flæ 0.5 -15.1319 bbdk10 0.5 -15.1319 invites 0.5 -15.1319 mwet 0.5 -15.1319 topdocs 0.5 -15.1319 41.23 0.5 -15.1319 gwas 0.5 -15.1319 catalysis, 0.5 -15.1319 projectivity(kuhlmannand 0.5 -15.1319 ure 0.5 -15.1319 reuters-21578 0.5 -15.1319 grain, 0.5 -15.1319 8since 0.5 -15.1319 logn) 0.5 -15.1319 noncue 0.5 -15.1319 12,782 0.5 -15.1319 (computer 0.5 -15.1319 672 0.5 -15.1319 77.38 0.5 -15.1319 alastvna}yp 0.5 -15.1319 72.38 0.5 -15.1319 tree-structured, 0.5 -15.1319 bodily 0.5 -15.1319 44.58 0.5 -15.1319 easier: 0.5 -15.1319 (v)) 0.5 -15.1319 n-gram/m-slm+n-gram/plsa 0.5 -15.1319 prowess 0.5 -15.1319 foltz 0.5 -15.1319 therapies, 0.5 -15.1319 (effect 0.25 -15.132 l∑ 0.0824742 -15.1331 propose an 0.5 -15.1344 (e1,e2) 0.5 -15.1344 gd(s), 0.5 -15.1344 graph-structured 0.5 -15.1344 manually-selected 0.5 -15.1344 hindito 0.5 -15.1344 hu) 0.5 -15.1344 1982), 0.5 -15.1344 irelandspecific 0.5 -15.1344 {s 0.5 -15.1344 pr¨asident 0.5 -15.1344 64.33 0.5 -15.1344 tseng 0.5 -15.1344 detections) 0.5 -15.1344 21.05 0.5 -15.1344 (forward) 0.5 -15.1344 8.33 0.5 -15.1344 sentence-specific 0.5 -15.1344 ��� 0.5 -15.1344 *department 0.5 -15.1344 719 0.5 -15.1344 yule 0.5 -15.1344 tense), 0.5 -15.1344 genomics, 0.5 -15.1344 programs: 0.5 -15.1344 1/d 0.5 -15.1344 [2008], 0.5 -15.1344 patternmatching 0.5 -15.1344 14.51 0.5 -15.1344 tokenization) 0.5 -15.1344 32.88 0.5 -15.1344 “bat” 0.5 -15.1344 marathi-tourism 0.5 -15.1344 numgen 0.5 -15.1344 (herring 0.5 -15.1344 imperfections 0.5 -15.1344 stringnet’s 0.5 -15.1344 09, 0.5 -15.1344 wiktionary1 0.5 -15.1344 (potential 0.5 -15.1344 rel(c1j,c2k) 0.5 -15.1344 (documents 0.5 -15.1344 2004] 0.5 -15.1368 colleague, 0.5 -15.1368 c(i,j)(ti,tj) 0.5 -15.1368 (attardi 0.5 -15.1368 (sarwar 0.5 -15.1368 unduly 0.5 -15.1368 2008], 0.5 -15.1368 clusterbased 0.5 -15.1368 vanguarders 0.5 -15.1368 samples, resulting 0.5 -15.1368 30.70 0.5 -15.1368 tsvms 0.5 -15.1368 (types 0.5 -15.1368 bt(φi) 0.5 -15.1368 “exit 0.5 -15.1368 54.79 0.5 -15.1368 4700 0.5 -15.1368 19.79 0.5 -15.1368 asr-related 0.5 -15.1368 portals 0.5 -15.1368 ϕlscript, 0.5 -15.1368 unstressed 0.5 -15.1368 research: 0.5 -15.1368 sensespecific 0.5 -15.1368 86.98 0.5 -15.1368 mslq 0.5 -15.1368 unsystematic 0.5 -15.1368 npj+1, 0.5 -15.1368 in-chain 0.5 -15.1368 screen-scrapping 0.5 -15.1368 multicolumn 0.5 -15.1368 83.66 0.5 -15.1368 85.44 0.5 -15.1368 0.237 0.5 -15.1368 lexicon’s 0.5 -15.1368 0.053 0.5 -15.1368 prepositions) 0.5 -15.1368 “guessed” 0.25 -15.1381 =∑ 0.0886076 -15.1389 arguments are 0.5 -15.1392 “laura 0.5 -15.1392 92.16 0.5 -15.1392 mind) 0.5 -15.1392 6(a) 0.5 -15.1392 “shape” 0.5 -15.1392 csp 0.5 -15.1392 5dd, 0.5 -15.1392 93.57 0.5 -15.1392 69.29 0.5 -15.1392 fipscoview: 0.5 -15.1392 apache, 0.5 -15.1392 19.82 0.5 -15.1392 bikes.) 0.5 -15.1392 rendell, 0.5 -15.1392 (97% 0.5 -15.1392 cfcs 0.5 -15.1392 entered” 0.5 -15.1392 mostcommon 0.5 -15.1392 (e,f), 0.5 -15.1392 barbara freethy? 0.5 -15.1392 nodoc 0.5 -15.1392 sixth, 0.5 -15.1392 “seed 0.5 -15.1392 25.17 0.5 -15.1392 κ) 0.5 -15.1392 mbol 0.5 -15.1392 breed 0.5 -15.1392 〈p 0.5 -15.1392 gave, 0.5 -15.1417 metalinguistic 0.5 -15.1417 wfsa, 0.5 -15.1417 90.11 0.5 -15.1417 info, 0.5 -15.1417 inanimate, 0.5 -15.1417 grammar” 0.5 -15.1417 sinatra 0.5 -15.1417 88.27 0.5 -15.1417 emiliano 0.5 -15.1417 gabani 0.5 -15.1417 (rudolph 0.5 -15.1417 3.06 0.5 -15.1417 happines, 0.5 -15.1417 dec, 0.5 -15.1417 remark, 0.5 -15.1417 with2-gramlm 0.5 -15.1417 “follows” 0.5 -15.1417 sets4 0.5 -15.1417 “relational 0.5 -15.1417 (references 0.5 -15.1417 5429 0.5 -15.1417 hypertagging 0.5 -15.1417 non-uniqueness 0.5 -15.1417 +orthofeatures 0.5 -15.1417 42.15 0.5 -15.1417 96.30 0.5 -15.1417 924 0.5 -15.1417 4,6 0.5 -15.1417 pr(p) 0.137931 -15.144 gratefully acknowledge 0.5 -15.1442 hakenberg 0.5 -15.1442 expand, 0.5 -15.1442 andais 0.5 -15.1442 world: 0.5 -15.1442 iia 0.5 -15.1442 regional, 0.5 -15.1442 [rb 0.5 -15.1442 renditions 0.5 -15.1442 paşca 0.5 -15.1442 38.36 0.5 -15.1442 (certain 0.5 -15.1442 39.65 0.5 -15.1442 believed, 0.5 -15.1442 smrz, 0.5 -15.1442 turtles 0.5 -15.1442 ℓ(w;(xt,yt)) 0.5 -15.1442 (peng 0.5 -15.1442 bery 0.5 -15.1442 surface-level 0.5 -15.1442 (dahlmeier 0.5 -15.1442 186,810 0.5 -15.1442 (combined) 0.5 -15.1442 guide, 0.5 -15.1442 (rumor) 0.5 -15.1442 above); 0.5 -15.1442 betweent 0.5 -15.1442 jj; 0.5 -15.1442 ˆei←eprimei 0.5 -15.1442 27.36 0.5 -15.1442 ans: 0.5 -15.1442 eprimei←insert(ecur,a,j) 0.5 -15.1442 cheonan 0.5 -15.1442 axioms; 0.5 -15.1442 ’train 0.1 -15.1447 many more 0.176471 -15.1454 entity recognition. 0.333333 -15.1458 lexconn 0.5 -15.1466 mpmpn 0.5 -15.1466 vl; 0.5 -15.1466 arity, 0.5 -15.1466 indonesian, 0.5 -15.1466 7th, 0.5 -15.1466 87.96 0.5 -15.1466 1,302 0.5 -15.1466 letters; 0.5 -15.1466 95.20 0.5 -15.1466 “asp” 0.5 -15.1466 88.09 0.5 -15.1466 (stem 0.5 -15.1466 top-scored 0.5 -15.1466 (relational) 0.5 -15.1466 impedes 0.5 -15.1466 (aauc) 0.5 -15.1466 70.87 0.5 -15.1466 engine) 0.5 -15.1466 break? 0.5 -15.1466 forbesriley 0.5 -15.1466 ocurrences 0.5 -15.1466 “local 0.5 -15.1466 69.63 0.5 -15.1466 uncovering 0.5 -15.1466 supervision) 0.5 -15.1466 100+ 0.5 -15.1466 (cotton 0.5 -15.1466 entity-origin, 0.5 -15.1466 1865, 0.5 -15.1466 22.61 0.5 -15.1466 41.38 0.5 -15.1466 diagnostics, 0.5 -15.1466 2,203 0.5 -15.1466 28.03 0.5 -15.1466 40.21 0.5 -15.1466 dualist: 0.5 -15.1466 accepting, 0.5 -15.1466 (iida 0.5 -15.1466 (below) 0.0342342 -15.1467 value of 0.5 -15.1491 “perfect 0.5 -15.1491 30.54 0.5 -15.1491 σ(n1,n2) 0.5 -15.1491 patients’ 0.5 -15.1491 dependency–parsed 0.5 -15.1491 1670 0.5 -15.1491 np-completeness 0.5 -15.1491 7news 0.5 -15.1491 transposition, 0.5 -15.1491 semexts 0.5 -15.1491 rel’n clust. 0.5 -15.1491 chat) 0.5 -15.1491 mapsto→z 0.5 -15.1491 promoters) 0.5 -15.1491 48.6% 0.5 -15.1491 spec 0.5 -15.1491 non-phrasal 0.5 -15.1491 16.39 0.5 -15.1491 re-examination 0.5 -15.1491 85.81 0.5 -15.1491 hills, 0.5 -15.1491 “around 0.5 -15.1491 ‘haiti 0.5 -15.1491 factivity 0.5 -15.1491 held, 0.5 -15.1491 tts) 0.5 -15.1491 displays, 0.5 -15.1491 62.75 0.5 -15.1491 vikas 0.5 -15.1491 (verbs, 0.5 -15.1491 68.54 0.5 -15.1491 45.80 0.5 -15.1491 sequence; 0.5 -15.1491 8.4% 0.5 -15.1491 3.0% 0.5 -15.1491 schemaas 0.5 -15.1491 acceptable; 0.5 -15.1491 0.623 0.5 -15.1491 0.13) 0.5 -15.1491 53.9% 0.25 -15.1504 karin 0.333333 -15.1508 sounding 0.5 -15.1516 (prepositional 0.5 -15.1516 pi: 0.5 -15.1516 aw, 0.5 -15.1516 (accuracy, 0.5 -15.1516 51.20 0.5 -15.1516 m=3 0.5 -15.1516 3way 0.5 -15.1516 nlf, 0.5 -15.1516 vista 0.5 -15.1516 engine; 0.5 -15.1516 “coordination”, 0.5 -15.1516 (lesk, 0.5 -15.1516 baral 0.5 -15.1516 ntp, 0.5 -15.1516 tokenise 0.5 -15.1516 9.78 0.5 -15.1516 yacc 0.5 -15.1516 outlined, 0.5 -15.1516 co-authors 0.5 -15.1516 wiktoria golik, 0.5 -15.1516 multi-valued 0.5 -15.1516 (keep 0.5 -15.1516 (hw), 0.5 -15.1516 presentverbal 0.5 -15.1516 552 0.5 -15.1516 42.36 0.5 -15.1516 25.70 0.5 -15.1516 6.2%, 0.5 -15.1516 goods, 0.5 -15.1516 meeting recorder 0.5 -15.1516 ger/eng1 0.5 -15.1516 eq[f(z)] 0.5 -15.1516 i=1p(xi) 0.5 -15.1516 cornetto 0.5 -15.1516 87.69 0.5 -15.1516 n×m 0.5 -15.1516 bavarian 0.5 -15.1516 96.88 0.5 -15.1516 toy-e 0.5 -15.1516 62.5%) 0.5 -15.1516 ¯o, 0.333333 -15.1524 word+nw 0.0582011 -15.1532 this means 0.5 -15.1541 63.50% 0.5 -15.1541 occasions, 0.5 -15.1541 65.48 0.5 -15.1541 systematizing 0.5 -15.1541 20.6% 0.5 -15.1541 [[utt]]c 0.5 -15.1541 formality: 0.5 -15.1541 14.22 0.5 -15.1541 ameliorates 0.5 -15.1541 (sima’an, 0.5 -15.1541 genpex; 0.5 -15.1541 [cc] 0.5 -15.1541 (passonneau 0.5 -15.1541 sonorants 0.5 -15.1541 (68.5%) 0.5 -15.1541 chery 0.5 -15.1541 errorcorrecting 0.5 -15.1541 (sogrpl) 0.5 -15.1541 big-o 0.5 -15.1541 paloma 0.5 -15.1541 (-0.75, 0.5 -15.1541 para tr´as 0.5 -15.1541 fence 0.5 -15.1541 nmi, 0.5 -15.1541 analyze, 0.5 -15.1541 mnullnullnullnullnull 0.5 -15.1541 nsga-ii 0.5 -15.1541 instance-labeling 0.5 -15.1541 em-lp-128 0.5 -15.1541 premier 0.5 -15.1541 10.39 0.5 -15.1541 distinguished: 0.5 -15.1541 get-optimal-action(p,o) 0.5 -15.1541 flp, 0.5 -15.1541 non-ascii 0.5 -15.1541 cj≥ 0.5 -15.1541 side-effects 0.5 -15.1541 (luhn, 0.5 -15.1541 73.87 0.5 -15.1541 merge: 0.5 -15.1541 tagging/parsing 0.25 -15.1554 #docs 0.333333 -15.1558 i’d 0.333333 -15.1558 lex-items 0.5 -15.1565 “tramped” 0.5 -15.1565 ‘no’s 0.5 -15.1565 c-structure 0.5 -15.1565 6here, 0.5 -15.1565 agreeableness, 0.5 -15.1565 reformulates 0.5 -15.1565 values); 0.5 -15.1565 amplifiers 0.5 -15.1565 kekäläinen, 0.5 -15.1565 (going 0.5 -15.1565 e(y) 0.5 -15.1565 lengel 0.5 -15.1565 87.05 0.5 -15.1565 analsysis 0.5 -15.1565 48.56 0.5 -15.1565 maximumnumber 0.5 -15.1565 (flesch, 0.5 -15.1565 pla 0.5 -15.1565 s-parser 0.5 -15.1565 affiliation, 0.5 -15.1565 metric), 0.5 -15.1565 62.84 0.5 -15.1565 en-cs, 0.5 -15.1565 cair 0.5 -15.1565 65.42 0.5 -15.1565 y∗(x) 0.5 -15.1565 >0.9 0.25 -15.1566 elizabeth 0.333333 -15.1574 self-learning 0.333333 -15.1574 notwithstanding 0.333333 -15.1574 sassano 0.5 -15.159 ci−1 0.5 -15.159 contacting 0.5 -15.159 sonar 0.5 -15.159 wheras 0.5 -15.159 82.41 0.5 -15.159 53.06 0.5 -15.159 jami 0.5 -15.159 niques 0.5 -15.159 alludes 0.5 -15.159 85.3% 0.5 -15.159 virulence, 0.5 -15.159 t[sj+1] 0.5 -15.159 tapanainen, 0.5 -15.159 far-misses 0.5 -15.159 [prompt] 0.5 -15.159 nigerian 0.5 -15.159 strtm’s 0.5 -15.159 lperm, 0.5 -15.159 thurlow 0.5 -15.159 til 0.5 -15.159 sue 0.5 -15.159 >); 0.5 -15.159 26.08 0.5 -15.159 (mc-closky 0.5 -15.159 (bunescu06) 0.5 -15.159 addresses) 0.5 -15.159 92.87 0.5 -15.159 cats, 0.5 -15.159 2×104 0.5 -15.159 mrf, 0.5 -15.159 (m3) 0.5 -15.159 φjoin(yr,z) 0.5 -15.159 “dropped 0.5 -15.159 56.06 0.5 -15.159 senna 0.5 -15.159 (ptb), 0.5 -15.159 91.66 0.5 -15.159 two-sentence 0.5 -15.159 (cristianini 0.5 -15.159 0.558 0.5 -15.159 acme 0.5 -15.159 saeger 0.5 -15.159 sherif 0.5 -15.159 chuck 0.333333 -15.1591 insulating 0.333333 -15.1591 $5,000 0.333333 -15.1591 treasure 0.333333 -15.1591 reward: 0.333333 -15.1591 pre-clusterpair 0.333333 -15.1591 near-synonym 0.333333 -15.1591 können 0.333333 -15.1591 ��'wt�luk��b. 0.333333 -15.1591 quadrants 0.333333 -15.1591 perform desired action 0.333333 -15.1591 dc cwcxbncycx 0.333333 -15.1591 predications. 0.333333 -15.1591 jordan. 0.333333 -15.1591 matcher. 0.333333 -15.1591 bpng 0.333333 -15.1591 weaknesses. 0.333333 -15.1591 geo:new 0.333333 -15.1591 chart[b, 0.333333 -15.1591 deadly 0.333333 -15.1591 heuristicrules 0.333333 -15.1591 conn+conn 0.333333 -15.1591 acl-08: hlt, pages 0.333333 -15.1591 [h1,i,h2h3,k] 0.333333 -15.1591 %imp 0.333333 -15.1591 had. 0.333333 -15.1591 bricht 0.333333 -15.1591 0.333333 -15.1591 unaries 0.333333 -15.1591 arcright 0.333333 -15.1591 npj, 0.333333 -15.1591 suicide notes. 0.333333 -15.1591 (s\np)/np 0.333333 -15.1591 r2: 0.176471 -15.1594 sou 0.333333 -15.1607 human–human 0.333333 -15.1607 guan, 0.333333 -15.1607 99.26 0.333333 -15.1607 377 0.333333 -15.1607 “intermediate,” 0.333333 -15.1607 783 0.333333 -15.1607 withdraw 0.333333 -15.1607 13.77 0.333333 -15.1607 humanistic 0.333333 -15.1607 (age 0.333333 -15.1607 t-class 0.333333 -15.1607 6.00 0.333333 -15.1607 combing 0.0420168 -15.1608 results show 0.133333 -15.1611 fifteenth conference 0.0843373 -15.1614 it could 0.5 -15.1615 file-based 0.5 -15.1615 units” 0.5 -15.1615 unknown/unspecified 0.5 -15.1615 0.5 -15.1818 (corney 0.5 -15.1818 jianxing 0.5 -15.1818 42.54 0.5 -15.1818 jours, 0.5 -15.1818 »q 0.5 -15.1818 c-lda-t, 0.5 -15.1818 (rohde, 0.5 -15.1818 corpus−based 0.5 -15.1818 5.30 0.5 -15.1818 47.45 0.5 -15.1818 thaiherbminer 0.5 -15.1818 (ignores 0.5 -15.1818 inextricably 0.5 -15.1818 np/n 0.5 -15.1818 61.88 0.5 -15.1818 undisputed 0.5 -15.1818 regress 0.5 -15.1818 kmm 0.5 -15.1818 corel 0.5 -15.1818 49.16 0.5 -15.1818 formations 0.5 -15.1818 slope, 0.5 -15.1818 length−→ 0.5 -15.1818 undergoer 0.5 -15.1818 /w/and, 0.5 -15.1818 (hermet 0.5 -15.1818 2,655 0.5 -15.1818 zadrozny, 0.5 -15.1818 “tax 0.5 -15.1818 twiter-based 0.5 -15.1818 68.06 0.333333 -15.1827 over-specified 0.333333 -15.1827 vp/nn 0.333333 -15.1827 heur-w 0.333333 -15.1827 misidentified 0.333333 -15.1827 2however, 0.333333 -15.1827 (”for 0.333333 -15.1827 neologisms 0.333333 -15.1827 complain, 0.0875 -15.1838 pattern. 0.25 -15.1844 has rank 0.333333 -15.1844 jerusalem, 0.333333 -15.1844 swo_m 0.333333 -15.1844 “sweep” 0.333333 -15.1844 33.47 0.333333 -15.1844 routledge 0.333333 -15.1844 kurdish 0.333333 -15.1844 65.46 0.333333 -15.1844 hm(x,y) 0.333333 -15.1844 sentence): 0.333333 -15.1844 graphof 0.333333 -15.1844 --------------- 0.333333 -15.1844 uprimek(s,u) 0.5 -15.1844 a8 0.5 -15.1844 64.48 0.5 -15.1844 95.10 0.5 -15.1844 wdhmm (he, 0.5 -15.1844 saldo, 0.5 -15.1844 parasense1 0.5 -15.1844 wcprimej 0.5 -15.1844 49.98 0.5 -15.1844 442 0.5 -15.1844 shakti 0.5 -15.1844 36.75 0.5 -15.1844 c(j) 0.5 -15.1844 38.55 0.5 -15.1844 was: 0.5 -15.1844 (schroeder 0.5 -15.1844 label(n) 0.5 -15.1844 907 0.5 -15.1844 communication: 0.5 -15.1844 4.) 0.5 -15.1844 ex, 0.5 -15.1844 viterbi-best 0.5 -15.1844 81.77 0.5 -15.1844 “explain” 0.5 -15.1844 demonym 0.5 -15.1844 (concepts) 0.5 -15.1844 (ro) 0.5 -15.1844 personalisation 0.5 -15.1844 94.30 0.5 -15.1844 nullid 0.5 -15.1844 tintarev, 0.5 -15.1844 (brenner 0.5 -15.1844 rau 0.5 -15.1844 causatives 0.5 -15.1844 3,179 0.5 -15.1844 “,” 0.5 -15.1844 down” 0.5 -15.1844 32.15 0.5 -15.1844 ratings) 0.5 -15.1844 cremation 0.5 -15.1844 pseudoprojective 0.5 -15.1844 utilisation 0.5 -15.1844 cases; 0.5 -15.1844 localmax, 0.5 -15.1844 (cs09) 0.5 -15.1844 (john 0.0952381 -15.1857  0.333333 -15.1861 appear) 0.333333 -15.1861 post) 0.333333 -15.1861 source) 0.333333 -15.1861 66.25 0.333333 -15.1861 “brush 0.333333 -15.1861 9.5% 0.333333 -15.1861 graspresponses 0.333333 -15.1861 vdss 0.333333 -15.1861 palm, 0.333333 -15.1861 tetsuya 0.333333 -15.1861 apologies 0.333333 -15.1861 t32 0.5 -15.1869 angular 0.5 -15.1869 αdk 0.5 -15.1869 0s, 0.5 -15.1869 r∈rδ(r) 0.5 -15.1869 40.06 0.5 -15.1869 19.83 0.5 -15.1869 respondent’s 0.5 -15.1869 (parse 0.5 -15.1869 cleanup 0.5 -15.1869 pepper 0.5 -15.1869 silently 0.5 -15.1869 91.88 0.5 -15.1869 1396 0.5 -15.1869 hopeful 0.5 -15.1869 16.2% 0.5 -15.1869 n-only 0.5 -15.1869 92.2% 0.5 -15.1869 hour’s 0.5 -15.1869 delattre, 0.5 -15.1869 kamps, 0.5 -15.1869 “roundtrip” 0.5 -15.1869 (voice 0.5 -15.1869 countries: 0.5 -15.1869 sensedisambiguated 0.5 -15.1869 85.80 0.5 -15.1869 -0.75 0.5 -15.1869 70.27 0.5 -15.1869 87.88 0.5 -15.1869 (e-j) 0.5 -15.1869 3.54 0.5 -15.1869 wi,wj 0.5 -15.1869 0.414 0.5 -15.1869 34.25 0.5 -15.1869 supposing 0.5 -15.1869 -bo 0.5 -15.1869 cookie, 0.5 -15.1869 modifications) 0.5 -15.1869 polka 0.5 -15.1869 3.5.2, 0.5 -15.1869 unscoped 0.5 -15.1869 caraballo 0.5 -15.1869 91.87 0.333333 -15.1878 review: 0.333333 -15.1878 refuses 0.333333 -15.1878 86.63 0.333333 -15.1878 ambassador 0.333333 -15.1878 (bf), 0.333333 -15.1878 (γ) 0.333333 -15.1878 stm10 0.333333 -15.1878 wel-known 0.333333 -15.1878 “politics” 0.333333 -15.1878 reliability: 0.333333 -15.1878 15.8% 0.333333 -15.1878 profile, 0.333333 -15.1878 2,200 0.0784314 -15.1881 itself. 0.25 -15.1883 (sa) 0.333333 -15.1895 94.23 0.333333 -15.1895 czech: 0.333333 -15.1895 pc(e) 0.333333 -15.1895 k(t1,t2) 0.333333 -15.1895 multiplications 0.333333 -15.1895 lexeme-synset 0.333333 -15.1895 87.37 0.333333 -15.1895 1326 0.333333 -15.1895 0.333 0.333333 -15.1895 xik 0.333333 -15.1895 np0,1 0.5 -15.1895 hm1hm2 0.5 -15.1895 modestly 0.5 -15.1895 graça 0.5 -15.1895 cpu-time 0.5 -15.1895 tightest 0.5 -15.1895 processes) 0.5 -15.1895 tsuji 0.5 -15.1895 (bazzi, 0.5 -15.1895 (kj 0.5 -15.1895 ,condition 0.5 -15.1895 elements userusage[i+1] 0.5 -15.1895 0.14% 0.5 -15.1895 expectations: 0.5 -15.1895 cepts 0.5 -15.1895 (cer), 0.5 -15.1895 parse-tree-based 0.5 -15.1895 definiens 0.5 -15.1895 machinegenerated 0.5 -15.1895 “value” 0.5 -15.1895 “peer” 0.5 -15.1895 aliens 0.5 -15.1895 64.75 0.5 -15.1895 u), 0.5 -15.1895 comment: 0.5 -15.1895 “srilm – 0.5 -15.1895 toss 0.5 -15.1895 emrs, 0.333333 -15.1913 wear 0.333333 -15.1913 fv 0.333333 -15.1913 89.38 0.333333 -15.1913 ys−1 0.333333 -15.1913 59.18 0.333333 -15.1913 (tagged 0.333333 -15.1913 16.53 0.333333 -15.1913 (g1) 0.333333 -15.1913 because) 0.333333 -15.1913 <1, 0.333333 -15.1913 34, 0.5 -15.1921 (22% 0.5 -15.1921 submission-ws 0.5 -15.1921 2n, 0.5 -15.1921 discharge sumaries 0.5 -15.1921 6580 0.5 -15.1921 67.12 0.5 -15.1921 parsers’ 0.5 -15.1921 (a)) 0.5 -15.1921 punc. 0.5 -15.1921 26.77 0.5 -15.1921 34.64 0.5 -15.1921 s’s 0.5 -15.1921 star fj,ek) 0.5 -15.1921 oav 0.5 -15.1921 26.1% 0.5 -15.1921 primitive holds, 0.5 -15.1921 improbable 0.5 -15.1921 non-zero, 0.5 -15.1921 bleu-4, 0.5 -15.1921 accomplishment 0.5 -15.1921 colour[white] 0.5 -15.1921 graph-ave 0.5 -15.1921 3,300 0.5 -15.1921 2007] 0.5 -15.1921 29.30 0.5 -15.1921 (hidden 0.5 -15.1921 ouputs’ 0.5 -15.1921 occurances 0.5 -15.1921 15.67 0.5 -15.1921 51.65 0.5 -15.1921 fulfiling 0.5 -15.1921 19.51 0.5 -15.1921 uturku09 0.5 -15.1921 $4 0.5 -15.1921 possvm 0.5 -15.1921 fl, 0.5 -15.1921 m=o(n 0.5 -15.1921 top.a, 0.5 -15.1921 einstein’s 0.5 -15.1921 (bairoch 0.5 -15.1921 (sentiment 0.5 -15.1921 bruijn, 0.5 -15.1921 k-best, 0.5 -15.1921 2.7) 0.5 -15.1921 38.63 0.5 -15.1921 suffers, 0.5 -15.1921 popularity rankings: 0.5 -15.1921 i(a) 0.5 -15.1921 g=(v 0.5 -15.1921 lexcat 0.333333 -15.193 2.60 0.333333 -15.193 +0.19 0.333333 -15.193 algorithmis 0.333333 -15.193 surounding 0.333333 -15.193 weps-3 0.333333 -15.193 rahul 0.333333 -15.193 nonexecutive 0.25 -15.1934 34.41 0.25 -15.1947 nicta 0.333333 -15.1947 10.13 0.333333 -15.1947 ly, 0.333333 -15.1947 averages, 0.333333 -15.1947 names), 0.333333 -15.1947 location), 0.333333 -15.1947 gardening, 0.333333 -15.1947 90.82 0.5 -15.1947 negative_regulation 0.5 -15.1947 stanford 26.56 0.5 -15.1947 orginal 0.5 -15.1947 substitution: 0.5 -15.1947 (giles, 0.5 -15.1947 pmajority 0.5 -15.1947 dist(v,h) 0.5 -15.1947 10.61 0.5 -15.1947 q-q 0.5 -15.1947 imports 0.5 -15.1947 d′′′ 0.5 -15.1947 jong-bok 0.5 -15.1947 (structural 0.5 -15.1947 l(h,r) 0.5 -15.1947 opinion-annotated 0.5 -15.1947 3,099 0.5 -15.1947 0.986 0.5 -15.1947 (english-hindi) 0.5 -15.1947 lug 0.5 -15.1947 27.46 0.5 -15.1947 (formerly 0.5 -15.1947 73.27 0.5 -15.1947 bigrams: 0.5 -15.1947 90.87 0.5 -15.1947 teslab 0.5 -15.1947 e1...el, 0.5 -15.1947 thisis 0.5 -15.1947 sub-domain 0.5 -15.1947 91.40 0.5 -15.1947 57.15 0.5 -15.1947 isomorphism, 0.5 -15.1947 36.42 0.5 -15.1947 lowconfidence 0.5 -15.1947 biber 0.5 -15.1947 substitution,a 0.5 -15.1947 (400 0.5 -15.1947 92.88 0.5 -15.1947 1,208 0.5 -15.1947 diferences, 0.5 -15.1947 semi-definite 0.0530973 -15.1952 referred 0.0346154 -15.1958 2009; 0.333333 -15.1964 x-from 0.333333 -15.1964 1397 0.333333 -15.1964 san-segundo 0.333333 -15.1964 “curb” 0.333333 -15.1964 33.98 0.333333 -15.1964 sbar\whnp 0.333333 -15.1964 22.13 0.5 -15.1973 goal-oriented 0.5 -15.1973 36.61 0.5 -15.1973 mets 0.5 -15.1973 ptb’s 0.5 -15.1973 55.19 0.5 -15.1973 82.95 0.5 -15.1973 wordbreak 0.5 -15.1973 correct: 0.5 -15.1973 lagging 0.5 -15.1973 fine-grainedness 0.5 -15.1973 five-gram 0.5 -15.1973 0.479 0.5 -15.1973 forster 0.5 -15.1973 13.99 0.5 -15.1973 olena 0.5 -15.1973 cities), 0.5 -15.1973 marrafa, 0.5 -15.1973 2e8 0.5 -15.1973 intraannotator 0.5 -15.1973 11.68 0.5 -15.1973 18.78 0.5 -15.1973 causality: 0.5 -15.1973 taxonomy: 0.5 -15.1973 management; 0.5 -15.1973 109,553 0.5 -15.1973 detection-error 0.5 -15.1973 snippets’ 0.5 -15.1973 premodifiers 0.5 -15.1973 caches, 0.5 -15.1973 65.81 0.5 -15.1973 θprime 0.5 -15.1973 52.76 0.5 -15.1973 joanne 0.5 -15.1973 deeper, 0.5 -15.1973 errortagged 0.5 -15.1973 (positive/negative) 0.5 -15.1973 hand,and 0.5 -15.1973 ˆψ 0.5 -15.1973 blindly 0.5 -15.1973 35.82 0.5 -15.1973 initialization; 0.0776699 -15.1979 ith 0.333333 -15.1982 799 0.333333 -15.1982 regulons 0.333333 -15.1982 andw 0.333333 -15.1982 mutation, 0.333333 -15.1982 diarmuid 0.333333 -15.1982 27.98 0.333333 -15.1982 blood port-au-prince 0.333333 -15.1982 vp→ 0.333333 -15.1982 therefore,we 0.333333 -15.1982 np” 0.333333 -15.1982 88.40 0.285714 -15.1985 faire 0.25 -15.1986 (maccartney 0.222222 -15.1988 shoes 0.0569948 -15.1989 2000; 0.0909091 -15.199 be learned 0.333333 -15.1999 bigram-based 0.333333 -15.1999 hod 0.333333 -15.1999 message-passing 0.333333 -15.1999 systemin 0.333333 -15.1999 1020 0.333333 -15.1999 79.34 0.333333 -15.1999 4these 0.333333 -15.1999 “browsing 0.333333 -15.1999 exact: 0.333333 -15.1999 f⟨1,e1,x⟩ 0.333333 -15.1999 leave-one-speaker-out 0.5 -15.1999 andz 0.5 -15.1999 huc 0.5 -15.1999 71.19 0.5 -15.1999 (krifka 0.5 -15.1999 l∪ 0.5 -15.1999 26.76 0.5 -15.1999 interrupted, 0.5 -15.1999 transporter 0.5 -15.1999 27.28 0.5 -15.1999 41.75 0.5 -15.1999 rishøj, 0.5 -15.1999 24.94 0.5 -15.1999 2,930 0.5 -15.1999 (finley 0.5 -15.1999 deliver, 0.5 -15.1999 (changeto’ 0.5 -15.1999 ishikawa 0.5 -15.1999 brogi 0.5 -15.1999 29.59 0.5 -15.1999 unconn 0.5 -15.1999 72.92 0.5 -15.1999 n-gramfeatures 0.5 -15.1999 words/utt 0.5 -15.1999 s/n, 0.5 -15.1999 (semi-)supervised 0.5 -15.1999 10.58 0.5 -15.1999 (chinchor 0.5 -15.1999 instr 0.5 -15.1999 embeddings 0.5 -15.1999 53.68 0.5 -15.1999 crl(p) 0.5 -15.1999 32.16 0.5 -15.1999 55.80 0.5 -15.1999 [festival 0.5 -15.1999 plan: 0.5 -15.1999 44.69 0.5 -15.1999 ℓ(w;(x,y)) 0.5 -15.1999 read– 0.5 -15.1999 30.32 0.5 -15.1999 “woman 0.5 -15.1999 figure9 0.5 -15.1999 (coverage 0.5 -15.1999 provider’s 0.5 -15.1999 subject-objectverb 0.5 -15.1999 reasoning: 0.5 -15.1999 “facebook” 0.2 -15.2009 (between 0.285714 -15.2014 ir-status 0.333333 -15.2016 ahah 0.333333 -15.2016 1763 0.333333 -15.2016 olympic 0.333333 -15.2016 x”, 0.333333 -15.2016 unique: 0.333333 -15.2016 majority, 0.5 -15.2025 calico 0.5 -15.2025 a1 ···an 0.5 -15.2025 potential-specific 0.5 -15.2025 classrooms 0.5 -15.2025 0.2349 0.5 -15.2025 kumano 0.5 -15.2025 education: 0.5 -15.2025 (seconds); 0.5 -15.2025 sfl 0.5 -15.2025 56.09 0.5 -15.2025 haddock 0.5 -15.2025 bankrupted 0.5 -15.2025 prudent 0.5 -15.2025 ey 0.5 -15.2025 r(e), 0.5 -15.2025 (putting 0.5 -15.2025 dallas, tx. 0.5 -15.2025 (more) 0.5 -15.2025 ]s3 0.5 -15.2025 68.71 0.5 -15.2025 guelph 0.5 -15.2025 34.98 0.5 -15.2025 manifold-ranking 0.5 -15.2025 renewing 0.5 -15.2025 revolutionized 0.5 -15.2025 “candidates” 0.5 -15.2025 (5)) 0.5 -15.2025 (lyons, 0.5 -15.2025 pscfg-based 0.5 -15.2025 set3 0.5 -15.2025 goalkeeper 0.5 -15.2025 bansalandklein(2010)mcp 0.333333 -15.2034 wfd 0.333333 -15.2034 lemmatiser 0.333333 -15.2034 ddeterminer 0.333333 -15.2034 37th 0.333333 -15.2034 “before”, 0.333333 -15.2034 x(n) 0.333333 -15.2034 neu 0.333333 -15.2034 ending, 0.333333 -15.2034 2.22 0.333333 -15.2034 yousg 0.333333 -15.2051 φcon 0.333333 -15.2051 55.60 0.333333 -15.2051 d(e))” 0.333333 -15.2051 lemur 0.333333 -15.2051 49.45 0.333333 -15.2051 (pair) 0.333333 -15.2051 (location 0.333333 -15.2051 contra 0.333333 -15.2051 designation 0.333333 -15.2051 eagles 0.333333 -15.2051 headdriven parsing 0.333333 -15.2051 mono 0.333333 -15.2051 ser, 0.5 -15.2051 traits, 0.5 -15.2051 30.41 0.5 -15.2051 λ-forests 0.5 -15.2051 going), 0.5 -15.2051 absurd 0.5 -15.2051 600,000 0.5 -15.2051 gmbh, 0.5 -15.2051 parsimony: 0.5 -15.2051 (hockenmaier, 0.5 -15.2051 eitheror, 0.5 -15.2051 1.8m 0.5 -15.2051 hwdep, 0.5 -15.2051 irp1 0.5 -15.2051 gov. 0.5 -15.2051 (equally 0.5 -15.2051 thus; 0.5 -15.2051 bids 0.5 -15.2051 misrecognized, 0.5 -15.2051 non-generic 0.5 -15.2051 (bolded), 0.5 -15.2051 56.43 0.5 -15.2051 p7tb 0.5 -15.2051 concluded, 0.5 -15.2051 drunk 0.5 -15.2051 phone) 0.5 -15.2051 segments; 0.5 -15.2051 1354 0.5 -15.2051 menu, pie, 0.5 -15.2051 varφ[x] 0.5 -15.2051 multiway 0.5 -15.2051 csr(α) 0.5 -15.2051 canonically 0.5 -15.2051 88.51 0.5 -15.2051 norel) 0.5 -15.2051 “raise 0.5 -15.2051 nissim+lexical+parse 0.5 -15.2051 above-median 0.5 -15.2051 2.1x 0.5 -15.2051 edus, 0.333333 -15.2069 sim(u,v) 0.333333 -15.2069 svd, 0.333333 -15.2069 baseline-lr 0.333333 -15.2069 spans: 0.333333 -15.2069 n-m 0.0212454 -15.2074 model, 0.5 -15.2077 93.07 0.5 -15.2077 “¸@ 0.5 -15.2077 (bio-crf 0.5 -15.2077 text- 0.5 -15.2077 µ(l)i 0.5 -15.2077 submenu 0.5 -15.2077 59.92 0.5 -15.2077 genolist 0.5 -15.2077 inappropriate, 0.5 -15.2077 machine; 0.5 -15.2077 /e/, 0.5 -15.2077 4.64 0.5 -15.2077 low-recall 0.5 -15.2077 damage) 0.5 -15.2077 unexpectedness 0.5 -15.2077 plaire 0.5 -15.2077 maxent: 0.5 -15.2077 evaluation”, 0.5 -15.2077 newtonian 0.5 -15.2077 ais 0.5 -15.2077 75.09 0.5 -15.2077 bx(x) 0.5 -15.2077 (lvms) 0.5 -15.2077 appositives 0.5 -15.2077 sub-divide 0.5 -15.2077 0.60, 0.5 -15.2077 array) 0.5 -15.2077 57.38 0.5 -15.2077 tfirfw 0.5 -15.2077 vrd 0.5 -15.2077 testament%1:10:00 0.5 -15.2077 undersample 0.5 -15.2077 3c 0.5 -15.2077 romanian-english 0.333333 -15.2086 753 0.333333 -15.2086 (yprime) 0.333333 -15.2086 urgency 0.333333 -15.2086 25k 0.333333 -15.2086 x1,...,xn 0.333333 -15.2086 ¯m), 0.0428135 -15.2088 portion 0.5 -15.2103 spaces), 0.5 -15.2103 73k 0.5 -15.2103 model): 0.5 -15.2103 87.18 0.5 -15.2103 bleu+1 0.5 -15.2103 nrl 0.5 -15.2103 ps−pbpb 0.5 -15.2103 divide-andconquer 0.5 -15.2103 builtpps={ρ(r1)···ρ(rn)}←{w0 0.5 -15.2103 wilbur 0.5 -15.2103 (heckerman, 0.5 -15.2103 detailedin 0.5 -15.2103 responsabilizar 0.5 -15.2103 professional, 0.5 -15.2103 pos(si) 0.5 -15.2103 (sha 0.5 -15.2103 &part-of 0.5 -15.2103 misses) 0.5 -15.2103 (shatkay 0.5 -15.2103 16.42 0.5 -15.2103 2or 0.5 -15.2103 subjects) 0.5 -15.2103 english-czech, 0.5 -15.2103 (citations, 0.5 -15.2103 recognized; 0.5 -15.2103 hohensee, 0.5 -15.2103 gest 0.5 -15.2103 “entity 0.5 -15.2103 films, 0.5 -15.2103 58.00 0.5 -15.2103 predicted: 0.5 -15.2103 miyao, 0.5 -15.2103 davison 0.5 -15.2103 as' 0.5 -15.2103 42.33 0.5 -15.2103 p(tnotarrowrighth) 0.5 -15.2103 tp+fp) 0.5 -15.2103 (σ, nprime, 0.5 -15.2103 d⇒w 0.5 -15.2103 oversegmentation 0.5 -15.2103 2.84 0.5 -15.2103 i7 0.5 -15.2103 statistically, 0.5 -15.2103 (vinokourov 0.5 -15.2103 wiegand 0.25 -15.2104 human) 0.333333 -15.2104 recap, 0.333333 -15.2104 (constructed 0.333333 -15.2104 “only 0.333333 -15.2104 53.98 0.333333 -15.2104 3.59 0.333333 -15.2104 authorities, 0.333333 -15.2104 “according 0.333333 -15.2104 q(y,z) 0.333333 -15.2104 atop 0.333333 -15.2104 0.457 0.333333 -15.2104 (2007, 0.333333 -15.2104 (pl) 0.333333 -15.2104 q2: 0.0448718 -15.2106 we would like 0.222222 -15.2116 jungen 0.333333 -15.2121 52.20 0.333333 -15.2121 partialy 0.333333 -15.2121 murder, 0.333333 -15.2121 shudo, 0.333333 -15.2121 l2s 0.333333 -15.2121 (bigrams) 0.333333 -15.2121 accu 0.333333 -15.2121 olney, 0.333333 -15.2121 +0.31 0.333333 -15.2121 1453 0.333333 -15.2121 (are 0.5 -15.213 ν, 0.5 -15.213 constrast 0.5 -15.213 5-v+f 0.5 -15.213 14.11 0.5 -15.213 licensees 0.5 -15.213 article,we 0.5 -15.213 1042 0.5 -15.213 jaccard’s 0.5 -15.213 bikel, 0.5 -15.213 mcrf 0.5 -15.213 (skantze, 0.5 -15.213 cute 0.5 -15.213 arrests 0.5 -15.213 cannot: 0.5 -15.213 exp{w 0.5 -15.213 “summary” 0.5 -15.213 ø¸ 0.5 -15.213 deve[should] 0.5 -15.213 nonlinearity 0.5 -15.213 −hatwidey 0.5 -15.213 etc.2 0.5 -15.213 welch’s 0.5 -15.213 (1,1,1) 0.5 -15.213 `b 0.5 -15.213 communism 0.5 -15.213 47.3% 0.5 -15.213 predictable, 0.5 -15.213 0.025) 0.5 -15.213 (augusto, 0.5 -15.213 46.30 0.5 -15.213 citizen 0.5 -15.213 initiative (si), 0.5 -15.213 86.94 0.5 -15.213 xtest 0.5 -15.213 non-terminals; 0.5 -15.213 (150 0.5 -15.213 humanand 0.5 -15.213 abvd 0.5 -15.213 and)c(and 0.0804598 -15.2131 total, 0.333333 -15.2139 adj-only 0.333333 -15.2139 inesc-id 0.333333 -15.2139 23k 0.285714 -15.2149 tomaž 0.5 -15.2156 itexpl, 0.5 -15.2156 patrik 0.5 -15.2156 “above” 0.5 -15.2156 “both” 0.5 -15.2156 london: 0.5 -15.2156 ($) 0.5 -15.2156 48.69 0.5 -15.2156 24.46 0.5 -15.2156 head: 0.5 -15.2156  0.5 -15.2156 ignite 0.5 -15.2156 15,001 0.5 -15.2156 40.92 0.5 -15.2156 f1j 0.5 -15.2156 subvector 0.5 -15.2156 44.52 0.5 -15.2156 parenleftbigsummationtextn 0.5 -15.2156 smrž 0.5 -15.2156 svm-light-tk 0.5 -15.2156 conv(t) 0.5 -15.2156 31.42 0.5 -15.2156 “second” 0.5 -15.2156 system-specific 0.5 -15.2156 (monson 0.5 -15.2156 doctor, 0.5 -15.2156 (j48) 0.5 -15.2156 hardware, 0.5 -15.2156 ˛º 0.5 -15.2156 (3)+(4)+(5) 0.333333 -15.2157 sums, 0.333333 -15.2157 thorsten 0.333333 -15.2157 33.5% 0.333333 -15.2157 φ(h,ξ) 0.333333 -15.2157 tires 0.333333 -15.2157 632 0.222222 -15.2163 nst 0.0697674 -15.2172 belongs 0.333333 -15.2174 1548 0.333333 -15.2174 k% 0.333333 -15.2174 technique; 0.333333 -15.2174 already, 0.333333 -15.2174 kamvar 0.5 -15.2183 2456 0.5 -15.2183 classificationand 0.5 -15.2183 35.59 0.5 -15.2183 '08 0.5 -15.2183 88.6% 0.5 -15.2183 poet, 0.5 -15.2183 month) 0.5 -15.2183 eliminate: 0.5 -15.2183 c+i 0.5 -15.2183 queried, 0.5 -15.2183 hale, 0.5 -15.2183 xh 0.5 -15.2183 i(i,j)(ti,tj) 0.5 -15.2183 iconographic 0.5 -15.2183 (c1000) 0.5 -15.2183 woefully 0.5 -15.2183 “seattle” 0.5 -15.2183 tramped 0.5 -15.2183 cfv=4,h=2 0.5 -15.2183 populist 0.5 -15.2183 “hope” 0.5 -15.2183 hyperpriors 0.5 -15.2183 li‡ 0.5 -15.2183 andnullnullnullnullnullnullnullr 0.5 -15.2183 mori, 0.5 -15.2183 19.63 0.5 -15.2183 51.95 0.5 -15.2183 e@ 0.5 -15.2183 29.69 0.5 -15.2183 26.21 0.5 -15.2183 senserelated 0.5 -15.2183 62.27% 0.5 -15.2183 guilty 0.5 -15.2183 48.54 0.5 -15.2183 kiosk 0.5 -15.2183 cros-validation, 0.5 -15.2183 1293 0.5 -15.2183 90.15 0.5 -15.2183 buchanan 0.5 -15.2183 15.83 0.5 -15.2183 89.40 0.333333 -15.2192 40.18 0.333333 -15.2192 regex 0.333333 -15.2192 “protein 0.333333 -15.2192 barrowhookleft→a 0.333333 -15.2192 (result 0.333333 -15.2192 sps 0.333333 -15.2192 flattering 0.333333 -15.2192 (+), 0.333333 -15.2192 activations 0.0227464 -15.2195 paper, 0.5 -15.2209 max-f 0.5 -15.2209 table4 0.5 -15.2209 1/4, 0.5 -15.2209 -pro 0.5 -15.2209 sj1, 0.5 -15.2209 p1: 0.5 -15.2209 taxes, 0.5 -15.2209 document, showing 0.5 -15.2209 56k 0.5 -15.2209 triangle: 0.5 -15.2209 sarsa 0.5 -15.2209 locatedin 0.5 -15.2209 convote, 0.5 -15.2209 sequitur, 0.5 -15.2209 blanc-f1: 0.5 -15.2209 18.77 0.5 -15.2209 (french, 0.5 -15.2209 58.63 0.5 -15.2209 22.93 0.5 -15.2209 mm, 0.5 -15.2209 88.97 0.5 -15.2209 (hot 0.5 -15.2209 (fern´andez 0.5 -15.2209 compromise: 0.5 -15.2209 ranlp 0.5 -15.2209 50.71 0.5 -15.2209 weka+fs 0.5 -15.2209 (1−α) 0.5 -15.2209 chomsky’s 0.5 -15.2209 68.75 0.5 -15.2209 steepness 0.5 -15.2209 deviations, 0.5 -15.2209 vnullv 0.5 -15.2209 sik, 0.5 -15.2209 fatally 0.5 -15.2209 26.93 0.5 -15.2209 (n=60) 0.5 -15.2209 s´anchez-marco 0.5 -15.2209 complementizer, 0.5 -15.2209 350k 0.285714 -15.221 5.3, 0.333333 -15.221 83.20 0.333333 -15.221 convergence: 0.333333 -15.221 cks 0.333333 -15.221 ndimensional 0.333333 -15.221 ifx 0.333333 -15.221 exempt 0.333333 -15.221 (moses) 0.333333 -15.221 southampton 0.2 -15.222 (martins 0.333333 -15.2227 29.58 0.333333 -15.2227 via: 0.333333 -15.2227 86.30 0.333333 -15.2227 σy 0.333333 -15.2227 (50%), 0.333333 -15.2227 -s 0.333333 -15.2227 74.75 0.333333 -15.2227 29th 0.5 -15.2236 4.0, 0.5 -15.2236 (leaf) 0.5 -15.2236 order-dependent 0.5 -15.2236 pattern/instance 0.5 -15.2236 ha-1 0.5 -15.2236 2009b, 0.5 -15.2236 0.963 0.5 -15.2236 xl, 0.5 -15.2236 context-insensitive 0.5 -15.2236 [second-pron] 0.5 -15.2236 46.06 0.5 -15.2236 mandalay 0.5 -15.2236 rei 0.5 -15.2236 optionality 0.5 -15.2236 stanley 0.5 -15.2236 stumble 0.5 -15.2236 16.13 0.5 -15.2236 unmanageable, 0.5 -15.2236 assignment) 0.5 -15.2236 aa; 0.5 -15.2236 (gasperin 0.5 -15.2236 euphemismand 0.5 -15.2236 (jenatton 0.5 -15.2236 15%, 0.5 -15.2236 19.2% 0.5 -15.2236 cfv=∞ 0.5 -15.2236 38.46 0.5 -15.2236 convey, 0.5 -15.2236 1146 0.5 -15.2236 intensifier, 0.5 -15.2236 0.942 0.5 -15.2236 ireland, 0.5 -15.2236 nullसह 0.5 -15.2236 dk: 0.5 -15.2236 .c0/jh, 0.5 -15.2236 (austin, 0.5 -15.2236 95.16 0.5 -15.2236 pi’s 0.5 -15.2236 (ja), 0.5 -15.2236 single-token 0.333333 -15.2245 query) 0.333333 -15.2245 rigau 0.333333 -15.2245 sincerely 0.333333 -15.2245 vectorβg 0.333333 -15.2245 null-word 0.333333 -15.2245 m3, 0.333333 -15.2263 neva 0.333333 -15.2263 38.39 0.333333 -15.2263 recognises 0.333333 -15.2263 a”, 0.333333 -15.2263 stresses 0.333333 -15.2263 cereus”, 0.333333 -15.2263 friesen, 0.5 -15.2263 501, 0.5 -15.2263 elderly man 0.5 -15.2263 garey 0.5 -15.2263 4,080 0.5 -15.2263 sign-language 0.5 -15.2263 dietterich 0.5 -15.2263 named, 0.5 -15.2263 1999b) 0.5 -15.2263 27.17 0.5 -15.2263 subsuming 0.5 -15.2263 persistently 0.5 -15.2263 32.45 0.5 -15.2263 awkwardness 0.5 -15.2263 1319 0.5 -15.2263 (c): 0.5 -15.2263 1480 0.5 -15.2263 35.71 0.5 -15.2263 5giza++ 0.5 -15.2263 681 0.5 -15.2263 6.1, 0.5 -15.2263 mlt 0.5 -15.2263 98.36 0.5 -15.2263 (levow 0.5 -15.2263 (lts-nolabels) 0.5 -15.2263 ?agent, 0.5 -15.2263 λlscript 0.5 -15.2263 haksayng-ul 0.5 -15.2263 scoreof 0.5 -15.2263 (asthma, 0.5 -15.2263 (points 0.5 -15.2263 (seconds) 0.5 -15.2263 mono-euro 0.5 -15.2263 nunberg 0.5 -15.2263 acl-08, 0.5 -15.2263 13:39:46 0.5 -15.2263 dönnes, 0.5 -15.2263 moiron 0.285714 -15.2271 robv 0.333333 -15.2281 square, 0.333333 -15.2281 h5: 0.333333 -15.2281 finnish), 0.0443038 -15.2282 al., 2009; 0.5 -15.2289 51.28 0.5 -15.2289 plaisant 0.5 -15.2289 [si, 0.5 -15.2289 1306 0.5 -15.2289 elapses 0.5 -15.2289 poly-co, 0.5 -15.2289 =〈k−1,s 0.5 -15.2289 18.36 0.5 -15.2289 imperfective, 0.5 -15.2289 category,and 0.5 -15.2289 spivey 0.5 -15.2289 ∈w 0.5 -15.2289 1·“sbk...sck”·s 0.5 -15.2289 (ringe 0.5 -15.2289 mtor 0.5 -15.2289 0.71, 0.5 -15.2289 directl+-style 0.5 -15.2289 dipped 0.5 -15.2289 9,984 0.5 -15.2289 circulated 0.5 -15.2289 mereologic 0.5 -15.2289 31st 0.5 -15.2289 10.16 0.5 -15.2289 nugget-based 0.5 -15.2289 (20% 0.5 -15.2289 ‘nivre’ 0.5 -15.2289 160000 0.5 -15.2289 baseline+lexical+ana 0.5 -15.2289 bigrams), 0.5 -15.2289 expedited 0.5 -15.2289 (so, 0.5 -15.2289 46.15 0.5 -15.2289 hildebrand 0.5 -15.2289 82.06 0.5 -15.2289 reordering-aspreprocessing 0.5 -15.2289 2english 0.5 -15.2289 commentator 0.5 -15.2289 nonincreasing 0.5 -15.2289 76.19 0.5 -15.2289 meteorological 0.5 -15.2289 gramar, 0.333333 -15.2299 maintain, 0.333333 -15.2299 b0 0.333333 -15.2299 54.28 0.333333 -15.2299 forgetting 0.333333 -15.2299 promoters, 0.333333 -15.2299 sl40 0.333333 -15.2299 quang 0.285714 -15.2302 un-observed 0.015186 -15.2305 shown 0.5 -15.2316 ˆ; 0.5 -15.2316 “/”) 0.5 -15.2316 §6 0.5 -15.2316 proposes, 0.5 -15.2316 17.01 0.5 -15.2316 clasifier) 0.5 -15.2316 27.03 0.5 -15.2316 +1.68 0.5 -15.2316 54.87 0.5 -15.2316 characterization, 0.5 -15.2316 figure5 0.5 -15.2316 tsvm-m 0.5 -15.2316 homophones: 0.5 -15.2316 rfe, 0.5 -15.2316 36.97 0.5 -15.2316 two-valued 0.5 -15.2316 readings: 0.5 -15.2316 boca 0.5 -15.2316 consolidates 0.5 -15.2316 inquire 0.5 -15.2316 5.08 0.5 -15.2316 upv-prhlt 0.5 -15.2316 texturing 0.5 -15.2316 meaningof 0.5 -15.2316 million)) 0.5 -15.2316 14.82 0.5 -15.2316 measurepos 0.5 -15.2316 34.54 0.5 -15.2316 winner” 0.5 -15.2316 [slot1,slot2,slot3], 0.5 -15.2316 preparenthesis 0.5 -15.2316 berlin/new 0.5 -15.2316 /i/, 0.5 -15.2316 (propbank) 0.5 -15.2316 4(c)), 0.333333 -15.2317 phrases  0.333333 -15.2317 lgs 0.333333 -15.2317 0.303 0.333333 -15.2317 λmax(s) 0.333333 -15.2317 12k 0.333333 -15.2317 gmentation 0.333333 -15.2317 76.27 0.333333 -15.2317 addressee, 0.333333 -15.2317 oc, 0.333333 -15.2317 pitching 0.333333 -15.2334 bantu 0.333333 -15.2334 56.52 0.333333 -15.2334 poslm 0.333333 -15.2334 economists 0.333333 -15.2334 „is 0.176471 -15.2336 sβ> 0.5 -15.2343 67.95 0.5 -15.2343 2994 0.5 -15.2343 mwi 0.5 -15.2343 (dk) 0.5 -15.2343 27.77 0.5 -15.2343 30.01 0.5 -15.2343 1975, 0.5 -15.2343 23.25 0.5 -15.2343 universit¨at stuttgart, 0.5 -15.2343 manifesting 0.5 -15.2343 linguistic-oriented 0.5 -15.2343 mpalmer@colorado.edu 0.5 -15.2343 hindi-tourism 0.5 -15.2343 32.52 0.5 -15.2343 1256 0.5 -15.2343 expl) 0.5 -15.2343 25.26 0.5 -15.2343 0.84, 0.5 -15.2343 prior:smoothed 0.5 -15.2343 (come) 0.5 -15.2343 ignore, 0.5 -15.2343 8-gram 0.5 -15.2343 23.95 0.5 -15.2343 authorityclaim 0.5 -15.2343 whnp, 0.5 -15.2343 ofgprime 0.5 -15.2343 c1) 0.5 -15.2343 alrand, 0.5 -15.2343 ‘all 0.5 -15.2343 66.11 0.5 -15.2343 58.15 0.5 -15.2343 bottom-left 0.5 -15.2343 26.94 0.5 -15.2343 “tell 0.5 -15.2343 heuristic”, 0.5 -15.2343 dk-2300 0.5 -15.2343 1041 0.5 -15.2343 tpl+-formulas 0.5 -15.2343 cut, 0.5 -15.2343 1−p 0.285714 -15.2348 mismatch: 0.333333 -15.2352 2:1 0.333333 -15.2352 λ-productions 0.333333 -15.2352 (ma) 0.333333 -15.2352 annotationmatcher 0.333333 -15.2352 rocking 0.333333 -15.2352 1.3/1 0.333333 -15.2352 sub-items 0.333333 -15.2352 “microsoft” 0.333333 -15.2352 “classic” 0.157895 -15.2359 returned. 0.333333 -15.237 kn, 0.333333 -15.237 slices, 0.333333 -15.237 72.45 0.333333 -15.237 38.48 0.333333 -15.237 n1,n2 0.333333 -15.237 pos1 0.333333 -15.237 furthermore,the 0.333333 -15.237 1583 0.333333 -15.237 undersegmentation 0.5 -15.237 1619 0.5 -15.237 butt 0.5 -15.237 data’s 0.5 -15.237 synchronizing 0.5 -15.237 m∈m 0.5 -15.237 inference.5 0.5 -15.237 naacl2009, 0.5 -15.237 genera, 0.5 -15.237 chinese-arabic 0.5 -15.237 behavioral, 0.5 -15.237 theoptimal 0.5 -15.237 59.39 0.5 -15.237 part–of–speech 0.5 -15.237 3,802 0.5 -15.237 generics, 0.5 -15.237 hyphenation 0.5 -15.237 n-gram/m-slm 0.5 -15.237 text+, 0.5 -15.237 769 0.5 -15.237 mechanic, 0.5 -15.237 isca 0.5 -15.237 p(i) 0.5 -15.237 expletives 0.5 -15.237 path-finding, 0.5 -15.237 x5650 0.5 -15.237 ideograms 0.5 -15.237 information-extraction 0.5 -15.237 ζ(z), 0.5 -15.237 φl 0.5 -15.237 surgical 0.5 -15.237 14.42% 0.5 -15.237 i5 0.5 -15.237 bcb 0.2 -15.2381 ov 0.0795455 -15.2386 far, 0.333333 -15.2388 (danescu-niculescu-mizil 0.333333 -15.2388 domain–specific 0.333333 -15.2388 (denoting 0.333333 -15.2388 replicative 0.333333 -15.2388 36.17 0.333333 -15.2388 t∈q 0.333333 -15.2388 ‘we 0.333333 -15.2388 ^vb, 0.333333 -15.2388 york” 0.333333 -15.2388 ‘s’ 0.333333 -15.2388 hsu, 0.0606061 -15.2392 acknowledgements 0.0909091 -15.2395 concept. 0.5 -15.2397 interactional, 0.5 -15.2397 ags 0.5 -15.2397 (301) 0.5 -15.2397 arive 0.5 -15.2397 rela 0.5 -15.2397 associativity: 0.5 -15.2397 r-precision 0.5 -15.2397 lansing, michigan, 0.5 -15.2397 dico. 0.5 -15.2397 sumary 0.5 -15.2397 conditionalprobability 0.5 -15.2397 32.29 0.5 -15.2397 20.97 0.5 -15.2397 92.17 0.5 -15.2397 1352 0.5 -15.2397 n[pi]∈ne 0.5 -15.2397 san−francisco 0.5 -15.2397 scoring; 0.5 -15.2397 nv, 0.5 -15.2397 (epsilon1) 0.5 -15.2397 27.82 0.5 -15.2397 pt,l 0.5 -15.2397 ˆα 0.5 -15.2397 deletion) 0.5 -15.2397 q(θna) 0.5 -15.2397 (weak 0.5 -15.2397 84.27 0.5 -15.2397 misinterprets 0.5 -15.2397 continuously, 0.5 -15.2397 endeavors 0.5 -15.2397 saw1,8 0.5 -15.2397 g∈d(g 0.5 -15.2397 xt1(r2 0.333333 -15.2407 (q1 0.333333 -15.2407 bleu_d 0.333333 -15.2407 0.094 0.333333 -15.2407 91.39 0.333333 -15.2407 x+1 0.333333 -15.2407 allotted 0.333333 -15.2407 literature), 0.333333 -15.2407 standard), 0.25 -15.2411 (category 0.5 -15.2424 19.45 0.5 -15.2424 jefferson 0.5 -15.2424 (sampson, 0.5 -15.2424 gdlm 0.5 -15.2424 use) 0.5 -15.2424 dordrecht, 0.5 -15.2424 klenner, 0.5 -15.2424 3:6b 0.5 -15.2424 scipy 0.5 -15.2424 -1.13 0.5 -15.2424 intervene 0.5 -15.2424 ptime 0.5 -15.2424 2bedroom 0.5 -15.2424 location-aware 0.5 -15.2424 1945; 0.5 -15.2424 (ranlp), 0.5 -15.2424 jensen 0.5 -15.2424 0.190 0.5 -15.2424 author”. 0.5 -15.2424 illness, 0.5 -15.2424 (pearson 0.5 -15.2424 ta: 0.5 -15.2424 u5 0.5 -15.2424 big-jump 0.5 -15.2424 1487 0.5 -15.2424 possibleto 0.5 -15.2424 (73% 0.5 -15.2424 3.67 0.5 -15.2424 awful 0.5 -15.2424 daalaa (put) 0.5 -15.2424 3as 0.5 -15.2424 (maxent), 0.5 -15.2424 soon” 0.5 -15.2424 place: 0.5 -15.2424 dp-tree 0.5 -15.2424 three) 0.5 -15.2424 (date, 0.5 -15.2424 warrior 0.5 -15.2424 crawling, 0.5 -15.2424 754 0.5 -15.2424 30.34 0.5 -15.2424 predict: 0.5 -15.2424 complex! 0.5 -15.2424 suffixes: 0.5 -15.2424 split) 0.5 -15.2424 tissue, 0.5 -15.2424 16.71 0.5 -15.2424 context-independent 0.25 -15.2425 rc 0.333333 -15.2425 influenceof 0.333333 -15.2425 survives 0.333333 -15.2425 tf/idf 0.333333 -15.2425 ous 0.333333 -15.2425 43.52 0.333333 -15.2425 thel-th 0.0701754 -15.2429 consisted 0.157895 -15.2436 y” 0.0380048 -15.2436 on this 0.148148 -15.244 :: 0.333333 -15.2443 n}, 0.333333 -15.2443 compr 0.333333 -15.2443 705 0.333333 -15.2443 0.333333 -15.2443 toyota 0.333333 -15.2443 708 0.0952381 -15.2451 noting that 0.5 -15.2451 constellation 0.5 -15.2451 22.2% 0.5 -15.2451 6.88 0.5 -15.2451 explicate 0.5 -15.2451 0.589 0.5 -15.2451 wright, 0.5 -15.2451 unknown), 0.5 -15.2451 “put 0.5 -15.2451 at-t 0.5 -15.2451 1043 0.5 -15.2451 ngram, 0.5 -15.2451 (8gb 0.5 -15.2451 83.3% 0.5 -15.2451 {, 0.5 -15.2451 40.85 0.5 -15.2451 15.45 0.5 -15.2451 ∆ptk(n1,n2) 0.5 -15.2451 81.91% 0.5 -15.2451 initiation 0.5 -15.2451 (cases 0.5 -15.2451 z1,4i 0.5 -15.2451 ep∈epc 0.5 -15.2451 estimation: 0.5 -15.2451 (10)) 0.5 -15.2451 (f(qak))(s,z)] 0.5 -15.2451 non-property 0.5 -15.2451 scopedisambiguated 0.5 -15.2451 auxk 0.5 -15.2451 34.1% 0.5 -15.2451 topicst 0.5 -15.2451 +f 0.5 -15.2451 39.10 0.5 -15.2451 lead, 0.5 -15.2451 outer-most 0.5 -15.2451 cake#3 0.5 -15.2451 drifts 0.5 -15.2451 ¬ 0.5 -15.2451 8.62 0.5 -15.2451 (paper 0.5 -15.2451 blocks: 0.5 -15.2451 1.6m 0.116279 -15.2453 i would 0.121951 -15.2456 dealt 0.333333 -15.2461 enclitic 0.333333 -15.2461 95.97 0.333333 -15.2461 tweet-based 0.333333 -15.2461 (dtks) 0.333333 -15.2461 courtesy 0.333333 -15.2461 “ein 0.333333 -15.2461 song” 0.333333 -15.2461 16.56 0.333333 -15.2461 season, 0.333333 -15.2461 gdlm, 0.333333 -15.2461 31.55 0.119048 -15.2461 resolution system 0.0695652 -15.2478 will use 0.1875 -15.2479 gap. 0.333333 -15.2479 86.88 0.333333 -15.2479 20.59 0.333333 -15.2479 hashtags: 0.333333 -15.2479 requested, 0.333333 -15.2479 76.9% 0.333333 -15.2479 use(during(e), 0.333333 -15.2479 analytically 0.333333 -15.2479 ˜y 0.5 -15.2479 inspiredby 0.5 -15.2479 (teacher 0.5 -15.2479 parent), 0.5 -15.2479 eprime, 0.5 -15.2479 fight, 0.5 -15.2479 implementational 0.5 -15.2479 c-arg, 0.5 -15.2479 greaterorequalslant 0.5 -15.2479 figure4 0.5 -15.2479 driven, 0.5 -15.2479 (unlexicalized) 0.5 -15.2479 viv 0.5 -15.2479 outedges 0.5 -15.2479 crispy 0.5 -15.2479 yprimeprime 0.5 -15.2479 topscoring 0.5 -15.2479 12.91 0.5 -15.2479 argm-ext 0.5 -15.2479 “none 0.5 -15.2479 mari¨et 0.5 -15.2479 (mccc-i 0.5 -15.2479 substantiated 0.5 -15.2479 36.60 0.5 -15.2479 inf-srf 0.5 -15.2479 45.58 0.5 -15.2479 1077 0.5 -15.2479 90.30 0.5 -15.2479 s5, 0.5 -15.2479 45.51 0.5 -15.2479 usp14 0.5 -15.2479 nullकाश 0.0332717 -15.2481 want 0.285714 -15.2487 (15), 0.333333 -15.2497 9all 0.333333 -15.2497 publisher 0.333333 -15.2497 cj,k 0.333333 -15.2497 company), 0.333333 -15.2497 away; 0.333333 -15.2497 evidentials 0.285714 -15.2503 county* 0.5 -15.2506 55.90 0.5 -15.2506 array-typed 0.5 -15.2506 1560 0.5 -15.2506 nlp) 0.5 -15.2506 **department 0.5 -15.2506 project∗ 0.5 -15.2506 90.16 0.5 -15.2506 (query, 0.5 -15.2506 python’s 0.5 -15.2506 compliment, 0.5 -15.2506 turk2 0.5 -15.2506 organizations), 0.5 -15.2506 mauro 0.5 -15.2506 conferences2 0.5 -15.2506 2681 0.5 -15.2506 (weissenbacher, 0.5 -15.2506 łpa 0.5 -15.2506 third-level 0.5 -15.2506 ¯q 0.5 -15.2506 terminal-aware 0.5 -15.2506 52.10 0.5 -15.2506 bntaml3 0.5 -15.2506 hiero0(s,σ,a) 0.5 -15.2506 consortium (ldc). 0.5 -15.2506 laughs, 0.5 -15.2506 visualiser 0.5 -15.2506 sys16 0.5 -15.2506 usa: 0.5 -15.2506 1208 0.5 -15.2506 limestone, 0.5 -15.2506 mwcl 0.5 -15.2506 ibm1sh 0.5 -15.2506 (counts 0.5 -15.2506 o(r), 0.5 -15.2506 patwardhan, 0.5 -15.2506 newssyscombtest2010 0.5 -15.2506 together) 0.5 -15.2506 progressing 0.5 -15.2506 〈s,t〉∈t 0.5 -15.2506 domain), 0.5 -15.2506 2,411 0.5 -15.2506 (beginning, 0.5 -15.2506 semeval-1 0.5 -15.2506 91.42 0.333333 -15.2516 (1970) 0.333333 -15.2516 59.23 0.333333 -15.2516 bisection 0.333333 -15.2516 41.46 0.333333 -15.2516 70.10 0.333333 -15.2516 facility, 0.333333 -15.2516 zum 0.333333 -15.2516 50.68 0.333333 -15.2516 sittichai 0.333333 -15.2516 polygons 0.333333 -15.2516 acquisitions, 0.333333 -15.2516 modelis 0.333333 -15.2534 nf-kappa 0.333333 -15.2534 (unsel) 0.333333 -15.2534 (s1 0.333333 -15.2534 wmt10, 0.333333 -15.2534 hurricane 0.333333 -15.2534 mmr-based 0.333333 -15.2534 gud 0.333333 -15.2534 “impact 0.333333 -15.2534 hard-disk 0.333333 -15.2534 component-whole, 0.5 -15.2534 (“sup”) 0.5 -15.2534 mx→f(x) 0.5 -15.2534 ∗now 0.5 -15.2534 94.34 0.5 -15.2534 e-rating 0.5 -15.2534 acl-coling 0.5 -15.2534 adaptation: 0.5 -15.2534 orient 0.5 -15.2534 mobile, 0.5 -15.2534 0.916 0.5 -15.2534 “the-bank 0.5 -15.2534 0.0161 0.5 -15.2534 multi-clas 0.5 -15.2534 memo, 0.5 -15.2534 rotation, 0.5 -15.2534 (gerner 0.5 -15.2534 lawyer, 0.5 -15.2534 87.16 0.5 -15.2534 moods, 0.5 -15.2534 al-grby-p” 0.5 -15.2534 euv 0.5 -15.2534 75.71 0.5 -15.2534 scot, 0.5 -15.2534 mknlm, 0.5 -15.2534 38.59 0.5 -15.2534 26.49 0.5 -15.2534 n= 0.5 -15.2534 tei, 0.5 -15.2534 reflexives, 0.5 -15.2534 tainan, 0.5 -15.2534 (nasr 0.5 -15.2534 reinvent 0.5 -15.2534 cold) 0.5 -15.2534 symmetrize 0.5 -15.2534 65.52 0.5 -15.2534 inn 0.5 -15.2534 20.41 0.5 -15.2534 word-align 0.137931 -15.2545 this demonstrates 0.176471 -15.2547 such pairs 0.285714 -15.255 randomly extracted 0.333333 -15.2552 has-a 0.333333 -15.2552 priors: 0.333333 -15.2552 don't 0.333333 -15.2552 import, 0.333333 -15.2552 (classes 0.333333 -15.2552 joachim 0.5 -15.2561 t 0.5 -15.2561 r,s 0.5 -15.2561 vectorci 0.5 -15.2561 non-sequential 0.5 -15.2561 nf,e 0.5 -15.2561 clayton, 0.5 -15.2561 (65 0.5 -15.2561 22.59 0.5 -15.2561 ida/center 0.5 -15.2561 0.69, 0.5 -15.2561 f1(m,2t) 0.5 -15.2561 prototypes, 0.5 -15.2561 kma 0.5 -15.2561 word–synonym 0.5 -15.2561 svd2 0.5 -15.2561 configurability, 0.5 -15.2561 chipman, 0.5 -15.2561 lms) 0.5 -15.2561 sented 0.5 -15.2561 mgm 0.5 -15.2561 83.15 0.5 -15.2561 concept) 0.5 -15.2561 12.63 0.5 -15.2561 20.80 0.5 -15.2561 milky 0.5 -15.2561 di∈d 0.5 -15.2561 ocurence 0.5 -15.2561 64.10 0.5 -15.2561 p(p) 0.5 -15.2561 948 0.5 -15.2561 62.99 0.5 -15.2561 svg 0.5 -15.2561 irregularly 0.5 -15.2561 decision making, 0.5 -15.2561 output’s 0.5 -15.2561 (segmentation, 0.5 -15.2561 personne 0.5 -15.2561 subsets: 0.5 -15.2561 relevent 0.5 -15.2561 89.23 0.5 -15.2561 s[wq]/(s[dcl]\np) 0.5 -15.2561 i-3 0.5 -15.2561 greenbacker, 0.5 -15.2561 looting 0.5 -15.2561 timeconsuming, 0.5 -15.2561 10.6m 0.5 -15.2561 69.80 0.285714 -15.2566 r4) 0.333333 -15.2571 (because, 0.333333 -15.2571 p= 0.333333 -15.2571 solar 0.333333 -15.2571 rapper 0.333333 -15.2571 53.55 0.333333 -15.2571 lateral 0.0568182 -15.2572 had a 0.157895 -15.2574 “michael 0.285714 -15.2582 (w/ 0.222222 -15.2583 trp 0.157895 -15.2583 source text. 0.333333 -15.2589 (hi) 0.333333 -15.2589 diabetic 0.333333 -15.2589 #3: 0.333333 -15.2589 52.46 0.333333 -15.2589 2-factor 0.333333 -15.2589 ϖk 0.333333 -15.2589 boldrini 0.5 -15.2589 fvk 0.5 -15.2589 74.47 0.5 -15.2589 38.75 0.5 -15.2589 _name1_ 0.5 -15.2589 29.39 0.5 -15.2589 fly” 0.5 -15.2589 (wsj), 0.5 -15.2589 entity-level 0.5 -15.2589 (4.1) 0.5 -15.2589 naacl/hlt 0.5 -15.2589 heuristiccost(ι) 0.5 -15.2589 exp3: 0.5 -15.2589 0.3)} 0.5 -15.2589 0.198 0.5 -15.2589 it 's 0.5 -15.2589 922 0.5 -15.2589 ecu, 0.5 -15.2589 2.66% 0.5 -15.2589 score(ti,j) 0.5 -15.2589 〈i,j〉, 0.5 -15.2589 negators, 0.5 -15.2589 egr-2 0.5 -15.2589 spv 0.5 -15.2589 m= 0.5 -15.2589 segmentations: 0.5 -15.2589 min(x,a) 0.5 -15.2589 stomach 0.5 -15.2589 entendres 0.5 -15.2589 tucson 0.5 -15.2589 cbs 0.5 -15.2589 (lan 0.5 -15.2589 obj-bd 0.5 -15.2589 mondal, 0.5 -15.2589 (7.7) 0.5 -15.2589 bigrams; 0.5 -15.2589 (ozgur 0.5 -15.2589 ceaf(e). 0.5 -15.2589 a(s) 0.5 -15.2589 1514 0.5 -15.2589 translators’ 0.5 -15.2589 (say) 0.5 -15.2589 ciravegna 0.5 -15.2589 updates: 0.0237643 -15.2594 system. 0.333333 -15.2607 bryant 0.333333 -15.2607 (than 0.333333 -15.2607 57.59 0.333333 -15.2607 +(1 0.285714 -15.2613 mods 0.5 -15.2616 49.51 0.5 -15.2616 35.30 0.5 -15.2616 (wmt10) 0.5 -15.2616 (semi-) 0.5 -15.2616 virtues 0.5 -15.2616 “cosavg” 0.5 -15.2616 entradas 0.5 -15.2616 synsets; 0.5 -15.2616 tokens’ 0.5 -15.2616 soft-clustering 0.5 -15.2616 loss; 0.5 -15.2616 hi,n 0.5 -15.2616 alliance 0.5 -15.2616 (lung 0.5 -15.2616 57.24 0.5 -15.2616 vertex: 0.5 -15.2616 honor, 0.5 -15.2616 inverse+ 0.5 -15.2616 79.33 0.5 -15.2616 heretofore 0.5 -15.2616 prosody-based 0.5 -15.2616 low-weight 0.5 -15.2616 619 0.5 -15.2616 interactions) 0.5 -15.2616 q-elaba(pi1,pi2) 0.5 -15.2616 ever-growing 0.5 -15.2616 “#basketball”, 0.5 -15.2616 wordlength 0.5 -15.2616 transcript, 0.5 -15.2616 steffen 0.5 -15.2616 cpa 0.5 -15.2616 baseline+ana 0.5 -15.2616 3.30 0.5 -15.2616 1568 0.5 -15.2616 “of” 0.5 -15.2616 76.60 0.5 -15.2616 1393 0.5 -15.2616 71.15 0.5 -15.2616 folk 0.5 -15.2616 alawst 0.5 -15.2616 matrix-vector 0.5 -15.2616 8our 0.5 -15.2616 63.77 0.5 -15.2616 hybridisation 0.5 -15.2616 flour 0.5 -15.2616 marcu's 0.5 -15.2616 (rudner 0.5 -15.2616 fpu 0.333333 -15.2626 0.25% 0.333333 -15.2626 16.28 0.333333 -15.2626 2.4m 0.333333 -15.2626 rabbit 0.333333 -15.2626 2university 0.333333 -15.2626 inference), 0.285714 -15.2629 winter 0.333333 -15.2644 20040060080010000 0.333333 -15.2644 investigate, 0.333333 -15.2644 at) 0.333333 -15.2644 controller 0.333333 -15.2644 (kl: 0.333333 -15.2644 yarowsky’s 0.333333 -15.2644 (name, 0.333333 -15.2644 (eck 0.5 -15.2644 maker, 0.5 -15.2644 pragmatic, 0.5 -15.2644 17.67 0.5 -15.2644 73.57 0.5 -15.2644 london/new 0.5 -15.2644 (dynamic 0.5 -15.2644 [w1,w2] 0.5 -15.2644 chart[0, 0.5 -15.2644 symmetrised 0.5 -15.2644 replicas 0.5 -15.2644 spliter, 0.5 -15.2644 tau) 0.5 -15.2644 33.26 0.5 -15.2644 15.12 0.5 -15.2644 totaled 0.5 -15.2644 92.50 0.5 -15.2644 28.82 0.5 -15.2644 ˆw, 0.5 -15.2644 23.57 0.5 -15.2644 complement), 0.5 -15.2644 e1...en, 0.5 -15.2644 49.39 0.5 -15.2644 heiskell’s 0.5 -15.2644 (belief 0.5 -15.2644 pas, 0.5 -15.2644 94.5%, 0.5 -15.2644 catalogs 0.5 -15.2644 (mr), 0.5 -15.2644 merge-split 0.5 -15.2644 @null, 0.5 -15.2644 universität 0.5 -15.2644 eq, 0.5 -15.2644 harald 0.5 -15.2644 (33% 0.5 -15.2644 reflect, 0.5 -15.2644 people), 0.5 -15.2644 internet-based 0.5 -15.2644 90.54 0.5 -15.2644 f(c, t) 0.285714 -15.2645 mohit 0.25 -15.2645 acids 0.333333 -15.2663 39.02 0.333333 -15.2663 clog 0.333333 -15.2663 -1.85 0.333333 -15.2663 flattening 0.333333 -15.2663 554 0.333333 -15.2663 4,4quality 0.333333 -15.2663 33.04 0.333333 -15.2663 (gender 0.333333 -15.2663 slat 0.5 -15.2672 7.27 0.5 -15.2672 6371 0.5 -15.2672 100-best list. 0.5 -15.2672 srinivas 0.5 -15.2672 give) 0.5 -15.2672 aer(a,s,p) 0.5 -15.2672 “low 0.5 -15.2672 circle, 0.5 -15.2672 35.42 0.5 -15.2672 〈i〉 0.5 -15.2672 arzucan 0.5 -15.2672 1986); 0.5 -15.2672 93.11 0.5 -15.2672 55.8% 0.5 -15.2672 18.76 0.5 -15.2672 16.92% 0.5 -15.2672 trickier 0.5 -15.2672 t-truncated 0.5 -15.2672 2n−2 0.5 -15.2672 14.60 0.5 -15.2672 fs1 0.5 -15.2672 (arg1) 0.5 -15.2672 ay 0.5 -15.2672 alamlyt” 0.5 -15.2672 gerold 0.5 -15.2672 21.69 0.5 -15.2672 62.76 0.5 -15.2672 sarcasm: 0.5 -15.2672 60.50 0.5 -15.2672 real ‘ground 0.5 -15.2672 (spanned) 0.5 -15.2672 7.06 0.5 -15.2672 82.46 0.5 -15.2672 94.05 0.285714 -15.2677 saj10a 0.333333 -15.2682 system1 0.333333 -15.2682 “objective” 0.333333 -15.2682 interleaved-expected-reconstruction-error-0.2interleaved-expected-reconstruction-error-0.4 interleaved-expected-reconstruction-error-0.6interleaved-expected-reconstruction-error-0.8 0.333333 -15.2682 p(zj 0.333333 -15.2682 en→cz 0.333333 -15.2682 s˜ao 0.333333 -15.2682 estimator, 0.121212 -15.2685  0.333333 -15.27 haben 0.333333 -15.27 66.03 0.333333 -15.27 “object 0.333333 -15.27 gold) 0.333333 -15.27 06 0.333333 -15.27 ()hz 0.333333 -15.27 “路滑 0.333333 -15.27 sanjeev 0.333333 -15.27 “green 0.5 -15.27 1* 0.5 -15.27 boarding 0.5 -15.27 29.34 0.5 -15.27 2,038 0.5 -15.27 46.70 0.5 -15.27 (moving 0.5 -15.27 restrictionson 0.5 -15.27 (tsa-a-te, 0.5 -15.27 0.173 0.5 -15.27 d50, 0.5 -15.27 switchboard) 0.5 -15.27 slides, 0.5 -15.27 timesz 0.5 -15.27 1s, 0.5 -15.27 06.5 0.5 -15.27 donthave_corr 0.5 -15.27 second) 0.5 -15.27 mouth, 0.5 -15.27 pair-specific 0.5 -15.27 foreign, 0.5 -15.27 infectiousdisease 0.5 -15.27 roukos 0.5 -15.27 percolate 0.5 -15.27 ϕsa 0.5 -15.27 adpositions 0.5 -15.27 best-known 0.5 -15.27 [tl]: 0.5 -15.27 major-league 0.5 -15.27 policies: 0.5 -15.27 (prakā 0.5 -15.27 irst2 0.157895 -15.2715 layer. 0.333333 -15.2719 chasen 0.333333 -15.2719 ]s2 0.333333 -15.2719 very, 0.333333 -15.2719 (daume 0.333333 -15.2719 calzolari, 0.333333 -15.2719 argues, 0.5 -15.2728 scorew−prosody(w) 0.5 -15.2728 (fresh 0.5 -15.2728 he entered 0.5 -15.2728 ju, 0.5 -15.2728 sweeps 0.5 -15.2728 12.81 0.5 -15.2728 zarrieß, 0.5 -15.2728 (covered 0.5 -15.2728 (know 0.5 -15.2728 hand-written, 0.5 -15.2728 80.2% 0.5 -15.2728 tales: 0.5 -15.2728 veletsianos, 0.5 -15.2728 h′ 0.5 -15.2728 n′, 0.5 -15.2728 >5 0.5 -15.2728 reflex 0.5 -15.2728 cimino, 0.5 -15.2728 tagset: 0.5 -15.2728 aggregation: 0.5 -15.2728 possessives: 0.5 -15.2728 leave, 0.5 -15.2728 consider: 0.5 -15.2728 100# 0.5 -15.2728 35.40 0.5 -15.2728 gundel 0.5 -15.2728 95.07 0.5 -15.2728 liu liu, 0.5 -15.2728 43.39 0.5 -15.2728 as), 0.5 -15.2728 s∈si 0.5 -15.2728 word-based) 0.5 -15.2728 contdep 0.5 -15.2728 topic-oriented 0.5 -15.2728 countering 0.5 -15.2728 en-cz, 0.5 -15.2728 verifiable 0.5 -15.2728 734 0.5 -15.2728 kai 0.5 -15.2728 agar, 0.5 -15.2728 nagao, 0.5 -15.2728 mcgill 0.5 -15.2728 sentence-segmented 0.5 -15.2728 better), 0.5 -15.2728 epfl 0.5 -15.2728 1865 0.5 -15.2728 0.00%) 0.5 -15.2728 rebuild 0.5 -15.2728 symbols), 0.5 -15.2728 meta-tag 0.333333 -15.2737 55.87 0.333333 -15.2737 ‰ 0.333333 -15.2737 527 0.333333 -15.2737 systemfor 0.333333 -15.2737 best-case 0.333333 -15.2737 cloning 0.333333 -15.2737 comb 0.333333 -15.2737 subrequests: 0.25 -15.2742 idss 0.25 -15.2742 (root 0.333333 -15.2756 ceaf-m 0.333333 -15.2756 parenleftbiggn 0.333333 -15.2756 52.11 0.333333 -15.2756 (eg, 0.333333 -15.2756 (29 0.333333 -15.2756 kbracehtipdownleft 0.333333 -15.2756 1,149 0.5 -15.2756 eighteenth 0.5 -15.2756 say: 0.5 -15.2756 st_ir 0.5 -15.2756 ‘k’ 0.5 -15.2756 49.71 0.5 -15.2756 (y, 0.5 -15.2756 srivastava, 0.5 -15.2756 dynasty 0.5 -15.2756 horizontal, 0.5 -15.2756 coling2010, 0.5 -15.2756 rq2 0.5 -15.2756 r1: 0.5 -15.2756 minimizer 0.5 -15.2756 →sc 0.5 -15.2756 1510 0.5 -15.2756 92.69 0.5 -15.2756 62.25 0.5 -15.2756 40.05 0.5 -15.2756 porter, 0.5 -15.2756 non-ambiguity 0.5 -15.2756 26.90 0.5 -15.2756 hand-created 0.5 -15.2756 (chawla, 0.5 -15.2756 52.7% 0.5 -15.2756 i1(x;y;z) 0.5 -15.2756 comfortably 0.5 -15.2756 (sentence, 0.5 -15.2756 quoc 0.5 -15.2756 44.51 0.5 -15.2756 exists), 0.5 -15.2756 60.24 0.5 -15.2756 (syed 0.5 -15.2756 (epi), 0.5 -15.2756 39.88 0.5 -15.2756 grids, 0.5 -15.2756 5ldc 0.5 -15.2756 20.12 0.5 -15.2756 822 0.5 -15.2756 (70%), 0.5 -15.2756 hd-comp-ph 0.5 -15.2756 kinase-26) 0.5 -15.2756 (np-sbj 0.5 -15.2756 bowyer 0.0244898 -15.2758 corpus, 0.333333 -15.2775 0.2; 0.333333 -15.2775 z∗(j)qh 0.333333 -15.2775 emotional, 0.333333 -15.2775 2−3 0.333333 -15.2775 +βli 0.333333 -15.2775 howard 0.333333 -15.2775 (ranked 0.333333 -15.2775 disconfirmed 0.0813953 -15.2778 ) = 0.5 -15.2784 tiles, 0.5 -15.2784 hesitation, 0.5 -15.2784 95.8% 0.5 -15.2784 06] 0.5 -15.2784 nonparagraph 0.5 -15.2784 (chazelle 0.5 -15.2784 anc1(r) 0.5 -15.2784 inflates 0.5 -15.2784 (kw) 0.5 -15.2784 dition 0.5 -15.2784 “poor”) 0.5 -15.2784 59.40 0.5 -15.2784 (es): 0.5 -15.2784 �” (tsune) 0.5 -15.2784 signficant 0.5 -15.2784 oni 0.5 -15.2784 50.27 0.5 -15.2784 wolter 0.5 -15.2784 0.19384 0.5 -15.2784 bitters 0.5 -15.2784 16.68 0.5 -15.2784 virb 0.5 -15.2784 subtrees: 0.5 -15.2784 comparisonof 0.5 -15.2784 (96 0.5 -15.2784 1d-mert, 0.5 -15.2784 62.46 0.5 -15.2784 also: 0.5 -15.2784 (object), 0.5 -15.2784 1-2, 0.5 -15.2784 hlt/emnlp 0.5 -15.2784 23.93 0.5 -15.2784 nounadjective 0.5 -15.2784 10−2 0.5 -15.2784 11a 0.5 -15.2784 othe 0.5 -15.2784 i‟m 0.5 -15.2784 impresion 0.5 -15.2784 ‘noun’ 0.5 -15.2784 transformational 0.5 -15.2784 avoidance 0.5 -15.2784 whilethe 0.5 -15.2784 (41%) 0.5 -15.2784 “floating 0.5 -15.2784 [contrast] 0.5 -15.2784 (ami) 0.5 -15.2784 develop, 0.5 -15.2784 dare 0.5 -15.2784 non-theme 0.5 -15.2784 pharmacology, 0.5 -15.2784 doubly-anchored 0.285714 -15.2789 grimm 0.333333 -15.2794 ghose 0.333333 -15.2794 efficient: 0.333333 -15.2794 “role 0.333333 -15.2794 str+met+lex2+ 0.333333 -15.2794 airfrance 0.333333 -15.2794 ‘dog’ 0.333333 -15.2794 impossible: 0.333333 -15.2794 answer’s 0.333333 -15.2794 half) 0.333333 -15.2794 pre-modifier 0.333333 -15.2794 [`k]φ0 0.285714 -15.2805 ,, 0.084507 -15.2805 dominated 0.222222 -15.2807 anne 0.333333 -15.2812 1,2, 0.333333 -15.2812 tial 0.333333 -15.2812 betterthan 0.333333 -15.2812 gilholy 0.333333 -15.2812 emp-org 0.333333 -15.2812 0.698 0.333333 -15.2812 subject(usp14) 0.333333 -15.2812 (add 0.333333 -15.2812 slice, 0.333333 -15.2812 gain: 0.333333 -15.2812 “block” 0.5 -15.2812 presentsthe 0.5 -15.2812 entity (gpe) 0.5 -15.2812 (hutchins 0.5 -15.2812 posts1 0.5 -15.2812 72.22 0.5 -15.2812 comparative/superlative 0.5 -15.2812 adjectivenoun, 0.5 -15.2812 pitcher 0.5 -15.2812 69.47 0.5 -15.2812 (margin) 0.5 -15.2812 nonconjugacy 0.5 -15.2812 886 0.5 -15.2812 miic 0.5 -15.2812 speeral 0.5 -15.2812 ically 0.5 -15.2812 30.49 0.5 -15.2812 “recall 0.5 -15.2812 amount” 0.5 -15.2812 ˜zl,θ(x) 0.5 -15.2812 similes, 0.5 -15.2812 (equivalent 0.5 -15.2812 1999)), 0.5 -15.2812 non-task-oriented 0.5 -15.2812 he-by 0.5 -15.2812 envelopes 0.5 -15.2812 19.02 0.5 -15.2812 recall-precision 0.5 -15.2812 (efl) 0.5 -15.2812 5-grams) 0.0502283 -15.2813 where each 0.25 -15.2813 passvp(expressed) 0.333333 -15.2831 _ yangtze 0.333333 -15.2831 madison 0.333333 -15.2831 mass, 0.333333 -15.2831 i± 0.333333 -15.2831 x〈c,dr〉 0.333333 -15.2831 “trending 0.333333 -15.2831 styles: 0.333333 -15.2831 oovs), 0.333333 -15.2831 grip 0.333333 -15.2831 neumann 0.333333 -15.2831 8.02 0.1875 -15.2834 (bojar 0.5 -15.284 decf1 0.5 -15.284 h-t 0.5 -15.284 51.48 0.5 -15.284 eachti 0.5 -15.284 (tiedemann, 0.5 -15.284 speled 0.5 -15.284 70.45 0.5 -15.284 resource3 0.5 -15.284 accuracy2 0.5 -15.284 anyone, 0.5 -15.284 restates 0.5 -15.284 emotion-bearing 0.5 -15.284 ngrami; 0.5 -15.284 (a1a2) 0.5 -15.284 c0), 0.5 -15.284 53.1% 0.5 -15.284 nulln(nullnullw) 0.5 -15.284 33.92 0.5 -15.284 0.5 -15.284 sense-topic 0.5 -15.284 benchmarked 0.5 -15.284 cuc 0.5 -15.284 24.12 0.5 -15.284 88.81 0.5 -15.284 χw 0.5 -15.284 illnesses, 0.5 -15.284 johnshopkinsuniversity 0.5 -15.284 1 0.5 -15.284 linkings 0.5 -15.284 main-efects 0.5 -15.284 importanceof 0.5 -15.284 0.63, 0.5 -15.284 s0r2 0.5 -15.284 102030405060708090 0.5 -15.284 90.17 0.5 -15.284 century) 0.5 -15.284 dominatedby 0.5 -15.284 mediocrity 0.5 -15.284 nonunderstanding, 0.5 -15.284  for 0.5 -15.284 optimization) 0.5 -15.284 (synonymy) 0.5 -15.284 okazaki, 0.5 -15.284 query-text 0.5 -15.284 viegas, 0.5 -15.284 pattern-based, 0.5 -15.284 2-layer 0.5 -15.284 5.34 0.5 -15.284 54.81 0.333333 -15.285 relatedto 0.333333 -15.285 car), 0.333333 -15.285 desegmenter 0.333333 -15.285 extensions: 0.333333 -15.285 2,034 0.333333 -15.285 (max) 0.333333 -15.285 pi−1 0.285714 -15.2853 $word 0.00260489 -15.2861 to 0.333333 -15.2869 (mcnemar’s 0.333333 -15.2869 (w1,w2,n,l,m,r) 0.333333 -15.2869 folders 0.333333 -15.2869 10.28 0.333333 -15.2869 cpd 0.333333 -15.2869 “rt 0.5 -15.2869 headset 0.5 -15.2869 funeral 0.5 -15.2869 now-standard 0.5 -15.2869 hawai’i 0.5 -15.2869 36.37 0.5 -15.2869 plans, 0.5 -15.2869 45.95 0.5 -15.2869 documentsw 0.5 -15.2869 74.51 0.5 -15.2869 1231 0.5 -15.2869 origine 0.5 -15.2869 …) 0.5 -15.2869 p3: 0.5 -15.2869 48.94 0.5 -15.2869 (commonly 0.5 -15.2869 p(m) 0.5 -15.2869 sched(j,i) 0.5 -15.2869 23.82 0.5 -15.2869 “lost” 0.5 -15.2869 78.1% 0.5 -15.2869 engelbrecht 0.5 -15.2869 scenario; 0.5 -15.2869 72.04 0.5 -15.2869 etc; 0.5 -15.2869 case-marking 0.5 -15.2869 10one 0.5 -15.2869 woz-08, 0.5 -15.2869 97.55 0.5 -15.2869 perplexity: 0.5 -15.2869 rrb 0.5 -15.2869 grandchildren 0.5 -15.2869 38.34 0.5 -15.2869 morally 0.5 -15.2869 fulfills 0.285714 -15.287 shasha 0.333333 -15.2888 22.73 0.333333 -15.2888 c(xl 0.333333 -15.2888 dui 0.333333 -15.2888 755 0.333333 -15.2888 irene 0.333333 -15.2888 haber 0.333333 -15.2888 (non-punctuation) 0.333333 -15.2888 “click 0.333333 -15.2888 essex 0.333333 -15.2888 hurried 0.333333 -15.2888 35.03 0.333333 -15.2888 93.4% 0.333333 -15.2888 parole 0.5 -15.2897 surrounds 0.5 -15.2897 mis-classification 0.5 -15.2897 is... 0.5 -15.2897 modulating 0.5 -15.2897 65.57 0.5 -15.2897 rolf 0.5 -15.2897 banter, 0.5 -15.2897 instruments, 0.5 -15.2897 15,383 0.5 -15.2897 (10%), 0.5 -15.2897 (15%) 0.5 -15.2897 overlaping 0.5 -15.2897 length(s) 0.5 -15.2897 62.09 0.5 -15.2897 http://en.wikipedia.org/wiki/list 0.5 -15.2897 traffic, 0.5 -15.2897 (prot) 0.5 -15.2897 chemicals 0.5 -15.2897 disgusted 0.5 -15.2897 hirao, 0.5 -15.2897 models,it 0.5 -15.2897 uterance: 0.5 -15.2897 conservative, 0.5 -15.2897 apex 0.5 -15.2897 (unconstrained) 0.5 -15.2897 62.78 0.5 -15.2897 labels.1 0.5 -15.2897 f-structure, 0.5 -15.2897 tr2str 0.5 -15.2897 paragraphs: 0.5 -15.2897 (eg 0.5 -15.2897 sentence’, 0.5 -15.2897 i-per-q, 0.5 -15.2897 is”, 0.5 -15.2897 96.87 0.5 -15.2897 step1: 0.5 -15.2897 sgt 0.5 -15.2897 cuis 0.5 -15.2897 russian: 0.5 -15.2897 bags-of-words 0.5 -15.2897 “snail,” 0.5 -15.2897 −agrc-lda-a 0.5 -15.2897 ipad), 0.5 -15.2897 filter) 0.5 -15.2897 ambitag 0.5 -15.2897 zukerman 0.5 -15.2897 ghani 0.5 -15.2897 jong-hoon 0.5 -15.2897 miami 0.5 -15.2897 60.9% 0.5 -15.2897 5to 0.5 -15.2897 sat.1 0.5 -15.2897 inter-article 0.222222 -15.2907 φn 0.333333 -15.2907 emanuele pianta 0.333333 -15.2907 27.45 0.333333 -15.2907 0.7.25 (snover 0.333333 -15.2907 chklovski 0.333333 -15.2907 “unknown 0.333333 -15.2907 fk(a) 0.333333 -15.2907 lyrics 0.333333 -15.2907 (fixed 0.333333 -15.2907 cup, 0.157895 -15.2911 kozareva 0.25 -15.2912 event representations 0.333333 -15.2926 cooked, 0.333333 -15.2926 chair, 0.333333 -15.2926 1034 0.333333 -15.2926 read” 0.333333 -15.2926 2.3.3 0.333333 -15.2926 decidable 0.333333 -15.2926 conceivably 0.5 -15.2926 72.44 0.5 -15.2926 broadway 0.5 -15.2926 hierarchy; 0.5 -15.2926 13.63 0.5 -15.2926 (aone 0.5 -15.2926 ...the 0.5 -15.2926 poe 0.5 -15.2926 existence, 0.5 -15.2926 pc(t,w) 0.5 -15.2926 “build 0.5 -15.2926 glm 0.5 -15.2926 21.16 0.5 -15.2926 regularization), 0.5 -15.2926 k1. 0.5 -15.2926 24.60 0.5 -15.2926 speedups, 0.5 -15.2926 16.86 0.5 -15.2926 75.65 0.5 -15.2926 30.59 0.5 -15.2926 print.v 0.5 -15.2926 (hirao 0.5 -15.2926 “ali” 0.5 -15.2926 1301 0.5 -15.2926 17.92 0.5 -15.2926 nullh 0.5 -15.2926 36.1% 0.5 -15.2926 coarse+patterns 0.5 -15.2926 keyness 0.5 -15.2926 p(xuv 0.5 -15.2926 p(lc) 0.5 -15.2926 vp: 0.5 -15.2926 elementwise 0.5 -15.2926 (dis)agree 0.5 -15.2926 crescent 0.5 -15.2926 trigrams; 0.5 -15.2926 annotatorto 0.5 -15.2926 rudzicz 0.25 -15.2941 .7 0.333333 -15.2945 81.41 0.333333 -15.2945 pages: 0.333333 -15.2945 inlink 0.333333 -15.2945 0.10, 0.333333 -15.2945 glasses 0.333333 -15.2945 lookingat 0.333333 -15.2945 ∗the 0.333333 -15.2945 lesk, 0.333333 -15.2945 produces, 0.333333 -15.2945 kids 0.333333 -15.2945 “allergic 0.333333 -15.2945 -0.41 0.0823529 -15.2949 future. 0.285714 -15.2951 d100 0.285714 -15.2951 κ0 0.5 -15.2954 noodles 0.5 -15.2954 sub-results, 0.5 -15.2954 (sb) 0.5 -15.2954 23.78 0.5 -15.2954 66.68 0.5 -15.2954 toolkit: 0.5 -15.2954 “assassinated”, 0.5 -15.2954 rendering, 0.5 -15.2954 1550 0.5 -15.2954 response’s 0.5 -15.2954 (2-a) 0.5 -15.2954 2893 0.5 -15.2954 49.49 0.5 -15.2954 nerve) 0.5 -15.2954 52.45 0.5 -15.2954 level), 0.5 -15.2954 important: 0.5 -15.2954 precisions, 0.5 -15.2954 tex 0.5 -15.2954 (kb), 0.5 -15.2954 “what, 0.5 -15.2954 coord: 0.5 -15.2954 f(wi,cj) 0.5 -15.2954 oun 0.5 -15.2954 xiaodan 0.5 -15.2954 moran 198, 0.5 -15.2954 88.01 0.5 -15.2954 0.17252 0.5 -15.2954 word-window 0.5 -15.2954 ptk-ct+csk 0.5 -15.2954 (value 0.5 -15.2954 prepositional, 0.5 -15.2954 gon 0.5 -15.2954 63.33 0.5 -15.2954 (all), 0.5 -15.2954 4×104 0.5 -15.2954 kampala 0.5 -15.2954 abbas 0.5 -15.2954 mostow, 0.5 -15.2954 i); 0.5 -15.2954 s/n 0.5 -15.2954 lscript2-penalty 0.333333 -15.2964 729 0.333333 -15.2964 honour 0.333333 -15.2964 dij 0.333333 -15.2964 pos-tagged, 0.046875 -15.2968 on all 0.25 -15.2969 40.64 0.166667 -15.2974 formally, let 0.157895 -15.2974 [np@1 0.333333 -15.2983 favourite 0.333333 -15.2983 minimum-cut approach 0.333333 -15.2983 high-precision, 0.333333 -15.2983 supp∈h0 0.333333 -15.2983 (r1 0.333333 -15.2983 <=20 0.333333 -15.2983 deepen 0.333333 -15.2983 1070 0.333333 -15.2983 determiners: 0.333333 -15.2983 “같 0.333333 -15.2983 94.52 0.333333 -15.2983 talp 0.5 -15.2983 rich-get-richer 0.5 -15.2983 p(e,f,a;θ) 0.5 -15.2983 child: 0.5 -15.2983 26.71 0.5 -15.2983 +ha 0.5 -15.2983 ø“ 0.5 -15.2983 possibilities: 0.5 -15.2983 0.3029 0.5 -15.2983 rentals 0.5 -15.2983 (380) 0.5 -15.2983 86.59 0.5 -15.2983 model,with 0.5 -15.2983 non-compositionality, 0.5 -15.2983 synchronic 0.5 -15.2983 pre-determined 0.5 -15.2983 57.63 0.5 -15.2983 “spot 0.5 -15.2983 21.14 0.5 -15.2983 at-l 0.5 -15.2983 spots, 0.5 -15.2983 1051 0.5 -15.2983 intrusive 0.5 -15.2983 seed-based 0.5 -15.2983 irrelevancy 0.5 -15.2983 20.06 0.5 -15.2983 (recall: 0.5 -15.2983 wmt’09 0.5 -15.2983 flip: 0.5 -15.2983 incoherent, 0.5 -15.2983 mediator 0.5 -15.2983 31.25 0.5 -15.2983 zurich, 0.5 -15.2983 zero), 0.5 -15.2983 23.36 0.5 -15.2983 million-word 0.5 -15.2983 3.3m 0.5 -15.2983 "%# 0.5 -15.2983 pronoun), 0.5 -15.2983 emotion-based 0.5 -15.2983 (i,j,l)∈a 0.5 -15.2983 70.05 0.5 -15.2983 paraphrrel 0.5 -15.2983 advertisements, 0.0695652 -15.3001 obtained with 0.333333 -15.3003 subgrammars 0.333333 -15.3003 57.27 0.333333 -15.3003 1599 0.333333 -15.3003 zs∈ys 0.333333 -15.3003 23.56 0.333333 -15.3003 34.26 0.333333 -15.3003 lastsibl(h,m,k) 0.333333 -15.3003 editable 0.176471 -15.3004 ,s 0.5 -15.3012 (evert 0.5 -15.3012 “livestock” 0.5 -15.3012 77.12 0.5 -15.3012 hyperarticulation 0.5 -15.3012 〈m,n,a〉, 0.5 -15.3012 17.88 0.5 -15.3012 (ties 0.5 -15.3012 51.01 0.5 -15.3012 meta-metrics 0.5 -15.3012 opensmile 0.5 -15.3012 “famous 0.5 -15.3012 young-sook 0.5 -15.3012 -ne 0.5 -15.3012 “xs 0.5 -15.3012 monologue-dialogue 0.5 -15.3012 formality, 0.5 -15.3012 5%), 0.5 -15.3012 j): 0.5 -15.3012 systematized 0.5 -15.3012 64.92 0.5 -15.3012 pitfallsof 0.5 -15.3012 centriod 0.5 -15.3012 popularities 0.5 -15.3012 blackberry, 0.5 -15.3012 rules/patterns 0.5 -15.3012 –1 0.5 -15.3012 briefings 0.5 -15.3012 crúbadán 0.5 -15.3012 1m, 0.5 -15.3012 y∗, 0.5 -15.3012 4,958 0.5 -15.3012 quote-response 0.222222 -15.3022 nocedal, 0.333333 -15.3022 systemsto 0.333333 -15.3022 ub 0.333333 -15.3022 0.776 0.333333 -15.3022 37k 0.333333 -15.3022 foo 0.333333 -15.3022 memory: 0.333333 -15.3022 (dang 0.333333 -15.3022 (1979) 0.333333 -15.3022 28.88 0.333333 -15.3022 [is 0.333333 -15.3022 busines 0.333333 -15.3022 23.20 0.1 -15.3024 army 0.15 -15.3027 post-translational 0.285714 -15.3033 λadj 0.333333 -15.3041 cu-bojar, 0.333333 -15.3041 painter 0.333333 -15.3041 (half 0.333333 -15.3041 deg, 0.333333 -15.3041 that's 0.333333 -15.3041 2008a, 0.333333 -15.3041 confront 0.333333 -15.3041 “two 0.5 -15.3041 rank(i) 0.5 -15.3041 semanticparser, 0.5 -15.3041 prizes 0.5 -15.3041 0.66, 0.5 -15.3041 veiled 0.5 -15.3041 “dfe 0.5 -15.3041 invocabulary 0.5 -15.3041 almthdp 0.5 -15.3041 directory, 0.5 -15.3041 “usability” 0.5 -15.3041 infrequency 0.5 -15.3041 ‘become’ 0.5 -15.3041 knowtator 0.5 -15.3041 31.57 0.5 -15.3041 cdec, 0.5 -15.3041 “reverse 0.5 -15.3041 tourism, 0.5 -15.3041 rte: 0.5 -15.3041 sentencepov 0.5 -15.3041 dec-all 0.5 -15.3041 (aluisio 0.5 -15.3041 belong, 0.5 -15.3041 smell, 0.5 -15.3041 rig, 0.5 -15.3041 typed-frequency 0.5 -15.3041 watershed 0.5 -15.3041 ugc, 0.5 -15.3041 lemma: 0.5 -15.3041 compared: 0.5 -15.3041 (84 0.5 -15.3041 ¯c 0.5 -15.3041 01/02 0.5 -15.3041 37.24 0.5 -15.3041 73.71 0.0721649 -15.3047 is relatively 0.2 -15.3052 word). 0.106383 -15.3055 validation. 0.102041 -15.3056 opinion word 0.03125 -15.3057 we then 0.333333 -15.306 saa 0.333333 -15.306 86.05 0.333333 -15.306 beam-width, 0.333333 -15.306 george w. 0.333333 -15.306 3%, 0.333333 -15.306 noisyand 0.333333 -15.306 pathto-root 0.285714 -15.3066 yee 0.5 -15.307 65.23 0.5 -15.307 jaakkola, 0.5 -15.307 evince 0.5 -15.307 sentences/tweets 0.5 -15.307 “?�(fahui)/demonstrate” 0.5 -15.307 arcs) 0.5 -15.307 m´exico 0.5 -15.307 used.1 0.5 -15.307 tart 0.5 -15.307 face”) 0.5 -15.307 nullpredicatenull: 0.5 -15.307 socialized 0.5 -15.307 ‘mixed 0.5 -15.307 metal 0.5 -15.307 łp@ 0.5 -15.307 cssvm 0.5 -15.307 ��� 0.5 -15.307 36.92 0.5 -15.307 g+cs 0.5 -15.307 flaws 0.5 -15.307 9.86 0.5 -15.307 14k 0.5 -15.307 diehl, 0.5 -15.307 hijacked 0.5 -15.307 57.71 0.5 -15.307 np[i169→i168] 0.5 -15.307 “quarterback 0.5 -15.307 floor; 0.5 -15.307 midwest 0.5 -15.307 885 0.5 -15.307 (transductive 0.5 -15.307 max planck 0.5 -15.307 arranges 0.5 -15.307 parataxis 0.5 -15.307 (24%) 0.5 -15.307 (pt, 0.5 -15.307 ghosh 0.5 -15.307 α· 0.5 -15.307 defend, 0.5 -15.307 disruption 0.5 -15.307 nullnull(nullnullnull) 0.5 -15.307 priors+em1 0.5 -15.307 referees) 0.5 -15.307 73.40 0.5 -15.307 ibardbl 0.5 -15.307 (1k 0.5 -15.307 niculescu-mizil, 0.5 -15.307 deceive 0.5 -15.307 vanishingly 0.333333 -15.308 yuret 0.333333 -15.308 receptors 0.333333 -15.308 dtnmod 0.0701754 -15.3083 be represented 0.285714 -15.3083 being, 0.09375 -15.3086 (smt) 0.15 -15.3096 weakness 0.333333 -15.3099 twolevel 0.333333 -15.3099 bi−1 0.333333 -15.3099 tonelli 0.333333 -15.3099 ci.indict 0.333333 -15.3099 i∈m 0.333333 -15.3099 15.36 0.333333 -15.3099 abbreviate 0.333333 -15.3099 fk(t,yt,yt−1,x) 0.5 -15.3099 micro-f 0.5 -15.3099 “prominent 0.5 -15.3099 §a.2 0.5 -15.3099 6lprq/dfrvwh-xolhq 0.5 -15.3099 assent 0.5 -15.3099 mouth” 0.5 -15.3099 numbers), 0.5 -15.3099 (hopefully) 0.5 -15.3099 unrestricted, 0.5 -15.3099 champaign, 0.5 -15.3099 x\y 0.5 -15.3099 ¯h, 0.5 -15.3099 semanticfeatures 0.5 -15.3099 66.6% 0.5 -15.3099 1403 0.5 -15.3099 1:3 0.5 -15.3099 estimators, 0.5 -15.3099 probably, 0.5 -15.3099 observational 0.5 -15.3099 19.72 0.5 -15.3099 136k 0.5 -15.3099 (13), 0.5 -15.3099 p=0.01 0.5 -15.3099 sennheiser 0.5 -15.3099 himself, 0.5 -15.3099 76.63 0.5 -15.3099 21.89 0.5 -15.3099 carefully, 0.5 -15.3099 “saw 0.5 -15.3099 2∗p∗rp+r 0.5 -15.3099 '06 0.5 -15.3099 (2001): 0.157895 -15.3111 decreases. 0.25 -15.3114 s51s48s48s48 0.084507 -15.3115 that had 0.285714 -15.3116 α1, 0.285714 -15.3116 non-verb 0.333333 -15.3118 (sampling 0.333333 -15.3118 (1979), 0.333333 -15.3118 detect, 0.333333 -15.3118 type1 0.333333 -15.3118 80.53 0.333333 -15.3118 (approximately) 0.2 -15.3122 tomoko 0.5 -15.3128 li† 0.5 -15.3128 (truth 0.5 -15.3128 56.90 0.5 -15.3128 csubj, 0.5 -15.3128 katzav 0.5 -15.3128 dip, 0.5 -15.3128 cercone, 0.5 -15.3128 guthrie, 0.5 -15.3128 fromu 0.5 -15.3128 x1,...,xn, 0.5 -15.3128 me) 0.5 -15.3128 1,695 0.5 -15.3128 vger, 0.5 -15.3128 constructs: 0.5 -15.3128 [between 0.5 -15.3128 33.71 0.5 -15.3128 66.00 0.5 -15.3128 24.15 0.5 -15.3128 hard-coded 0.5 -15.3128 smrg, 0.5 -15.3128 c+p 0.5 -15.3128 8.6m 0.5 -15.3128 76.12 0.5 -15.3128 subtotal 0.5 -15.3128 *% 0.5 -15.3128 admixture 0.5 -15.3128 alςab 0.5 -15.3128 obeys 0.5 -15.3128 english’ 0.5 -15.3128 trial-and-error 0.5 -15.3128 (sov) 0.5 -15.3128 boden 0.5 -15.3128 [farhadi 0.5 -15.3128 hm(e 0.5 -15.3128 uation 0.5 -15.3128 huggins-daines, 0.5 -15.3128 misrepresented 0.5 -15.3128 0.546 0.5 -15.3128 geo:instate(y,x) 0.5 -15.3128 i-2 0.5 -15.3128 ,→made 0.5 -15.3128 ρmonths 0.5 -15.3128 trans5 0.5 -15.3128 cost(ι) 0.285714 -15.3133 pustejovsky. 0.285714 -15.3133 gambell 0.285714 -15.3133 sulins, 0.285714 -15.3133 ma. 0.285714 -15.3133 bloehdorn 0.285714 -15.3133 icd-9-cm 0.285714 -15.3133 (kate 0.285714 -15.3133 magnini, 0.285714 -15.3133 university ithaca, 0.333333 -15.3138 {a, 0.333333 -15.3138 timet, 0.333333 -15.3138 peters 0.333333 -15.3138 1537 0.333333 -15.3138 este 0.0869565 -15.3139 however, when 0.0430464 -15.3156 end of 0.333333 -15.3157 ynq 0.333333 -15.3157 morethan 0.333333 -15.3157 i,cm,n 0.333333 -15.3157 upv-ehu 0.333333 -15.3157 qing 0.333333 -15.3157 scherrer 0.333333 -15.3157 (purpose) 0.333333 -15.3157 (next) 0.333333 -15.3157 before; 0.333333 -15.3157 wunderwald 0.5 -15.3157 tests” 0.5 -15.3157 zeitung” 0.5 -15.3157 91.75 0.5 -15.3157 attempto 0.5 -15.3157 29.32 0.5 -15.3157 abacha 0.5 -15.3157 k4a 0.5 -15.3157 1.32 0.5 -15.3157 ),(),( 0.5 -15.3157 (undirected 0.5 -15.3157 66.17 0.5 -15.3157 semantic-based 0.5 -15.3157 sbh 0.5 -15.3157 constraint; 0.5 -15.3157 usera 0.5 -15.3157 “stuffy”/“cold” 0.5 -15.3157 1067 0.5 -15.3157 81, 0.5 -15.3157 11.4%, 0.5 -15.3157 (modified 0.5 -15.3157 , 0.5 -15.3157 (xi,zi) 0.5 -15.3157 6561 0.5 -15.3157 null1: 0.5 -15.3157 h@q 0.5 -15.3157 jsummationdisplay 0.5 -15.3157 arkansas 0.5 -15.3157 aspectual/temporal 0.333333 -15.3177 brain, 0.333333 -15.3177 ‘s 0.333333 -15.3177 ∀x 0.333333 -15.3177 44.28 0.333333 -15.3177 maxent/ge 0.333333 -15.3177 81.3% 0.333333 -15.3177 200000 0.333333 -15.3177 53.47 0.333333 -15.3177 youngest 0.333333 -15.3177 separately; 0.5 -15.3186 (wubben 0.5 -15.3186 o(n3q3) 0.5 -15.3186 blurred 0.5 -15.3186 yüz 0.5 -15.3186 1954 0.5 -15.3186 (2008)† 0.5 -15.3186 non-defeasible 0.5 -15.3186 21.96 0.5 -15.3186 ‘weak 0.5 -15.3186 subcorpora, 0.5 -15.3186 line-edges 0.5 -15.3186 susanne 0.5 -15.3186 70.50 0.5 -15.3186 30.35 0.5 -15.3186 p+t+c, 0.5 -15.3186 “nose”/“cold”) 0.5 -15.3186 constraints; 0.5 -15.3186 network1 0.5 -15.3186 sentencew 0.5 -15.3186 ‘bad’ 0.5 -15.3186  0.5 -15.3186 o” 0.5 -15.3186 (vallabha 0.5 -15.3186 originalsentence 0.5 -15.3186 (etc) 0.5 -15.3186 (cat) 0.5 -15.3186 top–list 0.5 -15.3186 sox” 0.5 -15.3186 φ▽ψ(p(pi)) 0.5 -15.3186 answer), 0.5 -15.3186 (br), 0.5 -15.3186 i(c) 0.5 -15.3186 banning 0.5 -15.3186 1’ 0.5 -15.3186 spam: 0.5 -15.3186 communication) 0.5 -15.3186 desired reordered 0.5 -15.3186 3297 0.5 -15.3186 translated; 0.5 -15.3186 (muc, 0.5 -15.3186 22.32 0.5 -15.3186 θtraine 0.5 -15.3186 1029 0.25 -15.3187 ¯nw 0.06 -15.3193 absence 0.333333 -15.3196 benoˆıt 0.333333 -15.3196 taxonomically 0.333333 -15.3196 cop 0.333333 -15.3196 26.11 0.333333 -15.3196 orderto 0.333333 -15.3216 uncertain+incorrect 0.333333 -15.3216 acquisitions 0.333333 -15.3216 parked 0.333333 -15.3216 35.77 0.333333 -15.3216 popularity-based 0.333333 -15.3216 execution, 0.333333 -15.3216 participatory 0.333333 -15.3216 informationon 0.5 -15.3216 “same” 0.5 -15.3216 ∀p 0.5 -15.3216 (n, 0.5 -15.3216 26.38 0.5 -15.3216 conventional, 0.5 -15.3216 51,045 0.5 -15.3216 integrating mwu 0.5 -15.3216 tally 0.5 -15.3216 2], 0.5 -15.3216 (already 0.5 -15.3216 ngrams’, 0.5 -15.3216 48.90 0.5 -15.3216 tsc3 0.5 -15.3216 ―olga‖ 0.5 -15.3216 h 0.5 -15.3364 proportions, 0.5 -15.3364 33.97 0.5 -15.3364 (tratz 0.5 -15.3364 paraphrase), 0.5 -15.3364 -1.23 0.5 -15.3364 42.06 0.5 -15.3364 depression, 0.5 -15.3364 90.81 0.5 -15.3364 602 0.5 -15.3364 cfv=3,h=1 0.5 -15.3364 responses) 0.5 -15.3364 identified, 0.5 -15.3364 hashing (ssh) 0.5 -15.3364 2.44 0.0520833 -15.3367 usa, 0.333333 -15.3374 englisharabic 0.333333 -15.3374 step-size 0.333333 -15.3374 (v,e), 0.333333 -15.3374 100,000-word 0.0620155 -15.3375 table 6 0.0304054 -15.3376 work. 0.166667 -15.3385 author(s) 0.333333 -15.3394 bmi@asu 0.333333 -15.3394 ‘language 0.333333 -15.3394  0.333333 -15.3394 0.699 0.333333 -15.3394 younger, 0.333333 -15.3394 h(·) 0.333333 -15.3394 cyber 0.333333 -15.3394 whats 0.5 -15.3394 +8.7% 0.5 -15.3394 στ∈trans(w1 0.5 -15.3394 57.40 0.5 -15.3394 audio) 0.5 -15.3394 taxonomy; 0.5 -15.3394 31.48 0.5 -15.3394 61.66 0.5 -15.3394 42.92 0.5 -15.3394 range(r) 0.5 -15.3394 non-verbs 0.5 -15.3394 ostensible 0.5 -15.3394 bacchiani 0.5 -15.3394 loi, 0.5 -15.3394 sacrificed 0.5 -15.3394 1061 0.5 -15.3394 heller 0.5 -15.3394 rtp 0.5 -15.3394 discouraged 0.5 -15.3394 nw, 0.5 -15.3394 subpopulations, 0.5 -15.3394 disallows 0.5 -15.3394 “edit” 0.5 -15.3394 23.08 0.5 -15.3394 transcriptions) 0.5 -15.3394 3.78 0.5 -15.3394 s1···sn 0.5 -15.3394 50/50 0.5 -15.3394 combinatorics 0.5 -15.3394 (favre 0.5 -15.3394 exp.arg1 0.5 -15.3394 set— 0.5 -15.3394 formal”, 0.5 -15.3394 helping%1:13:00 0.5 -15.3394 overtime 0.5 -15.3394 ius 0.5 -15.3394 tm-euc 0.5 -15.3394 ray, 0.5 -15.3394 (precision) 0.5 -15.3394 lowing 0.5 -15.3394 determinations 0.176471 -15.3398 sir 0.125 -15.3401 latter, 0.333333 -15.3414 trade, 0.333333 -15.3414 sorkal 0.333333 -15.3414 θe, 0.333333 -15.3414 60.85 0.333333 -15.3414 43.75 0.333333 -15.3414 (≥ 0.333333 -15.3414 ، 0.333333 -15.3414 18.40 0.333333 -15.3414 charter 0.333333 -15.3414 +1.29 0.333333 -15.3414 arises: 0.333333 -15.3414 oard 0.5 -15.3424 structures”, 0.5 -15.3424 bounded-stack 0.5 -15.3424 subjects; 0.5 -15.3424 0.378 0.5 -15.3424 v1.0 0.5 -15.3424 0.04) 0.5 -15.3424 nn1 0.5 -15.3424 firefox 0.5 -15.3424 (annotation 0.5 -15.3424 hesitations 0.5 -15.3424 10.2m 0.5 -15.3424 subjective-positive 0.5 -15.3424 hy06 0.5 -15.3424 multi-dialect 0.5 -15.3424 effect); 0.5 -15.3424 codeword) 0.5 -15.3424 mester, 0.5 -15.3424 jonas 0.5 -15.3424 rmax 0.5 -15.3424 afinitor” 0.5 -15.3424 90.22 0.5 -15.3424 users) 0.5 -15.3424 badly, 0.5 -15.3424 (oflazer 0.5 -15.3424 services) 0.5 -15.3424 f∗; 0.5 -15.3424 nnpposjj 0.5 -15.3424 1,181 0.5 -15.3424 articles; 0.5 -15.3424 l´actea 0.5 -15.3424 numerate 0.5 -15.3424 85.95 0.5 -15.3424 source-tag 0.064 -15.3434 label. 0.333333 -15.3434 (persons, 0.333333 -15.3434 strips 0.333333 -15.3434 “global” 0.333333 -15.3434 20.37 0.333333 -15.3434 jens 0.333333 -15.3434 postpositions, 0.333333 -15.3454 −0.04 0.333333 -15.3454 l(λ) 0.333333 -15.3454 0.768 0.333333 -15.3454 stations 0.5 -15.3454 communication; 0.5 -15.3454 commissioner fischler, 0.5 -15.3454 ···rm} 0.5 -15.3454 17.27 0.5 -15.3454 (accurate 0.5 -15.3454 layer’s 0.5 -15.3454 c∈wc 0.5 -15.3454 belself, 0.5 -15.3454 refereed 0.5 -15.3454 deny, 0.5 -15.3454 9.10 0.5 -15.3454 2.1); 0.5 -15.3454 pre-filtered 0.5 -15.3454 0.0138 0.5 -15.3454 ga¢j p@q 0.5 -15.3454 than)” 0.5 -15.3454 memoir 0.5 -15.3454 contrib 0.5 -15.3454 hdp-lg 0.5 -15.3454 ∗lower 0.5 -15.3454 plus, 0.5 -15.3454 jitter, 0.5 -15.3454 1808 0.5 -15.3454 1491 0.5 -15.3454 £ 0.5 -15.3454 certification 0.5 -15.3454 66.42 0.5 -15.3454 51.90 0.5 -15.3454 entitled, 0.5 -15.3454 indicatethe 0.5 -15.3454 malicious 0.5 -15.3454 (lucy, 0.5 -15.3454 iis-0713448, 0.5 -15.3454 1602 0.5 -15.3454 modifier’s 0.5 -15.3454 white?” 0.5 -15.3454 annalu 0.5 -15.3454 connection, 0.5 -15.3454 kronecker’s 0.5 -15.3454 fass, 0.5 -15.3454 corectness 0.5 -15.3454 tures, 0.5 -15.3454 leftward 0.5 -15.3454 16.48 0.0320755 -15.3469 them. 0.333333 -15.3474 fea 0.333333 -15.3474 wubben’s 0.333333 -15.3474 insult, 0.333333 -15.3474 620 0.333333 -15.3474 ≥1 0.333333 -15.3474 手 0.5 -15.3484 candidate: 0.5 -15.3484 90.35 0.5 -15.3484 each), 0.5 -15.3484 ⇓ 0.5 -15.3484 prefix/suffix 0.5 -15.3484 fatigue, 0.5 -15.3484 holographic 0.5 -15.3484 chromosomal 0.5 -15.3484 abdullah 0.5 -15.3484 20.28 0.5 -15.3484 1402 0.5 -15.3484 0.53, 0.5 -15.3484 f(jiqiao)” 0.5 -15.3484 ntcir-ch 0.5 -15.3484 34.88 0.5 -15.3484 “abdominal 0.5 -15.3484 1,915 0.5 -15.3484 4although 0.5 -15.3484 =y, 0.5 -15.3484 28.32 0.5 -15.3484 1.5.5 0.5 -15.3484 ntcir-en 0.5 -15.3484 flat-bottomed 0.5 -15.3484 shih 0.5 -15.3484 gcrbeta 0.5 -15.3484 to-day 0.5 -15.3484 20.38 0.5 -15.3484 devices: 0.5 -15.3484 procedure), 0.5 -15.3484 polarity-based 0.5 -15.3484 simplified 0.5 -15.3484 (wright 0.5 -15.3484 daurate 0.333333 -15.3494 boullier, 0.333333 -15.3494 stan 0.333333 -15.3494 7.1% 0.333333 -15.3494 rwth-cder 0.333333 -15.3494 liking 0.5 -15.3514 joern 0.5 -15.3514 2011)) 0.5 -15.3514 63.79 0.5 -15.3514 bestc1000 0.5 -15.3514 ubiu (zhekova 0.5 -15.3514 bill clinton. 0.5 -15.3514 (non-rumor) 0.5 -15.3514 (ta) 0.5 -15.3514 (huet 0.5 -15.3514 (20ng) 0.5 -15.3514 structurethe 0.5 -15.3514 operations), 0.5 -15.3514 32.40 0.5 -15.3514 date=abs; 0.5 -15.3514 shachar 0.5 -15.3514 &month-in-daymonth 0.5 -15.3514 “demethylation” 0.5 -15.3514 steinbeck 0.5 -15.3514 0.979 0.5 -15.3514 0.8764 0.5 -15.3514 yasin 0.5 -15.3514 17.75 0.5 -15.3514 “clearly 0.5 -15.3514 (18), 0.5 -15.3514 (26%), 0.5 -15.3514 2.80 0.5 -15.3514 day=abs 0.5 -15.3514 0.09, 0.5 -15.3514 29.88 0.5 -15.3514 (digital 0.5 -15.3514 vp-noner 0.5 -15.3514 73.28 0.5 -15.3514 64.85 0.5 -15.3514 henning, 0.5 -15.3514 “non-fringe 0.5 -15.3514 33.65 0.5 -15.3514 42.28 0.5 -15.3514 lexicalstylistic 0.5 -15.3514 transductions, 0.5 -15.3514 f(xij) 0.5 -15.3514 3.81% 0.5 -15.3514 destroying 0.5 -15.3514 (pbmt), 0.5 -15.3514 coherency, 0.333333 -15.3515 pos(p) 0.333333 -15.3515 ∈i 0.333333 -15.3515 72.21 0.333333 -15.3515 insulin 0.333333 -15.3515 646 0.333333 -15.3515 head” 0.333333 -15.3515 k=0 0.333333 -15.3515 p2prime 0.333333 -15.3515 5410 0.333333 -15.3535 thus,the 0.333333 -15.3535 1111 0.333333 -15.3535 lh1(null), 0.333333 -15.3535 {} 0.333333 -15.3535 isst 0.333333 -15.3535 lecturer 0.333333 -15.3535 corn 0.333333 -15.3535 (kuˇcera 0.333333 -15.3535 title(keyp) 0.333333 -15.3535 abbrev 0.5 -15.3545 ip, 0.5 -15.3545 ogre] 0.5 -15.3545 3121 0.5 -15.3545 52.48 0.5 -15.3545 “vague” 0.5 -15.3545 (xprime(s) 0.5 -15.3545 interchanged 0.5 -15.3545 khemphwuthe-ka 0.5 -15.3545 38.54 0.5 -15.3545 succinctly: 0.5 -15.3545 67.65 0.5 -15.3545 holidays 0.5 -15.3545 framenet-style 0.5 -15.3545 (would 0.5 -15.3545 “starting 0.5 -15.3545 (%); 0.5 -15.3545 clp 0.5 -15.3545 sentiment-annotated 0.5 -15.3545 83.09 0.5 -15.3545 toward, 0.5 -15.3545 +1.76 0.5 -15.3545 uninverted 0.5 -15.3545 lct, 0.5 -15.3545 “too 0.5 -15.3545 unsen 0.5 -15.3545 basque: 0.5 -15.3545 look” 0.5 -15.3545 k(d1,d2) 0.5 -15.3545 years’ 0.5 -15.3545 party’ 0.5 -15.3545 “kevin 0.5 -15.3545 suffix: 0.5 -15.3545 unthank, 0.5 -15.3545 with l+) 0.5 -15.3545 3-voting 0.5 -15.3545 cldaesel,+ 0.5 -15.3545 72.97 0.5 -15.3545 (przybocki 0.5 -15.3545 91.7% 0.5 -15.3545 irregardless 0.5 -15.3545 ‘semantic 0.5 -15.3545 ‘topic 0.5 -15.3545 w-euc 0.5 -15.3545 “from”, 0.5 -15.3545 30%) 0.5 -15.3545 “where”, 0.5 -15.3545 23.13 0.5 -15.3545 (dreuw 0.5 -15.3545 side); 0.5 -15.3545 heuristically, 0.0672269 -15.3553 conclusions and future 0.333333 -15.3555 “of”) 0.333333 -15.3555 certain+incorrect 0.333333 -15.3555 subsequence, 0.333333 -15.3555 confusionnetwork 0.333333 -15.3555 mahalanobis 0.333333 -15.3555 46.76 0.1 -15.3557 associating 0.15 -15.356 other users 0.333333 -15.3575 away” 0.333333 -15.3575 lessmuch 0.333333 -15.3575 (ap 0.333333 -15.3575 rj, 0.5 -15.3575 averageflu 0.5 -15.3575 74.26 0.5 -15.3575 sys6 0.5 -15.3575 78.59 0.5 -15.3575 (maynard, 0.5 -15.3575 2,2can 0.5 -15.3575 (+ts), 0.5 -15.3575 sumary, 0.5 -15.3575 punctuation-induced 0.5 -15.3575 tirar 0.5 -15.3575 consultation, 0.5 -15.3575 folow, 0.5 -15.3575 1958), 0.5 -15.3575 2using 0.5 -15.3575 triad) 0.5 -15.3575 wh-element 0.5 -15.3575 usedas 0.5 -15.3575 0.668 0.5 -15.3575 beatrix 0.5 -15.3575 “as” 0.5 -15.3575 o(n8) 0.5 -15.3575 alif 0.5 -15.3575 (ldc2007t07) 0.5 -15.3575 79.31 0.5 -15.3575 ≤5 0.5 -15.3575 +0.13 0.5 -15.3575 “concession” 0.5 -15.3575 endpoints(i) 0.5 -15.3575 p17�p16 0.5 -15.3575 48.19 0.5 -15.3575 uncontroversial 0.5 -15.3575 cfv=4,h=1 0.5 -15.3575 2.64 0.5 -15.3575 74k 0.5 -15.3575 12%, 0.5 -15.3575 74.81 0.5 -15.3575 viennent” 0.5 -15.3575 janke 0.5 -15.3575 64.27 0.5 -15.3575 lurking 0.5 -15.3575 (bloomfield 0.333333 -15.3596 “signal 0.333333 -15.3596 sett 0.333333 -15.3596 gerunds, 0.333333 -15.3596 uprime 0.333333 -15.3596 dog, 0.333333 -15.3596 summationtextnj=1 0.333333 -15.3596 0.057 0.5 -15.3605 k1-th 0.5 -15.3605 trial: 0.5 -15.3605 people] 0.5 -15.3605 40.48 0.5 -15.3605 ldahs 0.5 -15.3605 between-group 0.5 -15.3605 1,2,3, 0.5 -15.3605 hay, 0.5 -15.3605 emphatic 0.5 -15.3605 avgtriangle: 0.5 -15.3605 tumor, 0.5 -15.3605 strings) 0.5 -15.3605 women; 0.5 -15.3605 83.47 0.5 -15.3605 sophisticatedand 0.5 -15.3605 north, 0.5 -15.3605 ‘at 0.5 -15.3605 21.76 0.5 -15.3605 133k 0.5 -15.3605 dubitative 0.5 -15.3605 77.98 0.5 -15.3605 delight 0.5 -15.3605 taxonomical 0.5 -15.3605 english-bulgarian 0.5 -15.3605 57.26 0.5 -15.3605 multivalued. 0.5 -15.3605 transl. 0.5 -15.3605 m-to-n 0.5 -15.3605 devastated 0.5 -15.3605 iii (born 0.5 -15.3605 ‘definite 0.5 -15.3605 87.25 0.5 -15.3605 iis-0968450, 0.5 -15.3605 ties), 0.5 -15.3605 0.26) 0.5 -15.3605 begin(t,l) 0.5 -15.3605 6.82 0.5 -15.3605 pairwiseκ 0.5 -15.3605 69.37 0.5 -15.3605 partitions: 0.5 -15.3605 diplomacy, 0.5 -15.3605 haidian 0.5 -15.3605 .c/, 0.181818 -15.3608 witten-bell 0.114286 -15.3611 compensate 0.2 -15.3612 null null, 0.333333 -15.3616 oscars 0.333333 -15.3616 vrp,r 0.333333 -15.3616 {a1,··· 0.333333 -15.3616 ostpa 0.0441176 -15.3617 language processing, 0.333333 -15.3636 figure6 0.333333 -15.3636 ceo-of(jobs, 0.333333 -15.3636 configuration: 0.333333 -15.3636 (mi), 0.333333 -15.3636 slop 0.5 -15.3636 henceforth), 0.5 -15.3636 problems/tests/treatments 0.5 -15.3636 something”, 0.5 -15.3636 00, 0.5 -15.3636 algebras, 0.5 -15.3636 guise 0.5 -15.3636 74.43 0.5 -15.3636 baging 0.5 -15.3636 guarantes 0.5 -15.3636 73.68 0.5 -15.3636 66.27 0.5 -15.3636 starting, 0.5 -15.3636 ate), 0.5 -15.3636 decision” 0.5 -15.3636 1-3, 0.5 -15.3636 simulation; 0.5 -15.3636 elias 0.5 -15.3636 79.65 0.5 -15.3636 exploding 0.5 -15.3636 lorg 0.5 -15.3636 (montague 0.5 -15.3636 1−5-grams 0.5 -15.3636 4.24 0.5 -15.3636 31.30 0.5 -15.3636 4.68 0.5 -15.3636 ps3 0.5 -15.3636 i([h1,i,h2h3,j]) 0.5 -15.3636 c} 0.5 -15.3636 irp3 0.5 -15.3636 data8 0.5 -15.3636 84.81 0.5 -15.3636 kaut 0.5 -15.3636 +0.36 0.5 -15.3636 problemof 0.5 -15.3636 2prp 0.5 -15.3636 exaggerated 0.5 -15.3636 ¯t: 0.5 -15.3636 overpredicts 0.5 -15.3636 α(w) 0.5 -15.3636 uninformative, 0.5 -15.3636 inconclusive, 0.5 -15.3636 msmpr) 0.5 -15.3636 confidence: 0.5 -15.3636 andres 0.5 -15.3636 (aslin 0.5 -15.3636 mamit 0.07 -15.3652 tasks such 0.333333 -15.3657 allow, 0.333333 -15.3657 36.54 0.333333 -15.3657 (klein, 0.333333 -15.3657 connectivity, 0.2 -15.3661 “different 0.5 -15.3667 distinguishbetween 0.5 -15.3667 1,345 0.5 -15.3667 77.07 0.5 -15.3667 phrases” 0.5 -15.3667 δuj 0.5 -15.3667 11.90 0.5 -15.3667 documentand 0.5 -15.3667 35.46 0.5 -15.3667 0.91, 0.5 -15.3667 ∼{x} 0.5 -15.3667 -8 0.5 -15.3667 stj 0.5 -15.3667 (come 0.5 -15.3667 calculate, 0.5 -15.3667 55.82 0.5 -15.3667 interlinked 0.5 -15.3667 (allport 0.5 -15.3667 noun”: 0.5 -15.3667 dcm/p´olya 0.5 -15.3667 31.87 0.5 -15.3667 microblogs: 0.5 -15.3667 voiced, 0.5 -15.3667 bathroom 0.5 -15.3667 49.55 0.5 -15.3667 rejections 0.5 -15.3667 permanently 0.5 -15.3667 jijkoun 0.5 -15.3667 91.94 0.5 -15.3667 (prior 0.5 -15.3667 lenzo 0.5 -15.3667 8,814 0.5 -15.3667 (jeong 0.333333 -15.3677 target-based 0.333333 -15.3677 floods, 0.333333 -15.3677 convergence; 0.333333 -15.3677 1953) 0.333333 -15.3677 0.333333 -15.3677 criticize, 0.0729167 -15.3678 brief 0.25 -15.3683 xia 0.333333 -15.3698 involuntary 0.333333 -15.3698 situates 0.333333 -15.3698 voice: 0.333333 -15.3698 51.24 0.333333 -15.3698 ∂θ 0.5 -15.3698 ∑== 0.5 -15.3698 97.15 0.5 -15.3698 “being 0.5 -15.3698 filter: 0.5 -15.3698 (values 0.5 -15.3698 f−1 0.5 -15.3698 41.76 0.5 -15.3698 nsubjpass 0.5 -15.3698 (ldcrf), 0.5 -15.3698 1,537 0.5 -15.3698 ofwj 0.5 -15.3698 ¯hni 0.5 -15.3698 (eagle, 0.5 -15.3698 goes, 0.5 -15.3698 miami, 0.5 -15.3698 frau 0.5 -15.3698 15.30 0.5 -15.3698 “srl 0.5 -15.3698 word” 0.5 -15.3698 47.50 0.5 -15.3698 description/examples 0.5 -15.3698 kucera, 0.5 -15.3698 “medium”, 0.5 -15.3698 [he]a0 0.5 -15.3698 (hdp; 0.5 -15.3698 .tdi 0.5 -15.3698 125k 0.5 -15.3698 girls, 0.5 -15.3698 subtle issues, 0.5 -15.3698 52.13 0.5 -15.3698 mixed-initiative, 0.5 -15.3698 discussion) 0.5 -15.3698 reassuring 0.5 -15.3698 “other”, 0.5 -15.3698 aptness 0.5 -15.3698 q0, 0.166667 -15.3699 boulder 0.0201394 -15.37 2010). 0.333333 -15.3719 “sxsw” 0.333333 -15.3719 θa124ma124d 0.333333 -15.3719 “nn 0.333333 -15.3719 gary 0.333333 -15.3719 boston” 0.5 -15.3729 yjk 0.5 -15.3729 parikh 0.5 -15.3729 sv1i 0.5 -15.3729 mark-up, 0.5 -15.3729 37.16 0.5 -15.3729 utrad(s) 0.5 -15.3729 eparl+nc+1091 0.5 -15.3729 “congratulate” 0.5 -15.3729 hang 0.5 -15.3729 85.98 0.5 -15.3729 gaume 0.5 -15.3729 conducive 0.5 -15.3729 %move: 0.5 -15.3729 5.25 0.5 -15.3729 arrives, 0.5 -15.3729 32.7% 0.5 -15.3729 “over 0.5 -15.3729 sidelines 0.5 -15.3729 24.62 0.5 -15.3729 hospitals, 0.5 -15.3729 s¨oderwalls 0.5 -15.3729 sourc 0.5 -15.3729 antecedent’s 0.5 -15.3729 phase), 0.5 -15.3729 generates, 0.5 -15.3729 wl1 0.5 -15.3729 ontario 0.5 -15.3729 58.77 0.5 -15.3729 +β) 0.5 -15.3729 word-space 0.5 -15.3729 1085 0.5 -15.3729 toasty 0.5 -15.3729 sr(r1) 0.5 -15.3729 userusage[i] 0.5 -15.3729 (parsers) 0.5 -15.3729 sigk 0.5 -15.3729 (50m), 0.5 -15.3729 w-skw 0.333333 -15.3739 {a 0.333333 -15.3739 umpire 0.333333 -15.3739 ―the 0.333333 -15.3739 0.054 0.333333 -15.3739 ge) 0.333333 -15.3739 brother) 0.333333 -15.3739 “prior 0.333333 -15.376 contentonly 0.333333 -15.376 (remaining 0.333333 -15.376 772 0.333333 -15.376 ds-based 0.333333 -15.376 de-facto 0.333333 -15.376 ¯w 0.333333 -15.376 confluences 0.5 -15.376 models1 0.5 -15.376 siemens 0.5 -15.376 affirmed 0.5 -15.376 20.60 0.5 -15.376 non-angry, 0.5 -15.376 “unnecessary” 0.5 -15.376 58.69 0.5 -15.376 1134 0.5 -15.376 baumann 0.5 -15.376 “user” 0.5 -15.376 imbalances 0.5 -15.376 adda 0.5 -15.376 twet, 0.5 -15.376 90.06 0.5 -15.376 vbg” 0.5 -15.376 c(w,n 0.5 -15.376 courtroom 0.5 -15.376 tradeoff; 0.5 -15.376 20.31 0.5 -15.376 overlap metric. 0.5 -15.376 backchannelpreceding 0.5 -15.376 rainer 0.5 -15.376 polling 0.5 -15.376 155.7 0.5 -15.376 500m, 0.5 -15.376 respons´avel 0.5 -15.376 c-ldaesel,+ 0.5 -15.376 andreevskaia, 0.5 -15.376 imagebased 0.5 -15.376 fillers: 0.5 -15.376 ιb 0.5 -15.376 15.92 0.5 -15.376 jripχ2 0.5 -15.376 popu+context 0.5 -15.376 tracts 0.5 -15.376 game” 0.5 -15.376 kashmir 0.5 -15.376 simonetta 0.5 -15.376 θq, 0.5 -15.376 ‘hoping 0.5 -15.376 v:ncsubj:n 0.5 -15.376 pegar 0.5 -15.376 (finn 0.5 -15.376 examinations, 0.5 -15.376 ti; 0.0744681 -15.3765 reason, 0.181818 -15.3765 of acl-08: 0.25 -15.3776 +0 0.0182161 -15.3777 model. 0.333333 -15.3781 bold-face 0.333333 -15.3781 “daughter”, 0.333333 -15.3781 37.63 0.333333 -15.3781 84.03 0.5 -15.3791 abrir 0.5 -15.3791 schilder 0.5 -15.3791 (daft 0.5 -15.3791 argadj, 0.5 -15.3791 uprimei 0.5 -15.3791 [haghighi 0.5 -15.3791 collaborative, 0.5 -15.3791 are” 0.5 -15.3791 aizawa 0.5 -15.3791 construct, 0.5 -15.3791 (ogrl) 0.5 -15.3791 0.8853 0.5 -15.3791 “keep” 0.5 -15.3791 period), 0.5 -15.3791 tree-bank 0.5 -15.3791 se1more 0.5 -15.3791 (until 0.5 -15.3791 (ss 0.5 -15.3791 vertices: 0.5 -15.3791 kidnapping, 0.5 -15.3791 lebanan 0.5 -15.3791 (fernandez 0.5 -15.3791 figure 13: 0.5 -15.3791 http://wapiti.limsi.fr 0.5 -15.3791 ramaswamy 0.5 -15.3791 str+ugr+met 0.5 -15.3791 86.92 0.5 -15.3791 (link, 0.5 -15.3791 behrang 0.5 -15.3791 amr 0.5 -15.3791 burden, 0.333333 -15.3801 (r1) 0.333333 -15.3801 ‘vice 0.333333 -15.3801 end-of-turn 0.333333 -15.3801 (4)⇐⇒ 0.333333 -15.3801 era, 0.333333 -15.3801 wipo-alpha 0.333333 -15.3801 swirl 0.333333 -15.3801 ω(θ) 0.333333 -15.3801 “general 0.25 -15.3807 wet 0.117647 -15.3815 nullnull null 0.129032 -15.3822 (lin et 0.333333 -15.3822 adik, 0.333333 -15.3822 <.05 0.333333 -15.3822 “black 0.333333 -15.3822 train) 0.333333 -15.3822 (normalized 0.333333 -15.3822 wars 0.5 -15.3822 argmaxe 0.5 -15.3822 1992): 0.5 -15.3822 (people, 0.5 -15.3822 z(e,f) 0.5 -15.3822 ‹q 0.5 -15.3822 94.98 0.5 -15.3822 sim(si) 0.5 -15.3822 84.03 (83.10/84.99) 0.5 -15.3822 64) 0.5 -15.3822 honor 0.5 -15.3822 clarkede 0.5 -15.3822 29.04 0.5 -15.3822 inflected, 0.5 -15.3822 comon, 0.5 -15.3822 non-alignment 0.5 -15.3822 78.79 0.5 -15.3822 raw+cm 0.5 -15.3822 longevity, 0.5 -15.3822 37.99 0.5 -15.3822 73.07 0.5 -15.3822 prp-prp→ich 0.5 -15.3822 δt,tprime 0.5 -15.3822 o([h1,i,h2h3,j]) 0.5 -15.3822 talker 0.5 -15.3822 out; 0.5 -15.3822 yhxh 0.5 -15.3822 2.29 0.5 -15.3822 custom, 0.5 -15.3822 (training+development). 0.5 -15.3822 15.10 0.5 -15.3822 88.20 0.5 -15.3822 expansion:conjunction 0.5 -15.3822 minute) 0.5 -15.3822 θr, 0.5 -15.3822 (seconds), 0.5 -15.3822 € 0.5 -15.3822 -4-2 0.5 -15.3822 transformedto 0.5 -15.3822 prospective, 0.5 -15.3822 phylogenies 0.5 -15.3822 syntactic) 0.5 -15.3822 ‘cause 0.5 -15.3822 mountains, 0.5 -15.3822 ˆri,j 0.333333 -15.3843 ciuli 0.333333 -15.3843 a3) 0.333333 -15.3843 adjunction, 0.333333 -15.3843 u1: 0.333333 -15.3843 pran 0.333333 -15.3843 (xp 0.5 -15.3853 [t]φbφa 0.5 -15.3853 26.02 0.5 -15.3853 seo 0.5 -15.3853 [all], 0.5 -15.3853 norton 0.5 -15.3853 (mykowiecka 0.5 -15.3853 57.01 0.5 -15.3853 scared 0.5 -15.3853 (χ2(2) 0.5 -15.3853 θlq, 0.5 -15.3853 37.80 0.5 -15.3853 dog) 0.5 -15.3853 61.71 0.5 -15.3853 lex-ex 0.5 -15.3853 likelihood(s,o) 0.5 -15.3853 18.88 0.5 -15.3853 mammoth 0.5 -15.3853 53.52 0.5 -15.3853 strung 0.5 -15.3853 otherwords,a’s 0.5 -15.3853 0,1,2,3 0.5 -15.3853 ˆβ 0.5 -15.3853 incarceramento 0.5 -15.3853 cyberinfrastructure 0.5 -15.3853 85.75 0.5 -15.3853 50p, 0.5 -15.3853 trigger” 0.5 -15.3853 puebla, 0.5 -15.3853 38.76 0.5 -15.3853 t′i 0.5 -15.3853 gum 0.5 -15.3853 (aic), 0.5 -15.3853 plan-elab, 0.5 -15.3853 noun-verb-noun 0.5 -15.3853 btree 0.5 -15.3853 28.37 0.5 -15.3853 r-rank 0.5 -15.3853 (zwicky, 0.5 -15.3853 49.17 0.2 -15.386 np@2null1null2] 0.333333 -15.3864 caps 0.333333 -15.3864 1797 0.333333 -15.3864 tasked 0.333333 -15.3864 →c 0.333333 -15.3864 argm 0.333333 -15.3864 ψ(x,y) 0.2 -15.3872 horizon 0.333333 -15.3885 “garbage 0.333333 -15.3885 (richards 0.333333 -15.3885 maximize: 0.333333 -15.3885 [of 0.333333 -15.3885 lad 0.333333 -15.3885 2(2), 0.333333 -15.3885 76.17% 0.333333 -15.3885 act-r, 0.333333 -15.3885 “who” 0.333333 -15.3885 lineage 0.333333 -15.3885 mel 0.333333 -15.3885 arc(h,m) 0.5 -15.3885 amba 0.5 -15.3885 nhdp 0.5 -15.3885 test), 0.5 -15.3885 6.4% 0.5 -15.3885 82.22 0.5 -15.3885 topicstotal 0.5 -15.3885 61.70 0.5 -15.3885 affording 0.5 -15.3885 town-hall 0.5 -15.3885 bleumt02 0.5 -15.3885 w. bush” 0.5 -15.3885 project), (fp7/2011-2014) 0.5 -15.3885 74.16 0.5 -15.3885 insertion: 0.5 -15.3885 91.70 0.5 -15.3885 6.78 0.5 -15.3885 20.81 0.5 -15.3885 wouldnot 0.5 -15.3885 opteron 0.5 -15.3885 mcc-maximizing 0.5 -15.3885 “history” 0.5 -15.3885 (200) 0.5 -15.3885 20.26 0.5 -15.3885 de=german, 0.5 -15.3885 tridiagonal 0.5 -15.3885 srihari 0.5 -15.3885 ceptual 0.5 -15.3885 1872 0.5 -15.3885 ii’ 0.5 -15.3885 clearest 0.5 -15.3885 (special 0.5 -15.3885 whoever 0.5 -15.3885 1,056 0.5 -15.3885 legs, 0.5 -15.3885 “zu 0.5 -15.3885 1986): 0.5 -15.3885 86.55 0.5 -15.3885 ji+1 0.5 -15.3885 bracketrightbiggbracketleftbigga 0.5 -15.3885 pe(t) 0.5 -15.3885 somebody) 0.5 -15.3885 74.77 0.5 -15.3885 onwards 0.5 -15.3885 1-5, 0.333333 -15.3906 followup 0.333333 -15.3906 “sarah 0.333333 -15.3906 28-32 0.333333 -15.3906 danielle 0.333333 -15.3906 n,v,s,p 0.333333 -15.3906 ∈y 0.333333 -15.3906 26.5% 0.333333 -15.3906 o(nm) 0.0535714 -15.3915 setting, 0.5 -15.3916 αu 0.5 -15.3916 fruit) 0.5 -15.3916 ahmad, 0.5 -15.3916 shift–reduce 0.5 -15.3916 amazon, 0.5 -15.3916 (being 0.5 -15.3916 configuring 0.5 -15.3916 (dev), 0.5 -15.3916 15.38 0.5 -15.3916 idf(w) 0.5 -15.3916 surprisingly) 0.5 -15.3916 ‘#’ 0.5 -15.3916 (probabilities) 0.5 -15.3916 nsz 0.5 -15.3916 (1957), 0.5 -15.3916 47.65 0.5 -15.3916 court→made 0.5 -15.3916 21.24 0.5 -15.3916 76.32 0.5 -15.3916 k+ 0.5 -15.3916 banse, 0.5 -15.3916 t32, 0.5 -15.3916 pirt(w) 0.5 -15.3916 1unless 0.5 -15.3916 34.24 0.5 -15.3916 casillas 0.5 -15.3916 context-restricted 0.5 -15.3916 opposes 0.5 -15.3916 91.41 0.5 -15.3916 w∈wd 0.5 -15.3916 bathroom” 0.5 -15.3916 n(¯x−(i,j)negationslash=∅)+α, 0.5 -15.3916 2793 0.5 -15.3916 saffran, 0.5 -15.3916 (morrissey, 0.5 -15.3916 lob 0.5 -15.3916 gradientbased 0.5 -15.3916 1epsilon12. 0.0759494 -15.3921 r f 0.333333 -15.3927 trianglerighttriangleleft1,2 0.333333 -15.3927 pascale 0.333333 -15.3927 pmi-ir 0.333333 -15.3927 determine: 0.333333 -15.3927 wfsts, 0.117647 -15.3934 (gao 0.0650407 -15.3936 hypothesize 0.333333 -15.3948 110m 0.333333 -15.3948 orthonormal 0.333333 -15.3948 sneath 0.333333 -15.3948 australasian 0.333333 -15.3948 frequency; 0.333333 -15.3948 pmi(x,y) 0.5 -15.3948 baseline(en): 0.5 -15.3948 verbalised 0.5 -15.3948 rithms 0.5 -15.3948 “doctor”, 0.5 -15.3948  0.5 -15.3948 sašo 0.5 -15.3948 hirschman, 0.5 -15.3948 8(b) 0.5 -15.3948 promote, 0.5 -15.3948 “hidden 0.5 -15.3948 (svcs) 0.5 -15.3948 18.59 0.5 -15.3948 vbd, 0.5 -15.3948 11.40 0.5 -15.3948 smt-like 0.5 -15.3948 −nawn 0.5 -15.3948 (sibling 0.5 -15.3948 “showed”, 0.5 -15.3948 µ(ν), 0.5 -15.3948 25.43 0.5 -15.3948 hmm 0.752 0.5 -15.3948 robaldo 0.5 -15.3948 (chunks, 0.5 -15.3948 “commercial 0.5 -15.3948 p(oi 0.5 -15.3948 fighting 0.5 -15.3948 2.” 0.5 -15.3948 plain-text, 0.5 -15.3948 omissions 0.5 -15.3948 slid 0.5 -15.3948 them-th 0.5 -15.3948 chameleon 0.5 -15.3948 79.9% 0.5 -15.3948 religions 0.5 -15.3948 boat, 0.0892857 -15.3951 way: 0.333333 -15.3969 adaptors 0.333333 -15.3969 anni 0.333333 -15.3969 0.333333 -15.3969 ‘and’ 0.333333 -15.3969 consistsof 0.333333 -15.3969 computerized 0.333333 -15.3969 39.55 0.333333 -15.3969 89.55 0.333333 -15.3969 nasa 0.333333 -15.3969 outside, 0.5 -15.3979 ∧before(x[1],x[2])∧return(x[2]) 0.5 -15.3979 baselines) 0.5 -15.3979 directly) 0.5 -15.3979 noth, 0.5 -15.3979 x(i,j) 0.5 -15.3979 0.68, 0.5 -15.3979 ε] 0.5 -15.3979 “johnny 0.5 -15.3979 svm), 0.5 -15.3979 cropped 0.5 -15.3979 1151 0.5 -15.3979 48.34 0.5 -15.3979 (res) 0.5 -15.3979 mansion, 0.5 -15.3979 19.46 0.5 -15.3979 fluency; 0.5 -15.3979 52.6% 0.5 -15.3979 fuxman 0.5 -15.3979 fl2 0.5 -15.3979 newsweek 0.5 -15.3979 stylized 0.5 -15.3979 55.56 0.5 -15.3979 thank, 0.5 -15.3979 34.07 0.5 -15.3979 (αins 0.5 -15.3979 coefficient) 0.5 -15.3979 feature-adapted 0.5 -15.3979 70.18 0.5 -15.3979 sumarization 0.5 -15.3979 supermarket 0.5 -15.3979 mccrae 0.5 -15.3979 georgian 0.5 -15.3979 (vi,v 0.5 -15.3979 19.73 0.5 -15.3979 75.79 0.5 -15.3979 wprime, 0.5 -15.3979 892 0.5 -15.3979 sdp: 0.333333 -15.399 alocalisation 0.333333 -15.399 ˆpintu 0.333333 -15.399 slot’s 0.333333 -15.399 wordnet-2 0.333333 -15.399 misread 0.0980392 -15.4002 * * 0.5 -15.4011 wieder 0.5 -15.4011 homonymy, 0.5 -15.4011 d(a) 0.5 -15.4011 mt05baseline 0.5 -15.4011 30.1% 0.5 -15.4011 color), 0.5 -15.4011 devoir 0.5 -15.4011 mei-hua 0.5 -15.4011 river( 0.5 -15.4011 glyphs, 0.5 -15.4011 ut(i) 0.5 -15.4011 oo-rel, 0.5 -15.4011 -0.25 0.5 -15.4011 ‘#’, 0.5 -15.4011 (malt 0.5 -15.4011 ckj 0.5 -15.4011 2,707 0.5 -15.4011 comedy, 0.5 -15.4011 migraine 0.5 -15.4011 −20 0.5 -15.4011 salah 0.5 -15.4011 adventures 0.5 -15.4011 "paid" 0.5 -15.4011 (weimer 0.5 -15.4011 jokes, 0.5 -15.4011 coordination: 0.5 -15.4011 traduction 0.5 -15.4011 speaker” 0.5 -15.4011 2.50 0.5 -15.4011 idf-weighted 0.5 -15.4011 attila 0.5 -15.4011 hu august 0.5 -15.4011 91.32 0.5 -15.4011 limin 0.5 -15.4011 wt1 0.5 -15.4011 17.96 0.5 -15.4011 statisticson 0.5 -15.4011 “moscow” 0.5 -15.4011 s[wq]/(s[q]/np) 0.5 -15.4011 vinyl 0.5 -15.4011 nificantly 0.5 -15.4011 (oh, 0.5 -15.4011 22,187 0.166667 -15.4012 decade 0.1 -15.4012 approach was 0.333333 -15.4012 37.02% 0.333333 -15.4012 37.75 0.333333 -15.4012 digit] 0.0538922 -15.4013 been proposed 0.333333 -15.4033 improvementin 0.333333 -15.4033 18.32 0.333333 -15.4033 ‘most 0.333333 -15.4033 naturally represented 0.333333 -15.4033 “battery 0.333333 -15.4033 kanji 0.333333 -15.4033 finegrained, 0.333333 -15.4033 “completely 0.333333 -15.4033 wisconsin 0.333333 -15.4033 teen 0.333333 -15.4033 sentenceis 0.333333 -15.4033 (90% 0.333333 -15.4033 blurry 0.142857 -15.4039 performance between 0.5 -15.4043 slight, 0.5 -15.4043 mtd 0.5 -15.4043 lel 0.5 -15.4043 cousin 0.5 -15.4043 cyc, 0.5 -15.4043 (carl 0.5 -15.4043 •a 0.5 -15.4043 *robot: 0.5 -15.4043 pag1 0.5 -15.4043 29.51 0.5 -15.4043 5,7and 0.5 -15.4043 (binding) 0.5 -15.4043 9.35 0.5 -15.4043 kaufen, 0.5 -15.4043 yt− 0.5 -15.4043 pp-clr 0.5 -15.4043 sporadic, 0.5 -15.4043 (sparse) 0.5 -15.4043 user-system 0.5 -15.4043 nullrst 0.5 -15.4043 crucial: 0.5 -15.4043 (million) 0.5 -15.4043 timebank: 0.5 -15.4043 burfoot, 0.5 -15.4043 intrasentence 0.5 -15.4043 institutionalization, 0.5 -15.4043 dice(y,tj) 0.5 -15.4043 awesome 0.5 -15.4043 noticable 0.5 -15.4043 defensive 0.5 -15.4043 birds, 0.5 -15.4043 92.38 0.5 -15.4043 tripadvisor 0.5 -15.4043 badanie ‘examination’ 0.5 -15.4043 ct[1] 0.5 -15.4043 76k 0.5 -15.4043 rally 0.5 -15.4043 monocyte 0.5 -15.4043 geometrical, 0.5 -15.4043 6.32 0.5 -15.4043 33.64 0.5 -15.4043 1039 0.333333 -15.4054 concessive 0.333333 -15.4054 persisted 0.333333 -15.4054 all-purpose 0.0382353 -15.4058 is then 0.333333 -15.4075 771 0.333333 -15.4075 u.s.a. 0.333333 -15.4075 totalling 0.333333 -15.4075 ˆx2, 0.5 -15.4075 p(w)q(w) 0.5 -15.4075 sp-token 0.5 -15.4075 spos, 0.5 -15.4075 52.21 0.5 -15.4075 92.24 0.5 -15.4075 score(·) 0.5 -15.4075 andrade 0.5 -15.4075 1,749 0.5 -15.4075 (‘control’) 0.5 -15.4075 bmore 0.5 -15.4075 newsarticle 0.5 -15.4075 i2 0.5 -15.4075 71.18 0.5 -15.4075 96.50 0.5 -15.4075 16.24 0.5 -15.4075 non-erasing 0.5 -15.4075 1-270 0.5 -15.4075 fro 0.5 -15.4075 ngd-google-w 0.5 -15.4075 91.60 0.5 -15.4075 80.11 0.5 -15.4075 overthe 0.5 -15.4075 labelsthe 0.5 -15.4075 “fair” 0.5 -15.4075 snowball,cnlp 0.5 -15.4075 1356 0.5 -15.4075 sub-categories: 0.5 -15.4075 explained: 0.5 -15.4075 ofpl 0.5 -15.4075 36.45 0.5 -15.4075 parsing’ 0.5 -15.4075 alias, 0.5 -15.4075 gradients: 0.5 -15.4075 sentence.1 0.5 -15.4075 tag-set, 0.5 -15.4075 pitch-accented 0.5 -15.4075 "killed" 0.5 -15.4075 clemens 0.5 -15.4075 15.55%) 0.5 -15.4075 counts” 0.5 -15.4075 1405 0.5 -15.4075 0.29) 0.5 -15.4075 viewer, 0.5 -15.4075 insightsdeveloped 0.0552147 -15.408 have also 0.136364 -15.4093 art. 0.166667 -15.4097 (soon 0.333333 -15.4097 (density: 0.333333 -15.4097 tagme 0.333333 -15.4097 g(y) 0.333333 -15.4097 (connected 0.333333 -15.4097 820 0.333333 -15.4097 vii 0.333333 -15.4097 28.58 0.333333 -15.4097 normalization: 0.333333 -15.4097 1965), 0.0892857 -15.4099 ± 0.0634921 -15.4104 question. 0.5 -15.4107 (cause 0.5 -15.4107 min(c, 0.5 -15.4107 dare, 0.5 -15.4107 backward” 0.5 -15.4107 s0r 0.5 -15.4107 gorkana 0.5 -15.4107 (maximal 0.5 -15.4107 xor 0.5 -15.4107 (catalog 0.5 -15.4107 +4k 0.5 -15.4107 chafe, 0.5 -15.4107 95.4% 0.5 -15.4107 bars) 0.5 -15.4107 3424 0.5 -15.4107 trial-averaged 0.5 -15.4107 svmrank, 0.5 -15.4107 spark 0.5 -15.4107 galician 0.5 -15.4107 exemplary 0.5 -15.4107 ≻x,φ′,z= 0.5 -15.4107 5000, 0.5 -15.4107 g2: 0.5 -15.4107 τ(v) 0.5 -15.4107 0.040 0.5 -15.4107 i∈ 0.5 -15.4107 follower, 0.5 -15.4107 p+a 0.5 -15.4107 (the) 0.5 -15.4107 imprecise 0.5 -15.4107 +0.29 0.5 -15.4107 circuits 0.5 -15.4107 goo 0.5 -15.4107 concept-level 0.5 -15.4107 kübler, 0.5 -15.4107 argumenttype 0.5 -15.4107 f`e 0.5 -15.4107 {com 0.5 -15.4107 ‡national 0.5 -15.4107 parents(n) 0.5 -15.4107 14.13 0.333333 -15.4118 sloppy 0.333333 -15.4118 in+out 0.333333 -15.4118 ¯q1 0.333333 -15.4118 69.66 0.333333 -15.4118 ‘by 0.333333 -15.4118 aller 0.333333 -15.4118 commit(pi,dn 0.0810811 -15.4122 volume 37, 0.2 -15.4127 denies 0.5 -15.4139 tokenrole 0.5 -15.4139 (out-of-domain) 0.5 -15.4139 p0” 0.5 -15.4139 85.22 0.5 -15.4139 34.97 0.5 -15.4139 (randles 0.5 -15.4139 non-alphabetic 0.5 -15.4139 84.69 0.5 -15.4139 paani 0.5 -15.4139 “preferential 0.5 -15.4139 xiaoyun 0.5 -15.4139 5m, 0.5 -15.4139 mentionsare 0.5 -15.4139 “talk.religion.misc” 0.5 -15.4139 furu 0.5 -15.4139 cost(ri) 0.5 -15.4139 (postech 0.5 -15.4139 “stuttgarter 0.5 -15.4139 ...), 0.5 -15.4139 28.53 0.5 -15.4139 4based 0.5 -15.4139 ]: 0.5 -15.4139 (quarterback, 0.5 -15.4139 practicaly 0.5 -15.4139 relationship; 0.5 -15.4139 stemsuffix 0.5 -15.4139 romance, 0.5 -15.4139 pulses, 0.5 -15.4139 (gud), 0.5 -15.4139 multiparticipant 0.5 -15.4139 optimiz 0.5 -15.4139 79.60 0.5 -15.4139 83.63% 0.5 -15.4139 unbound 0.5 -15.4139 lattice-type-r, 0.5 -15.4139 ¯θ 0.5 -15.4139 slice) 0.5 -15.4139 lines), 0.5 -15.4139 date/number 0.5 -15.4139 hashing (weiss 0.5 -15.4139 2.06 0.5 -15.4139 0.1284 0.5 -15.4139 grammaticalrelation 0.5 -15.4139 rtls 0.333333 -15.414 right-arcl 0.333333 -15.414 c7, 0.333333 -15.414 inequality: 0.333333 -15.414 spanish-, 0.333333 -15.414 ∈rn 0.333333 -15.4161 prefixes: 0.333333 -15.4161 (kruengkrai 0.333333 -15.4161 “normal” 0.0460251 -15.4166 with its 0.102041 -15.4166 italy 0.5 -15.4172 positive  0.5 -15.4172 +summationtextm∈e,n/∈eψr(m,n) 0.5 -15.4172 preprocessingresults 0.5 -15.4172 super-, 0.5 -15.4172  from 0.5 -15.4172 onb 0.5 -15.4172 climbed 0.5 -15.4172 pii,j 0.5 -15.4172 project; 0.5 -15.4172 “enemy 0.5 -15.4172 ot-rel, 0.5 -15.4172 activate-labeled 0.5 -15.4172 triangledownsld 0.5 -15.4172 0.144 0.5 -15.4172 j,k 0.5 -15.4172 (kinyon 0.5 -15.4172 1,321 0.5 -15.4172 “popular” 0.5 -15.4172 (cause) 0.5 -15.4172 parser7 0.5 -15.4172 nullr(null) 0.5 -15.4172 1406 0.5 -15.4172 natarajan 0.5 -15.4172 1nullnullnullnullnull 0.5 -15.4172 72.5% 0.5 -15.4172 focused, 0.5 -15.4172 performanceon 0.5 -15.4172 32.30 0.5 -15.4172 2453 0.5 -15.4172 93.05 0.5 -15.4172 (loc), 0.5 -15.4172 quartile, 0.5 -15.4172 1470 0.5 -15.4172 fotinea, 0.5 -15.4172 cornerback 0.5 -15.4172 jaime 0.5 -15.4172 (optional) 0.5 -15.4172 παλμοκωδική 0.5 -15.4172 44m 0.5 -15.4172 φ(e[c(z)] 0.5 -15.4172 son, 0.5 -15.4172 atnullnullnullnullnull 0.5 -15.4172 knlm 0.5 -15.4172 hema 0.333333 -15.4183 relation-independent 0.333333 -15.4183 sub-intervals 0.333333 -15.4183 [general] 0.333333 -15.4183 (600 0.333333 -15.4183 functorargument 0.333333 -15.4183 eukaryote 0.333333 -15.4183 precision= 0.15 -15.4191 former, 0.181818 -15.4192 reference set. 0.333333 -15.4204 hierok−1(s,σ,a) 0.333333 -15.4204 averaging, 0.333333 -15.4204 bmj 0.333333 -15.4204 1378 0.333333 -15.4204 53.81 0.111111 -15.4204 xylitol 0.5 -15.4204 h´el`ene 0.5 -15.4204 “missing 0.5 -15.4204 dobra, 0.5 -15.4204 rack 0.5 -15.4204 institution, 0.5 -15.4204 signs, 0.5 -15.4204 guilt, 0.5 -15.4204 41.20 0.5 -15.4204 a“¸@ 0.5 -15.4204 macqueen, 0.5 -15.4204 record/document 0.5 -15.4204 (cd), 0.5 -15.4204 marciniak 0.5 -15.4204 387k 0.5 -15.4204 andr(el) 0.5 -15.4204 “shotgun” 0.5 -15.4204 20p, 0.5 -15.4204 #tweetdebate 0.5 -15.4204 hruschka, 0.5 -15.4204 +0.18 0.5 -15.4204 0.3058 0.5 -15.4204 48.27 0.5 -15.4204 prod-rule[stanford]*(n=all) 0.5 -15.4204 1302 0.5 -15.4204 untouched, 0.5 -15.4204 abandoning 0.5 -15.4204 undefined, 0.5 -15.4204 +0.06 0.5 -15.4204 理 0.5 -15.4204 3.68 0.5 -15.4204 stative 0.5 -15.4204 projects: 0.5 -15.4204 (day 0.5 -15.4204 cost; 0.5 -15.4204 (sq): 0.5 -15.4204 troubled 0.5 -15.4204 67.28 0.5 -15.4204 zooms 0.5 -15.4204 descriptions; 0.5 -15.4204 fehe 0.5 -15.4204 multi-character 0.5 -15.4204 scoreson 0.5 -15.4204 rosy 0.5 -15.4204 grammatik 0.5 -15.4204 nava 0.333333 -15.4226 siga 0.333333 -15.4226 96.42 0.333333 -15.4226 [np-dt-jj-*] 0.333333 -15.4226 eachof 0.333333 -15.4226 entrar 0.333333 -15.4226 nyt95 0.333333 -15.4226 (noisy 0.333333 -15.4226 proponents 0.333333 -15.4226 expθ 0.333333 -15.4226 trans–ptb 0.222222 -15.4234 (regardless 0.5 -15.4237 10summationdisplay 0.5 -15.4237 andstrube 0.5 -15.4237 2.66ghz 0.5 -15.4237 −→p 0.5 -15.4237 pre-computing 0.5 -15.4237 workshop’s 0.5 -15.4237 82.20% 0.5 -15.4237 senting 0.5 -15.4237 containat 0.5 -15.4237 uma 0.5 -15.4237 threat, 0.5 -15.4237 combine: 0.5 -15.4237 washington’s turing 0.5 -15.4237 defining 0.5 -15.4237 nourishment 0.5 -15.4237 (thematic 0.5 -15.4237 units; 0.5 -15.4237 state-side 0.5 -15.4237 prone, 0.5 -15.4237 12.22 0.5 -15.4237 potato 0.5 -15.4237 referent’s 0.5 -15.4237 strategies’ 0.5 -15.4237 “where 0.5 -15.4237 touched∩a 0.5 -15.4237 xylitol’s 0.5 -15.4237 trec-2004 0.5 -15.4237 full-order 0.5 -15.4237 kathrin 0.5 -15.4237 euthanasia 0.5 -15.4237 15.04 0.5 -15.4237 ps: 0.5 -15.4237 24.72 0.5 -15.4237 catherine, 0.5 -15.4237 [dell 0.5 -15.4237 threelink 0.5 -15.4237 implementthree 0.5 -15.4237 43.64% 0.5 -15.4237 armchair 0.333333 -15.4248 [prp 0.333333 -15.4248 figure7 0.333333 -15.4248 1our 0.333333 -15.4248 outputsof 0.333333 -15.4248 jobs”, 0.333333 -15.4248 x∈ 0.0538922 -15.4256 (p 0.5 -15.4269 package: 0.5 -15.4269 (does 0.5 -15.4269 arte 0.5 -15.4269 40.43 0.5 -15.4269 init(f,u)1 0.5 -15.4269 wherea 0.5 -15.4269 5.48 0.5 -15.4269 0.5 -15.4269 frontier, 0.5 -15.4269 “stanford”, 0.5 -15.4269 18we 0.5 -15.4269 abstrmax 0.5 -15.4269 crush 0.5 -15.4269 50.79 0.5 -15.4269 (perrault 0.5 -15.4269 757 0.5 -15.4269 18.73 0.5 -15.4269 (2abc). 0.5 -15.4269 1,136 0.5 -15.4269 prod-rule[c&j](n=1000) 0.5 -15.4269 miwa10, 0.5 -15.4269 pole 0.5 -15.4269 (acesed 0.5 -15.4269 (dersc 0.5 -15.4269 99.69 0.5 -15.4269 criteria “e+p*+a*” 0.5 -15.4269 (1976), 0.5 -15.4269 null=nullnullnullnullnullnull(nullnullnull) 0.5 -15.4269 66.28 0.5 -15.4269 3.7m 0.5 -15.4269 ∀(vectorxifollowsr 0.5 -15.4269 rivera, 0.5 -15.4269 selectedas 0.5 -15.4269 non-terminal; 0.5 -15.4269 (pearce, 0.5 -15.4269 pseudo-probability 0.5 -15.4269 dec-all, 0.5 -15.4269 1-role 0.5 -15.4269 tfij 0.5 -15.4269 ht,lscript 0.5 -15.4269 0.370 0.5 -15.4269 zz 0.333333 -15.427 869 0.333333 -15.427 transits 0.333333 -15.427 identified: 0.333333 -15.427 ·) 0.333333 -15.427 o11 0.333333 -15.427 “shot 0.142857 -15.4288 forth. 0.333333 -15.4291 emnlp-conll 0.333333 -15.4291 fish”, 0.333333 -15.4291 yours 0.333333 -15.4291 hypothesesare 0.2 -15.4296 auch 0.5 -15.4302 21.02 0.5 -15.4302 (zubaryeva 0.5 -15.4302 (buitelaar 0.5 -15.4302 strötgen 0.5 -15.4302 wherey 0.5 -15.4302 6-8 0.5 -15.4302 95.85 0.5 -15.4302 w2”. 0.5 -15.4302 16.37 0.5 -15.4302 i−2 0.5 -15.4302 66.95 0.5 -15.4302 3best 0.5 -15.4302 1610 0.5 -15.4302 firstname 0.5 -15.4302 20.40 0.5 -15.4302 minority, 0.5 -15.4302 1418 0.5 -15.4302 80.57 0.5 -15.4302 megyesi 0.5 -15.4302 fully-automatic 0.5 -15.4302 bachorowski, 0.5 -15.4302 92.51 0.5 -15.4302 anand 0.5 -15.4302 knowledge” 0.5 -15.4302 (lu), 0.5 -15.4302 queue; 0.5 -15.4302 4during 0.0356234 -15.4305 rely 0.333333 -15.4313 partha 0.333333 -15.4313 2,600 0.333333 -15.4313 (byrd 0.333333 -15.4313 flavor 0.333333 -15.4313 reddy, 0.0353535 -15.4334 0 0 0.333333 -15.4335 (accuracy) 0.333333 -15.4335 si’s 0.333333 -15.4335 “english 0.333333 -15.4335 computerassisted 0.2 -15.4335 lexical-stylistic 0.5 -15.4335 55.50 0.5 -15.4335 cmu) 0.5 -15.4335 (march 0.5 -15.4335 theprobability 0.5 -15.4335 “universal” 0.5 -15.4335 mercedes 0.5 -15.4335 21.95 0.5 -15.4335 diamond, 0.5 -15.4335 pro-mult 0.5 -15.4335 parenthesis) 0.5 -15.4335 bsorf 0.5 -15.4335 pueblo 0.5 -15.4335 205), 0.5 -15.4335 89.15 0.5 -15.4335 regularised 0.5 -15.4335 ilar 0.5 -15.4335 hpa 0.5 -15.4335 quad-core 0.5 -15.4335 kreuz, 0.5 -15.4335 42.95 0.5 -15.4335 restaurants) 0.5 -15.4335 segmented) 0.5 -15.4335 1977 0.5 -15.4335 hodges, 0.5 -15.4335 φ⊗(n−k) 0.5 -15.4335 adaptations) 0.5 -15.4335 “so-tp” 0.5 -15.4335 ccr 0.5 -15.4335 boswel 0.5 -15.4335 (range: 0.5 -15.4335 8only 0.5 -15.4335 70.34 0.5 -15.4335 “which 0.5 -15.4335 [:ˆ1-*0 0.5 -15.4335 l<0 0.5 -15.4335 niu, 0.5 -15.4335 68.37 0.5 -15.4335 54.65 0.5 -15.4335 formaly, 0.5 -15.4335 “migs” 0.5 -15.4335 time-stamps 0.5 -15.4335 nullaligned 0.5 -15.4335 99.47 0.130435 -15.4338 element. 0.333333 -15.4357 estimate: 0.333333 -15.4357 implementationof 0.333333 -15.4357 (190) 0.0206009 -15.4364 we are 0.25 -15.4368 reduction) 0.5 -15.4368 57.31 0.5 -15.4368 parse2 0.5 -15.4368 7i 0.5 -15.4368 ∀n 0.5 -15.4368 hist, 0.5 -15.4368 87.77 0.5 -15.4368 armenian 0.5 -15.4368 (appelt 0.5 -15.4368 tm-js 0.5 -15.4368 pubmed-scale 0.5 -15.4368 goal), 0.5 -15.4368 (marking 0.5 -15.4368 np0,2 0.5 -15.4368 italics: 0.5 -15.4368 typej 0.5 -15.4368 (malakasiotis, 0.5 -15.4368 pointing, 0.5 -15.4368 41.01 0.5 -15.4368 “when”, 0.5 -15.4368 emotionfelt 0.5 -15.4368 andxj 0.5 -15.4368 2dimensional 0.5 -15.4368 10-25% 0.5 -15.4368 animal” 0.5 -15.4368 olmayan 0.5 -15.4368 1176 0.5 -15.4368 motifs 0.5 -15.4368 8,031 0.5 -15.4368 depthfirst 0.5 -15.4368 c−1 0.5 -15.4368 52.90 0.5 -15.4368 ¯mni 0.5 -15.4368 28.92 0.5 -15.4368 clinton” 0.5 -15.4368 v-beta 0.5 -15.4368 references brown, 0.5 -15.4368 1194 0.5 -15.4368 ‘today’], 0.5 -15.4368 +b 0.5 -15.4368 cm; 0.5 -15.4368 original–original 0.5 -15.4368 direct/projected 0.5 -15.4368 ‘straight’ 0.102564 -15.4369 account, 0.333333 -15.4379 links; 0.333333 -15.4379 [some 0.333333 -15.4379 triphones 0.333333 -15.4379 o(θ) 0.333333 -15.4379 (chapter 0.25 -15.4401 malaria 0.333333 -15.4401 φj 0.333333 -15.4401 1,414 0.333333 -15.4401 72.39 0.333333 -15.4401 cooperativity 0.333333 -15.4401 uλ(y,s,w) 0.333333 -15.4401 (120) 0.333333 -15.4401 engulfed 0.5 -15.4401 +pr+1 0.5 -15.4401 w)) 0.5 -15.4401 (24 0.5 -15.4401 spanje 0.5 -15.4401 textgrounder 0.5 -15.4401 atributed 0.5 -15.4401 nulljnullhj 0.5 -15.4401 zhou1 0.5 -15.4401 square4 0.5 -15.4401  f-­‐score 0.5 -15.4401 µi−1 0.5 -15.4401 column’s 0.5 -15.4401 gretzel, 0.5 -15.4401 0.653 0.5 -15.4401 54m 0.5 -15.4401 (label) 0.5 -15.4401 26.05 0.5 -15.4401 reporters, 0.5 -15.4401 57%, 0.5 -15.4401 strawman 0.5 -15.4401 propbank-based 0.5 -15.4401 multi-purpose 0.5 -15.4401 2058 0.5 -15.4401 82.97 0.5 -15.4401 per:employee 0.5 -15.4401 modifiers: 0.5 -15.4401 bennett, 0.5 -15.4401 θ(t) 0.5 -15.4401 ancestors(n) 0.5 -15.4401 (clarke, 0.5 -15.4401 found; 0.5 -15.4401 peterson, 0.5 -15.4401 1228 0.5 -15.4401 discomfort 0.5 -15.4401 1.00, 0.5 -15.4401 (entailed) 0.5 -15.4401 775 0.5 -15.4401 40.03 0.5 -15.4401 tempo 0.5 -15.4401 (potet 0.5 -15.4401 sensorial 0.5 -15.4401 58.79 0.5 -15.4401 {health, 0.5 -15.4401 weighted: 0.25 -15.4418 θq 0.333333 -15.4423 fables 0.333333 -15.4423 owl: 0.333333 -15.4423 tverb, 0.333333 -15.4423 2x 0.333333 -15.4423 bowler 0.333333 -15.4423 σ0 0.333333 -15.4423 unknowns 0.2 -15.4427 pc-trans 0.5 -15.4434 didn't 0.5 -15.4434 misrecognitions 0.5 -15.4434 73.90 0.5 -15.4434 1/22/2010 0.5 -15.4434 tage 0.5 -15.4434 (45%) 0.5 -15.4434 18-22s 0.5 -15.4434 ‘check 0.5 -15.4434 cyclosporin 0.5 -15.4434 ceauşu 0.5 -15.4434 707 0.5 -15.4434 (dutrey 0.5 -15.4434 “target”, 0.5 -15.4434 well-understood, 0.5 -15.4434 combinesthe 0.5 -15.4434 u(n) 0.5 -15.4434 rmprimenegationslash= 0.5 -15.4434 nasals 0.5 -15.4434 lmin 0.5 -15.4434 first/last 0.5 -15.4434 accurat 0.5 -15.4434 in/on/at/to 0.5 -15.4434 mammals, 0.5 -15.4434 upfront 0.5 -15.4434 kybots, 0.5 -15.4434 atis, 0.5 -15.4434 “internet 0.5 -15.4434 organising 0.5 -15.4434 (down 0.5 -15.4434 +root 0.333333 -15.4445 prp: 0.333333 -15.4445 “reference” 0.333333 -15.4445 spammers, 0.333333 -15.4445 spinach 0.333333 -15.4445 gnue 0.057554 -15.4449 comments. 0.5 -15.4467 (full-short) 0.5 -15.4467 (possible) 0.5 -15.4467 “base 0.5 -15.4467 clumping 0.5 -15.4467 moser 0.5 -15.4467 72.34 0.5 -15.4467 turku09 0.5 -15.4467 belt 0.5 -15.4467 (reference 0.5 -15.4467 dreams 0.5 -15.4467 tianjin 0.5 -15.4467 tr-light-v 0.5 -15.4467 pakistan’s 0.5 -15.4467 pi(s) 0.5 -15.4467 nkjp 0.5 -15.4467 web-based, 0.5 -15.4467 sacred 0.5 -15.4467 (video 0.5 -15.4467 tsvm-b 0.5 -15.4467 (dtd) 0.5 -15.4467 findings; 0.5 -15.4467 pasted 0.5 -15.4467 “two” 0.5 -15.4467 effortlessly 0.5 -15.4467 c(q) 0.5 -15.4467 -subj 0.5 -15.4467 71.9% 0.5 -15.4467 “good”, 0.5 -15.4467 28,835 0.5 -15.4467 0.079 0.5 -15.4467 (µ1 0.5 -15.4467 78.36 0.5 -15.4467 vectorui, 0.5 -15.4467 obtained; 0.5 -15.4467 pt(w) 0.5 -15.4467 infrequent, 0.5 -15.4467 1007 0.25 -15.4468 haifa, 0.25 -15.4468 ?s 0.25 -15.4468 clark’s 0.25 -15.4468 (reichart 0.333333 -15.4468 pc(v,e,o) 0.333333 -15.4468 23.80 0.333333 -15.4468 douglas 0.222222 -15.4483 amherst, 0.25 -15.4484 vectorr1 0.25 -15.4484 tepperman, 0.25 -15.4484 dtd 0.25 -15.4484 intra-class 0.25 -15.4484 kipper 0.25 -15.4484 longrange 0.25 -15.4484 texting, 0.25 -15.4484 immersion 0.25 -15.4484 abduction” 0.25 -15.4484 wikipedia), 0.25 -15.4484 celestial 0.333333 -15.449 m+ 0.333333 -15.449 wir 0.333333 -15.449 milan 0.333333 -15.449 reminded 0.333333 -15.449 indebted 0.333333 -15.449 81.63 0.25 -15.4501 publishers 0.25 -15.4501 s2j 0.25 -15.4501 cause> 0.25 -15.4501 skeptic 0.25 -15.4501 ltt 0.25 -15.4501 manual: 0.25 -15.4501 sariyer, 0.25 -15.4501 zesch, 0.25 -15.4501 schubert, 0.25 -15.4501 können:may 0.2 -15.4507 task). 0.125 -15.4509 multipliers 0.333333 -15.4512 [pp 0.333333 -15.4512 time-to-event 0.333333 -15.4512 record) 0.333333 -15.4512 1365 0.333333 -15.4512 64.97 0.222222 -15.4513 enlarge 0.25 -15.4518 shadow 0.25 -15.4518 %# 0.25 -15.4518 ∆i 0.25 -15.4518 a46 0.25 -15.4535 periphrases 0.25 -15.4535 „c1 0.25 -15.4535 seki 0.25 -15.4535 97.1 0.25 -15.4535 later), 0.333333 -15.4535 moran 0.333333 -15.4535 br-cs 0.333333 -15.4535 s/p 0.333333 -15.4535 “block”, 0.333333 -15.4535 northeast 0.333333 -15.4535 blackstone 0.25 -15.4551 (bellegarda, 0.25 -15.4551 1000k 0.25 -15.4551 milk. 0.333333 -15.4557 i., 0.333333 -15.4557 unhandled 0.333333 -15.4557 exp2 0.333333 -15.4557 (geng 0.333333 -15.4557 θk,i 0.25 -15.4568 “true 0.25 -15.4568 higher: 0.333333 -15.4579 deprivation 0.333333 -15.4579 “typical” 0.333333 -15.4579 43.55 0.25 -15.4585 habits 0.333333 -15.4602 ‘data 0.333333 -15.4602 81.39 0.333333 -15.4602 ∆σ(n1,n2) 0.333333 -15.4602 hotels, 0.333333 -15.4602 thak 0.333333 -15.4602 instrument) 0.333333 -15.4602 nathanael 0.333333 -15.4624 parents] 0.333333 -15.4624 ldc2009t08, 0.333333 -15.4624 event-atoms 0.333333 -15.4624 erik 0.222222 -15.4632 路 0.25 -15.4636 (m-1) 0.25 -15.4636 0.690 0.25 -15.4636 sentiment  0.25 -15.4636 85.74 0.0892857 -15.4636 parenleftbigg 0.333333 -15.4647 simply: 0.333333 -15.4647 cu-sempos-bleu 0.333333 -15.4647 imagination 0.333333 -15.4647 chierchia 0.333333 -15.4647 ple, 0.25 -15.4653 (like) 0.222222 -15.4662 haiti, 0.0339806 -15.4663 used by 0.25 -15.467 brother, 0.25 -15.467 hunt, 0.333333 -15.467 cesta 0.333333 -15.467 mdparser 0.333333 -15.467 wo-pcfg 0.333333 -15.467 -0.47 0.25 -15.4687 ships, 0.333333 -15.4692 1.44 0.333333 -15.4692 navy 0.333333 -15.4692  ast 0.333333 -15.4692 mysore, 0.333333 -15.4692 sparse: 0.333333 -15.4692 conva,b(t) 0.333333 -15.4692 28.13 0.25 -15.4704 “?�(fahui)” 0.333333 -15.4715 (13): 0.333333 -15.4715 1299 0.333333 -15.4715 wid, 0.333333 -15.4715  rev 0.333333 -15.4715 reservations 0.333333 -15.4715 quests 0.25 -15.4721 head(e) 0.25 -15.4721 ˆl(v) 0.25 -15.4738 ga] 0.25 -15.4738 steal 0.333333 -15.4738 [[u]]c 0.333333 -15.4738 ticker 0.333333 -15.4738 cuvplus 0.333333 -15.4738 plays, 0.333333 -15.4738 p(θ 0.333333 -15.4738 iprimenegationslash= 0.333333 -15.4738 extracted: 0.333333 -15.4738 6.38 0.333333 -15.4738 characterise 0.333333 -15.4738 yulan 0.333333 -15.4761 53.58 0.333333 -15.4761 “man 0.25 -15.4772 200 distinct 0.333333 -15.4784 arantza 0.333333 -15.4784 xuan 0.333333 -15.4784 g-index 0.333333 -15.4784 gradient, 0.333333 -15.4784 (zi 0.333333 -15.4784 metamap+ 0.0529412 -15.4788 from these 0.25 -15.479 “by 0.333333 -15.4807 61.05 0.333333 -15.4807 wave 0.333333 -15.4807 paramor 0.136364 -15.481 4.3. 0.0526316 -15.481 has an 0.25 -15.4824 (w1 0.0243902 -15.4827 task, 0.222222 -15.483 level: 0.333333 -15.483 oil) 0.333333 -15.483 ˆτ 0.333333 -15.483 como 0.333333 -15.483 84.59 0.333333 -15.483 addressees, 0.333333 -15.483 surface-form 0.25 -15.4841 cat, 0.25 -15.4841 sprout 0.333333 -15.4853 14.5% 0.333333 -15.4853 label), 0.25 -15.4859 g(i,j) 0.0689655 -15.4871 markov model 0.25 -15.4876 approachto 0.25 -15.4876 tproductdisplay 0.333333 -15.4876 guillaume 0.333333 -15.4876 fr(ns(s1),ns(s2)) 0.333333 -15.4876 prokaryotic 0.333333 -15.4876 arc-length 0.0679612 -15.4885 significant. 0.25 -15.4893 120,000 0.333333 -15.4899 documentsand 0.333333 -15.4899 qm, 0.333333 -15.4899 yprimec 0.333333 -15.4899 auxiliary, 0.222222 -15.4907 das lied 0.25 -15.4911 71.03 0.12 -15.4918 (ner) 0.333333 -15.4922 hypernym, 0.333333 -15.4922 selectors 0.333333 -15.4922 apps, 0.333333 -15.4922 usedin 0.333333 -15.4922 s◦r 0.333333 -15.4922 rural 0.333333 -15.4922 [l 0.333333 -15.4922 embodies 0.25 -15.4928 affordable 0.25 -15.4946 21.8% 0.333333 -15.4946 s(ci) 0.333333 -15.4946 tmi 0.333333 -15.4946 mesages 0.333333 -15.4946 learn: 0.333333 -15.4946 tw 0.333333 -15.4946 63.65 0.333333 -15.4946 s(xp,i,j) 0.333333 -15.4946 homeland 0.333333 -15.4946 coals 0.333333 -15.4946 616 0.222222 -15.4954 (re) 0.25 -15.4963 berv 0.25 -15.4963 benv 0.25 -15.4963 iconix, 0.25 -15.4963 im, 0.333333 -15.4969 [firstname], 0.333333 -15.4969 (backoff) 0.333333 -15.4969 dbpedia, 0.333333 -15.4969 p15 0.333333 -15.4969 tonic 0.25 -15.4981 (2005): 0.25 -15.4981 0.112 0.0660377 -15.4984 if one 0.333333 -15.4992 s[wq] 0.333333 -15.4992 status/response 0.333333 -15.4992 log-probabilities 0.333333 -15.4992 machine (fsm) 0.333333 -15.4992 spaghetti 0.333333 -15.4992 63.41 0.25 -15.4998 1b, 0.222222 -15.5 khapra et 0.166667 -15.5016 bicknell et 0.25 -15.5016 3.2% 0.25 -15.5016 〈arg0〉 0.25 -15.5016 cd(w) 0.25 -15.5016 f(x,y) 0.25 -15.5016 “伞 0.333333 -15.5016 τa 0.333333 -15.5016 81.42 0.333333 -15.5016 divrank, 0.333333 -15.5016 (ungrammatical) 0.333333 -15.5016 85.29 0.333333 -15.5016 βv 0.333333 -15.5016 casa 0.333333 -15.5016 1,326 0.333333 -15.5016 “president 0.333333 -15.5016 deutschen 0.333333 -15.5016 tags): 0.25 -15.5034 directness 0.333333 -15.5039 “p 0.333333 -15.5039 1532 0.333333 -15.5039 q(cj1) 0.130435 -15.5041 contains about 0.333333 -15.5063 1288 0.333333 -15.5063 1(b), 0.333333 -15.5063 maara 0.333333 -15.5086 96.07 0.333333 -15.5086 extrinsic, 0.333333 -15.5086 ev 0.333333 -15.5086 hw(i,j) 0.333333 -15.5086 couldbe 0.333333 -15.5086 454 0.136364 -15.5086 are possible. 0.25 -15.5087 22.18 0.166667 -15.5099 onull 0.0598291 -15.5107 thanks 0.333333 -15.511 induce, 0.333333 -15.511 nasty 0.181818 -15.5118 instances belonging 0.25 -15.5122 0.837 0.333333 -15.5134 exam 0.333333 -15.5134 computethe 0.333333 -15.5134 (secs 0.333333 -15.5134 η) 0.333333 -15.5134 vector-in-context 0.333333 -15.5134 root-f1 0.25 -15.514 (hobbs 0.25 -15.5158 kuzman 0.333333 -15.5158 gesture’s 0.333333 -15.5158 α), 0.333333 -15.5158 calculatethe 0.25 -15.5176 m12 0.046875 -15.518 however, this 0.333333 -15.5181 56/c 0.333333 -15.5181 “external 0.333333 -15.5181 jersey 0.333333 -15.5181 “mean 0.0759494 -15.5192 because there 0.25 -15.5194 27.15 0.25 -15.5194 25.69 0.333333 -15.5205 7.75 0.333333 -15.5205 arcs: 0.333333 -15.5205 gericht 0.333333 -15.5205 “affect” 0.333333 -15.5205 weigand 0.333333 -15.5205 +pattern 0.25 -15.5211 77.3% 0.25 -15.5211 0.25 -15.5211 timeneed(x, 0.2 -15.5215 (dzikovska 0.333333 -15.5229 2>3, 0.333333 -15.5229 componentwise 0.333333 -15.5229 tm(f 0.333333 -15.5229 λns 0.333333 -15.5229 (first, 0.333333 -15.5229 * 0.333333 -15.5229 tions, 0.25 -15.5247 "wuensch" 0.25 -15.5247 2416 0.25 -15.5247 pardo 0.333333 -15.5253 bφ 0.333333 -15.5253 t(f 0.333333 -15.5253 a→α 0.333333 -15.5253 ϕga 0.181818 -15.5262 slot3, 0.25 -15.5265 boys invited 0.2 -15.5272 ∗department 0.333333 -15.5277 nml 0.333333 -15.5277 -th 0.333333 -15.5277 et, 0.333333 -15.5277 biological, 0.333333 -15.5277 gaifman 0.333333 -15.5277 comfort 0.25 -15.5284 core metric 0.333333 -15.5301 boto 0.333333 -15.5301 efthimiou 0.333333 -15.5301 term-by-document 0.333333 -15.5301 21.52 0.333333 -15.5301 æ¸ 0.25 -15.5302 you), 0.136364 -15.5311 collins et 0.333333 -15.5326 (wikipedia 0.333333 -15.5326 martine 0.333333 -15.5326 /0 0.333333 -15.5326 gifts 0.333333 -15.5326 (mt04) 0.333333 -15.535 (⊔) 0.333333 -15.535 [nvda 0.333333 -15.535 availabilities 0.0505618 -15.5358 is available 0.25 -15.5374 3.39 0.25 -15.5374 acquis 0.25 -15.5374 +bf 0.333333 -15.5374 60m 0.333333 -15.5374 “money 0.333333 -15.5374 airport, 0.333333 -15.5374 sdom(a) 0.333333 -15.5374 heather 0.333333 -15.5374 funny, 0.333333 -15.5374 figured 0.12 -15.538 consuming 0.25 -15.5393 program: 0.130435 -15.5397 stream. 0.333333 -15.5399 (s,t,e) 0.333333 -15.5399 nancy-universit´e 0.333333 -15.5399 ups 0.333333 -15.5399 time-span 0.333333 -15.5423 ~20 0.333333 -15.5423 (wer 0.333333 -15.5423 wordwi 0.333333 -15.5423 l→ψ 0.333333 -15.5423 (lexicalized 0.102564 -15.5425 “play 0.333333 -15.5447 [it 0.333333 -15.5447 σinfix(w,g) 0.333333 -15.5447 boilerplate 0.333333 -15.5447 531 0.333333 -15.5447 merriam-webster 0.333333 -15.5447 subtilis; 0.333333 -15.5447 æm 0.333333 -15.5447 24.35 0.333333 -15.5447 (hudson, 0.0833333 -15.546 accounting 0.0368098 -15.5463 be seen 0.333333 -15.5472 feat(nullnullnullnullnullnullnull 0.333333 -15.5472 bekalan-bekalan 0.333333 -15.5472 ϕ(a) 0.333333 -15.5472 (2%), 0.333333 -15.5472 nullnull: 0.25 -15.5484 ree 0.333333 -15.5497 jacques 0.333333 -15.5497 phonebook, 0.333333 -15.5497 lappin 0.333333 -15.5497 0.0819672 -15.5848 analysis for 0.136364 -15.585 noah 0.153846 -15.5852 chunk. 0.25 -15.586 0.7.25 0.333333 -15.5873 threshold factor 0.333333 -15.5873 devlin 0.333333 -15.5873 statopenccg 0.333333 -15.5873 840 0.2 -15.5883 demethylation 0.105263 -15.5891 tended 0.181818 -15.5898 “red 0.25 -15.5899 3.66 0.25 -15.5899 37.5% 0.333333 -15.5899 die, 0.333333 -15.5899 orsay, 0.333333 -15.5899 wh-question, 0.333333 -15.5899 @q 0.333333 -15.5899 (chou 0.333333 -15.5899 +lm 0.333333 -15.5899 14.4% 0.25 -15.5918 praat 0.333333 -15.5924 impactof 0.333333 -15.5924 8this 0.333333 -15.5924 bass 0.0033121 -15.5926 that 0.0583333 -15.5942 this would 0.333333 -15.595 bridget 0.333333 -15.595 “crude 0.25 -15.5957 3null4 0.25 -15.5957 1468 0.25 -15.5957 insteadof 0.25 -15.5957 never, 0.25 -15.5957 tor 0.0631579 -15.5964 we look 0.0757576 -15.5969 linguistics and 0.25 -15.5976 61.76 0.333333 -15.5976 “address” 0.333333 -15.5976 jprime 0.333333 -15.5976 +mt 0.333333 -15.5976 displacement 0.333333 -15.5976 you] 0.333333 -15.5976 derived, 0.333333 -15.5976 “defeated” 0.333333 -15.5976 intents 0.0529801 -15.5992 techniques to 0.25 -15.5995 sar 0.2 -15.6007 vouk, 0.25 -15.6015 unspecialized 0.333333 -15.6028 pasbio 0.333333 -15.6028 0-2 0.333333 -15.6028 “a(b 0.333333 -15.6028 80%) 0.333333 -15.6028 decipherments 0.333333 -15.6028 ≥0, 0.333333 -15.6028 a[the] 0.153846 -15.603 algorithm (dempster 0.25 -15.6034 i←1 0.166667 -15.6041 equation, 0.25 -15.6054 impress 0.25 -15.6054 ej, 0.142857 -15.6054 watson research 0.333333 -15.6054 0.575 0.333333 -15.6054 84.71 0.333333 -15.6054 -は 0.25 -15.6073 -と 0.333333 -15.608 =log 0.333333 -15.608 bike), 0.333333 -15.608 asts, 0.25 -15.6093 ∗school 0.333333 -15.6106 mohan 0.333333 -15.6106 (32 0.333333 -15.6106 shari 0.25 -15.6113 avgsl 0.25 -15.6113 composition: 0.181818 -15.6125 source-to-target 0.25 -15.6132 logios 0.333333 -15.6132 systemcombination 0.333333 -15.6132 koˇcka 0.333333 -15.6132 ≻ 0.25 -15.6152 covenants 0.25 -15.6152 word-word 0.333333 -15.6158 exploded 0.333333 -15.6158 bay, 0.333333 -15.6158 mnegationslash= 0.333333 -15.6158 f(c, 0.333333 -15.6158 nestedness 0.0869565 -15.6163 hand. 0.25 -15.6172 84.09 0.25 -15.6172 0.842 0.333333 -15.6185 (formal 0.333333 -15.6185 model,we 0.333333 -15.6185 god’s 0.25 -15.6192 i8 0.25 -15.6192 sex, 0.25 -15.6192 h−1−m 0.00898613 -15.6202 same 0.25 -15.6211 87.51 0.333333 -15.6211 race, 0.333333 -15.6211 without/with 0.333333 -15.6211 4] 0.333333 -15.6211 scfg g 0.333333 -15.6211 1278 0.333333 -15.6211 26.56 0.333333 -15.6211 partir 0.333333 -15.6211 luisa 0.0493827 -15.6218 capable 0.25 -15.6231 ‘be 0.333333 -15.6238 i/prp 0.333333 -15.6238 loria/inria, 0.333333 -15.6238 rachel 0.333333 -15.6238 戽 0.2 -15.6243 iwslt, 0.25 -15.6251 leftarc 0.25 -15.6251 comunication, 0.333333 -15.6264 greetings 0.333333 -15.6264 madv 0.333333 -15.6264 (spearman’s 0.25 -15.6271 j» 0.136364 -15.6276 (ramage 0.153846 -15.6285 => 0.25 -15.6291 trad-r 0.333333 -15.6291 meta-features 0.333333 -15.6291 ); 0.333333 -15.6291 hearst-i 0.25 -15.6311 p(wi) 0.25 -15.6311 definitions: 0.333333 -15.6318 diversity-aware 0.333333 -15.6318 1507 0.333333 -15.6318 %hq7dvndu 0.333333 -15.6318 (obj), 0.333333 -15.6318 afternoon, 0.333333 -15.6318 (which, 0.181818 -15.6327 arrested 0.0659341 -15.6332 account. 0.181818 -15.6342 auto-generated 0.333333 -15.6344 21.97 0.142857 -15.6348 (hockenmaier 0.25 -15.6351 strict-unsup 0.25 -15.6371 ancn1(r)) 0.333333 -15.6371 17.06 0.333333 -15.6371 ye. 0.333333 -15.6371 ::= 0.333333 -15.6371 rele 0.333333 -15.6371 “five 0.25 -15.6392 (si), 0.25 -15.6392 vasileios 0.25 -15.6392 santos 0.25 -15.6392 oj, 0.25 -15.6392 response: 0.333333 -15.6398 gratitude 0.333333 -15.6398 emiterations 0.333333 -15.6398 akinori 0.333333 -15.6398 (mutual 0.333333 -15.6398 ekaterina 0.25 -15.6412 non-regional 0.25 -15.6412 prep(in) 0.2 -15.6419 “happy 0.333333 -15.6425 [h1,i,h2h3,j] 0.333333 -15.6425 dieser 0.333333 -15.6425 potts 0.0909091 -15.6438 2011; 0.078125 -15.6442 grateful 0.25 -15.6452 psss, 0.333333 -15.6452 the/s 0.333333 -15.6452 identify: 0.333333 -15.6452 (nc) 0.333333 -15.6452 manualevaluation 0.25 -15.6473 p@q 0.25 -15.6473 mehr 0.0564516 -15.6475 experiment, we 0.333333 -15.6479 0,2 0.333333 -15.6479 13.07 0.333333 -15.6479 beer 0.333333 -15.6479 nuevo 0.333333 -15.6479 borrelia 0.25 -15.6493 s-p 0.333333 -15.6506 230,864 0.12 -15.6511 [ 0.25 -15.6513 stepsizes 0.107143 -15.6531 june, 2011. 0.25 -15.6534 p(r1 0.25 -15.6534 inductively 0.25 -15.6534 1211 0.25 -15.6534 c2), 0.333333 -15.6534 itj 0.333333 -15.6534 il(s) 0.333333 -15.6534 changesthat 0.333333 -15.6534 lgpl 0.0173716 -15.6535 proceedings of the 0.166667 -15.6548 fisher’s 0.0300926 -15.6549 under the 0.333333 -15.6561 click-through 0.333333 -15.6561 a2l, 0.333333 -15.6561 shining 0.25 -15.6575 nomlex 0.25 -15.6575 85.65 0.2 -15.6583 cow, 0.0675676 -15.6587 use only 0.333333 -15.6589 oneof 0.333333 -15.6589 ftc 0.333333 -15.6589 vectortildewidet 0.333333 -15.6589 strogatz, 0.333333 -15.6589 0.333333 -15.6589 koc¸ 0.333333 -15.6589 notations, 0.333333 -15.6589 jos´e 0.25 -15.6596 dedo 0.25 -15.6596 768 0.2 -15.6599 “/ 0.0359477 -15.661 compared to the 0.25 -15.6616 ∆p 0.25 -15.6616 rnr 0.333333 -15.6616 dact 0.333333 -15.6616 colleague 0.333333 -15.6616 zaenen 0.333333 -15.6616 e1e2 0.333333 -15.6616 aprime(m,n,k) 0.2 -15.6616 relatedwork 0.0457143 -15.663 tends 0.25 -15.6637 set-valued 0.25 -15.6637 1/1 0.25 -15.6637 (ν,µ) 0.142857 -15.664 base system 0.333333 -15.6644 direction; 0.333333 -15.6644 substitutions: 0.333333 -15.6644 (mapping 0.333333 -15.6644 exp3 0.333333 -15.6644 =null 0.333333 -15.6644 karlsruhe, 0.333333 -15.6644 thatt 0.333333 -15.6644 equator 0.333333 -15.6644 element-wise 0.0364238 -15.6664 this model 0.333333 -15.6671 right-handside 0.333333 -15.6671 10-cluster 0.333333 -15.6671 f↑ 0.333333 -15.6671 interleaved-expected-reconstruction-error-0.2interleaved-expected-reconstruction-error-0.4 0.333333 -15.6671 dull, 0.333333 -15.6671 multi-conditional 0.25 -15.6678 2.873 0.25 -15.6678 0.113 0.2 -15.6682 simhash 0.25 -15.6699 oe, 0.0833333 -15.6699 drawback 0.333333 -15.6699 hij 0.333333 -15.6699 caption, 0.333333 -15.6699 68-form languages. 0.2 -15.6715 supertreelet 0.115385 -15.6719 trento, 0.25 -15.672 section), 0.25 -15.672 62.15 0.25 -15.672 technology hyderabad, india 0.333333 -15.6727 philippine 0.333333 -15.6727 dependency–based 0.333333 -15.6727 leuven, 0.333333 -15.6727 organisation, 0.333333 -15.6727 ciphertext: 0.333333 -15.6727 summands 0.0735294 -15.673 1 , 0.2 -15.6732 55.72 0.2 -15.6732 nice, 0.2 -15.6732 vector” 0.2 -15.6732 volkswagen 0.2 -15.6732 dabei 0.2 -15.6749 (pestian 0.333333 -15.6755 conan 0.333333 -15.6755 di fabrizio, 0.333333 -15.6755 89.32 0.333333 -15.6755 anton 0.333333 -15.6755 boch 0.333333 -15.6755 typesof 0.333333 -15.6755 detonate 0.25 -15.6762 baseline1 0.2 -15.6766 logloss 0.2 -15.6766 memory). 0.2 -15.6766 comunity 0.333333 -15.6783 t(mult(e)) 0.333333 -15.6783 smrž, 0.333333 -15.6783 amountof 0.2 -15.6799 mt08-nw 0.333333 -15.6811 -た 0.333333 -15.6811 taro 0.0277778 -15.6811 annual 0.0145553 -15.6817 used in 0.25 -15.6825 inaddition, 0.0451977 -15.6832 experiments in 0.333333 -15.6839 irp3-omaprox2-ts 0.333333 -15.6839 si2 0.25 -15.6846 vp) 0.25 -15.6846 “he”, 0.133333 -15.6856 gene/protein 0.25 -15.6867 durme1,2 0.333333 -15.6867 elementsof 0.333333 -15.6867 zt 0.333333 -15.6867 across-the-board 0.0308483 -15.6881 contrast, 0.25 -15.6889 #relevant 0.25 -15.6889 -0.24 0.25 -15.6889 0.473 0.25 -15.6889 36.63 0.25 -15.6889 t29 0.133333 -15.689 defer 0.333333 -15.6895 c[i,k] 0.333333 -15.6895 wright 0.333333 -15.6895 [we 0.333333 -15.6895 robin 0.130435 -15.6908 closer look 0.25 -15.691 average_f 0.2 -15.6918 +0.8 0.2 -15.6918 name: 0.333333 -15.6924 [first 0.333333 -15.6924 lars 0.333333 -15.6924 relativ 0.333333 -15.6924 3535 0.333333 -15.6924 1095 0.25 -15.6931 diagnoses 0.0952381 -15.6932 (hall 0.2 -15.6935 “learning 0.25 -15.6952 c1), 0.333333 -15.6952 transferred, 0.333333 -15.6952 smoker, 0.333333 -15.6952 constructing, 0.333333 -15.6952 day=rel 0.2 -15.6969 (lowest 0.25 -15.6974 anais 0.25 -15.6974 inter-annotation 0.333333 -15.6981 manually-created 0.333333 -15.6981 factorisation 0.333333 -15.6981 e+ 0.333333 -15.6981 ψc,t 0.2 -15.6986 made* 0.25 -15.6995 handbok 0.153846 -15.6998 asia 0.2 -15.7003 alicante 0.333333 -15.7009 [q−r(sτ)] 0.333333 -15.7009 “let 0.25 -15.7017 @æ 0.25 -15.7017 θ1 0.2 -15.7021 mitesh 0.2 -15.7021 (barrault, 0.181818 -15.7022 instead, they 0.103448 -15.7032 2.2. 0.25 -15.7038 carberry 0.333333 -15.7038 clever, 0.333333 -15.7038 post-edits 0.2 -15.7038 amod 0.2 -15.7055 bad; 0.25 -15.706 unine 0.25 -15.706 76.71 0.333333 -15.7067 itexpl 0.333333 -15.7067 proposal: 0.2 -15.7072 sentchunk 0.2 -15.7072 be::v 0.0405405 -15.7075 denotes the 0.2 -15.709 nikhil 0.333333 -15.7096 831 0.333333 -15.7096 d∗ 0.333333 -15.7096 delineates 0.333333 -15.7096 ‘ground 0.333333 -15.7096 reservoir, 0.142857 -15.71 vorbei:by 0.25 -15.7103 (default 0.25 -15.7103 (christ, 0.25 -15.7125 0.587 0.25 -15.7125 employees, 0.333333 -15.7125 a1:intend 0.333333 -15.7125 obj1 0.333333 -15.7125 at-loc 0.2 -15.7142 63.50 0.2 -15.7142 disputed 0.25 -15.7147 dehydroxylation 0.25 -15.7147 ffopffr 0.25 -15.7147 second kind 0.142857 -15.715 zij 0.333333 -15.7154 53k 0.333333 -15.7154 pianta 0.25 -15.7168 eckert, 0.25 -15.7168 figure 4: examples 0.0231548 -15.7169 which we 0.103448 -15.7176 set). 0.333333 -15.7183 webster 0.333333 -15.7183 (trec) 0.333333 -15.7183 isa-i 0.25 -15.719 countt(f,e)summationtext 0.25 -15.719 rescores 0.25 -15.719 barriers 0.25 -15.719 “has 0.25 -15.719 l1crf 0.2 -15.7194 samsung 0.25 -15.7212 s; 0.25 -15.7212 wal 0.333333 -15.7212 klaren 0.333333 -15.7212 ―ك 0.333333 -15.7212 interleaved-expected-reconstruction-error-0.6interleaved-expected-reconstruction-error-0.8 0.25 -15.7234 “barack 0.25 -15.7234 weiß 0.333333 -15.7241 inc) 0.333333 -15.7241 gw, 0.333333 -15.7241 (represent 0.333333 -15.7241 g(p) 0.333333 -15.7241 killers 0.333333 -15.7241 half-court 0.333333 -15.7241 18.65 0.333333 -15.7241 mario 0.333333 -15.7241 site(entity)?,csite(entity)? 0.333333 -15.7241 thyme 0.2 -15.7247 0.518 0.25 -15.7256 91.50 0.181818 -15.726 cartesian 0.333333 -15.7271 p(ti 0.333333 -15.7271 presidency 0.333333 -15.7271 textcat 0.333333 -15.7271 portal, 0.25 -15.7278 query-focused summarization 0.25 -15.7278 0.122 0.2 -15.7282 libya 0.2 -15.7282 0.458 0.333333 -15.73 top-3 0.333333 -15.73 recscore+bias 0.2 -15.73 [last 0.2 -15.73 0.984 0.181818 -15.7308 agency, 0.181818 -15.7308 3.0. 0.181818 -15.7308 hussein 0.2 -15.7318 a50♦triangle 0.25 -15.7322 αprime 0.25 -15.7322 nasukawa, 0.25 -15.7322 kapil 0.25 -15.7322 85.53 0.333333 -15.733 epsilon1> 0.25 -15.7345 33.18 0.333333 -15.7359 84.85 0.333333 -15.7359 ,v 0.333333 -15.7359 는 0.181818 -15.7372 news text 0.166667 -15.7375 important issue 0.333333 -15.7389 nullprpnull 0.333333 -15.7389 ¯r(x) 0.333333 -15.7389 multi-nonterminal 0.2 -15.7389 kristy 0.0526316 -15.7408 this information 0.25 -15.7412 nuschime 0.25 -15.7412 lieberman 0.333333 -15.7419 15-7 0.333333 -15.7419 word+n-gram 0.333333 -15.7419 two  0.333333 -15.7419 (m2) 0.25 -15.7434 replace, 0.333333 -15.7449 60.92 0.25 -15.7456 y2y 0.2 -15.7478 messages: 0.25 -15.7479 lambert, 0.25 -15.7479 w∈si 0.333333 -15.7479 ‘they 0.047619 -15.7487 section 2, 0.2 -15.7496 macedonian 0.25 -15.7501 58.33 0.25 -15.7501 w/or 0.333333 -15.7509 731 0.333333 -15.7509 77.12% 0.333333 -15.7509 chakrabarti, 0.333333 -15.7509 lettertran 0.333333 -15.7509 bveptm 0.2 -15.7515 ldc2002e18, 0.25 -15.7524 717 0.0816327 -15.7528 sake of 0.0300752 -15.7536 department 0.333333 -15.7539 “items” 0.333333 -15.7539 althoughthe 0.333333 -15.7539 systemswas 0.333333 -15.7539 specification: 0.2 -15.7551 (decodepat 0.0545455 -15.7553 in cases 0.25 -15.7569 3a) 0.133333 -15.7569 (berg-kirkpatrick 0.333333 -15.7569 bedside 0.333333 -15.7569 -in 0.2 -15.7569 mlc 0.2 -15.7569 d5 0.2 -15.7587 rk(ρ,p) 0.25 -15.7592 −1.0 0.25 -15.7592 oracle-oov 0.166667 -15.76 (boyer, 0.333333 -15.76 v/s 0.333333 -15.76 curly 0.333333 -15.76 „can 0.2 -15.7605 (c2) 0.25 -15.7615 1249 0.25 -15.7615 gen(x) 0.333333 -15.763 3.1m 0.25 -15.7638 vieira 0.25 -15.7638 my 0.25 -15.7638 game: 0.25 -15.7638 beň 0.181818 -15.7652 fowler 0.181818 -15.7652 perp. 0.333333 -15.766 0.753 0.2 -15.766 penn2malt 0.25 -15.7661 dendrograms 0.25 -15.7661 29.9% 0.2 -15.7679 (verbs), 0.0288462 -15.7684 expect 0.333333 -15.7691 v´eronique 0.333333 -15.7691 [for 0.333333 -15.7691 p(si 0.25 -15.7707 muhua 0.333333 -15.7722 a subtreelet 0.333333 -15.7722 iron, 0.333333 -15.7722 (agent, 0.153846 -15.7724 (fan 0.25 -15.773 gayaa 0.25 -15.773 maise’s 0.25 -15.773 franklin 0.2 -15.7734 recall/precision/f1 0.2 -15.7734 em-80 0.2 -15.7734 slovenia 0.1 -15.7753 year’s 0.25 -15.7753 45.31 0.25 -15.7753 sophie 0.333333 -15.7753 lema 0.333333 -15.7753 fori←1 0.0309859 -15.7783 depending 0.333333 -15.7784 tpl, 0.333333 -15.7784 unix 0.333333 -15.7784 (einstein, 0.333333 -15.7784 19) 0.333333 -15.7784 notethat 0.0657895 -15.7799 (pang 0.333333 -15.7815 “a0 0.333333 -15.7815 (whissell, 0.333333 -15.7815 6.42 0.333333 -15.7815 featuref 0.0503597 -15.7821 a key 0.25 -15.7823 hala 0.25 -15.7823 vanlehn, 0.25 -15.7823 lattice-token-r 0.2 -15.7827 runs), 0.2 -15.7827 0.036 0.25 -15.7846 chilean dictator 0.25 -15.7846 smu-sis 0.25 -15.7846 4.62 0.333333 -15.7846 dennis 0.333333 -15.7846 volk 0.25 -15.7869 (arnold 0.25 -15.7869 82.35 0.25 -15.7869 recomendations 0.333333 -15.7877 多 0.333333 -15.7877 wounded 0.333333 -15.7877 lest 0.25 -15.7893 x3, 0.2 -15.7902 interinvlist 0.2 -15.7902 tscore 0.333333 -15.7908 adapted, 0.333333 -15.7908 (5, 0.333333 -15.7908 186.5 0.333333 -15.7908 “it’s 0.333333 -15.7908 ···wn 0.25 -15.7916 verb-phrase 0.25 -15.7916 (of) 0.25 -15.7916 eibe 0.2 -15.792 spanish: 0.2 -15.792 stuttgart, 0.2 -15.792 avignon, 0.25 -15.794 (***) 0.25 -15.794 cen 0.25 -15.794 +hier 0.333333 -15.794 0,0 0.333333 -15.794 h(t) 0.333333 -15.794 lscript1/lscriptp 0.2 -15.7958 n-gram: 0.25 -15.7963 negotiate” 0.333333 -15.7971 interfaculty 0.333333 -15.7971 ‘slow’ 0.333333 -15.7971 multi-player 0.0694444 -15.7975 we intend 0.2 -15.7977 (code 0.25 -15.7987 implementations: 0.25 -15.7987 rush, 0.2 -15.7996 f7 0.2 -15.7996 mosquitoes 0.166667 -15.8003 $. 0.333333 -15.8003 systems,and 0.333333 -15.8003 \w*{w2}\w*_(\d+) 0.0625 -15.8005 r p 0.25 -15.8011 min( 0.25 -15.8011 8a 0.25 -15.8011 36.9% 0.333333 -15.8034 start item 0.2 -15.8034 m5 0.2 -15.8034 hugo 0.107143 -15.8035 thus, we can 0.25 -15.8035 93.26 0.25 -15.8035 postdoctoral 0.25 -15.8035 in; 0.0437158 -15.8042 led 0.25 -15.8058 h(e) 0.25 -15.8058 dorow 0.333333 -15.8066 leuven 0.333333 -15.8066 “posterior 0.333333 -15.8066 (increased 0.333333 -15.8066 guergana 0.333333 -15.8066 walde, 0.2 -15.8072 κ0, 0.25 -15.8082 svm+ssrs 0.25 -15.8082 w(xi,yj) 0.25 -15.8082 plsa+prior 0.25 -15.8082 toulouse 0.2 -15.8091 z0 0.2 -15.8091 migration 0.333333 -15.8098 fkgl 0.0882353 -15.8101 xinhua 0.25 -15.8106 ,q3urfhhglqjv 0.25 -15.8106 0.327 0.25 -15.8106 91.28 0.0486111 -15.8123 opposed 0.0107203 -15.8129 example, 0.25 -15.813 73.53 0.25 -15.813 babytalk 0.333333 -15.813 1199 0.333333 -15.813 “central 0.2 -15.813 instrument-agency 0.2 -15.8149 jn 0.25 -15.8154 arteriolar 0.0291005 -15.8161 the training data 0.333333 -15.8162 “cat” 0.333333 -15.8162 ,q3urfhhglqjvri 0.333333 -15.8162 ‘show’ 0.333333 -15.8162 thmm 0.25 -15.8178 4.13 0.25 -15.8178 afosr 0.2 -15.8188 39.36 0.0595238 -15.8189 the first one 0.333333 -15.8194 853 0.333333 -15.8194 resource: 0.333333 -15.8194 0.110 0.0182371 -15.8202 have a 0.2 -15.8207 currencies 0.142857 -15.8213 0.2 -15.8785 bp,q=1 0.2 -15.8785 tran 0.25 -15.8827 have 0.25 -15.8827 prem 0.25 -15.8827 gp 0.25 -15.8827 1.7m 0.125 -15.8852 topic information 0.25 -15.8853 subnode 0.166667 -15.8879 ∀ci∈s,(li 0.25 -15.8879 daniele 0.25 -15.8879 w ��pk� 0.25 -15.8879 :d 0.25 -15.8879 91.77 0.0666667 -15.8883 be inferred 0.25 -15.8905 2006: 0.25 -15.8905 -0.10 0.2 -15.893 revolution 0.2 -15.893 r(e1,e2) 0.25 -15.8931 ≤2 0.25 -15.8957 vessels 0.153846 -15.8963 be paraphrased 0.2 -15.8972 bt-nurse 0.2 -15.8972 p(s1 0.25 -15.8983 “stretchy 0.25 -15.8983 24.38 0.142857 -15.899 nsf under 0.117647 -15.8992 (eng. 0.2 -15.8993 [0 0.25 -15.9009 radj 0.25 -15.9009 1,2in 0.25 -15.9009 adam, 0.25 -15.9009 palestine 0.2 -15.9014 hamza 0.2 -15.9014 nyc 0.166667 -15.9018 javier 0.25 -15.9035 non-rebuttals 0.2 -15.9035 ib;j 0.25 -15.9062 0.25 -15.9062 detector-i 0.166667 -15.9071 (mr) 0.25 -15.9088 perturbation 0.25 -15.9088 0.05; 0.25 -15.9088 dictator 0.166667 -15.9089 enciphered 0.25 -15.9115 2w 0.25 -15.9115 monash 0.2 -15.9141 c3, 0.0555556 -15.9147 is being 0.105263 -15.9164  0.25 -15.9168 (wps), 0.25 -15.9168 rds 0.25 -15.9168 memt 0.25 -15.9168 44.21 0.25 -15.9168 mlp 0.2 -15.9183 [y 0.2 -15.9183 (en), 0.25 -15.9195 lieutenant: 0.25 -15.9195 0.056 0.25 -15.9195 73.70 0.25 -15.9195 catching 0.25 -15.9195 “event 0.2 -15.9205 hiroshi 0.2 -15.9205 (answer: 0.2 -15.9205 flies 0.153846 -15.9207 neighbours. 0.166667 -15.9213 ∂wj 0.25 -15.9221 73.9% 0.25 -15.9221 (nn, 0.25 -15.9221 drawings 0.2 -15.9226 dp1 0.25 -15.9248 (1977) 0.25 -15.9248 enullnullnullnullnull 0.25 -15.9248 geo:flowsthrough 0.2 -15.9269 nn-prop 0.25 -15.9275 macbook 0.166667 -15.9284 finkel, 0.133333 -15.9298 specified. 0.25 -15.9302 (ri, 0.25 -15.9302 fieldwork 0.25 -15.9302 64.26 0.25 -15.9302 1.55 0.25 -15.9302 p(r2 0.0555556 -15.9308 subject and 0.166667 -15.932 stoyanchev 0.0384615 -15.9327 we refer 0.0655738 -15.9335 framework that 0.0595238 -15.9341 point. 0.25 -15.9356 99.4 0.25 -15.9356 efron 0.2 -15.9378 to overcome this, 0.2 -15.9378 node(null) 0.2 -15.9399 enjoy, 0.133333 -15.94 chung 0.25 -15.9411 vbg, 0.25 -15.9411 sucs 0.25 -15.9411 ladies 0.0182648 -15.9414 and then 0.2 -15.9421 p  0.2 -15.9421  have 0.166667 -15.9429 kanssa 0.166667 -15.9429 warg 0.25 -15.9438 d,l 0.2 -15.9443 cv, 0.0967742 -15.9448 (munteanu 0.142857 -15.945 to implement, 0.142857 -15.945 goldwater et 0.2 -15.9465 k∑ 0.2 -15.9465 ` 0.2 -15.9465 webber 0.25 -15.9466 who answered 0.25 -15.9493 ”hot 0.25 -15.9493 cs, 0.142857 -15.9497 package. 0.0377358 -15.9509 that there is 0.020202 -15.9518 us to 0.166667 -15.9521 (learning 0.25 -15.9521 source-sided 0.25 -15.9521 xdiamsa 0.25 -15.9521 1025 0.25 -15.9521 discourses, 0.2 -15.9531 dx 0.2 -15.9531 84.9% 0.2 -15.9531 rafael 0.25 -15.9548 participatedin 0.25 -15.9548 mention), 0.25 -15.9548 25.75 0.166667 -15.9558 zur 0.0526316 -15.957 we use an 0.166667 -15.9576 cosyne 0.25 -15.9576 wmtsmall 0.2 -15.9576 socer 0.142857 -15.9592 director 0.0259434 -15.9596 1 shows 0.2 -15.9598 arabic: 0.25 -15.9604 parenleftbiggproductdisplay 0.048 -15.9614 problem in 0.166667 -15.9614 refering to 0.023166 -15.9623 conference on 0.25 -15.9632 563 0.2 -15.9643 q¸@ 0.25 -15.966 88.80 0.25 -15.966 e,a 0.25 -15.966 2nd-person 0.00963082 -15.9676 the number of 0.25 -15.9688 coling/acl 0.2 -15.971 shooter’s 0.2 -15.971 figure4: 0.25 -15.9716 generatethe 0.125 -15.973 (harris, 0.2 -15.9732 +0.41 0.2 -15.9732 singers 0.25 -15.9744 restricted-many-to-many 0.25 -15.9744 strmatch 0.25 -15.9744 20.70 0.0260664 -15.9745 method is 0.25 -15.9772 cnc 0.25 -15.9772 g-cut 0.25 -15.9772 16.19 0.25 -15.9772 corpus-, 0.2 -15.98 an author, 0.25 -15.9801 per-topic 0.25 -15.9801 inchoative 0.166667 -15.982 1.96 0.25 -15.9829 cros-chain 0.25 -15.9829 89.42 0.2 -15.9846 δi 0.117647 -15.9855 campaign. 0.25 -15.9858 ‘when 0.25 -15.9858 u2(x,y,di,p,s) 0.25 -15.9858 trans4 0.2 -15.9869 stubbs 0.2 -15.9869 eve 0.166667 -15.9877 (yan 0.25 -15.9886 arc length 0.25 -15.9886 polymerase 0.2 -15.9892 −0.50 0.2 -15.9892 taxa 0.0967742 -15.9893 data consists 0.0645161 -15.9903 wang et 0.0701754 -15.9913 prone 0.25 -15.9915 (person), 0.25 -15.9915 55.29 0.2 -15.9915 α↓i 0.166667 -15.9935 0.583 0.25 -15.9944 full-dimensional 0.25 -15.9944 cristina 0.25 -15.9944 rust 0.2 -15.9961 p(e1)p(e2) 0.2 -15.9961 pal 0.25 -15.9973 respondable 0.1 -15.9996 heritage, 0.25 -16.0002 0.7477 0.25 -16.0002 u.pos 0.2 -16.0007 ordbok 0.0549451 -16.0013 intend 0.0143613 -16.0039 is an 0.25 -16.006 5,5and 0.25 -16.006 (ncsubj 0.25 -16.006 “during 0.0380435 -16.0077 one is 0.2 -16.0077 group; 0.0311419 -16.0083 first, we 0.0882353 -16.0088 lattice. 0.25 -16.0089 wide-scope 0.25 -16.0089 a∼ 0.25 -16.0089 16gb 0.25 -16.0089 w(i,j) 0.25 -16.0089 tanmoy 0.25 -16.0089 limi→∞ 0.0271739 -16.009 the result 0.142857 -16.0097 0.142857 -16.0097 inter- 0.25 -16.0118 φm 0.25 -16.0118 possess, 0.25 -16.0118 s56s48 0.142857 -16.0147 (the, 0.2 -16.0147 “de 0.25 -16.0148 ∈{nn} 0.25 -16.0148 verbalpresent: 0.25 -16.0148 p1r5 0.25 -16.0148 39.09 0.0689655 -16.0167 grants 0.2 -16.0171 grzegorz 0.2 -16.0171 )()( 0.25 -16.0177 varna 0.142857 -16.0181 “plant” 0.0237581 -16.0191 thank 0.0689655 -16.0192 increases. 0.25 -16.0207 deniz 0.25 -16.0207 insulation 0.25 -16.0207 all  0.25 -16.0207 α2] 0.25 -16.0207 28.65 0.25 -16.0207 adji 0.2 -16.0218 [firstname] 0.111111 -16.0226 shortness 0.25 -16.0236 srbya 0.25 -16.0236 (≈ 0.166667 -16.0266 hate, 0.166667 -16.0266 88.06 0.00365016 -16.0266 be 0.25 -16.0266 11.2% 0.0384615 -16.0266 thus, the 0.142857 -16.0283 taiwan, 0.166667 -16.0286 1983) 0.25 -16.0296 16.51 0.25 -16.0296 state* 0.25 -16.0296 28.97 0.25 -16.0296 abs 0.166667 -16.0306 99.2 0.2 -16.0313 (sbar 0.2 -16.0313 echocardiogram 0.142857 -16.0317 rsing 0.00785105 -16.0325 as a 0.25 -16.0326 (64m 0.25 -16.0326 +fp 0.25 -16.0326 total: 0.2 -16.0337 amateur 0.2 -16.0337 f-ratios 0.25 -16.0356 men* 0.25 -16.0356 φ(x, 0.25 -16.0356 2+ 0.166667 -16.0366 matteo 0.2 -16.0385 descriptionof 0.25 -16.0386 stochastically 0.25 -16.0386 pprime0 0.25 -16.0386 grm, 0.25 -16.0386 0.704 0.25 -16.0386 conviction 0.25 -16.0386 mitogenic 0.0731707 -16.0398 each user 0.25 -16.0416 out-of-the-box 0.25 -16.0416 18.27 0.166667 -16.0426 bmm 0.105263 -16.0443 ----- 0.25 -16.0446 f-score(%) 0.166667 -16.0447 18.45 0.2 -16.0458 sussex 0.2 -16.0458 0.823 0.25 -16.0477 (born 0.117647 -16.0479 robot. 0.25 -16.0507 41.66 0.25 -16.0507 hearing-impaired 0.25 -16.0507 ipw 0.25 -16.0507 (schein 0.166667 -16.0528 emanuele 0.25 -16.0537 *: 0.25 -16.0537 bosma, 0.25 -16.0537 für 0.25 -16.0537 sinica 0.142857 -16.0559 54, 0.166667 -16.0568 c(w) 0.25 -16.0568 icelandic 0.2 -16.058 researcher, 0.166667 -16.0589 cape 0.25 -16.0599 ¶ 0.25 -16.0599 p(gt 0.2 -16.0629 1425 0.2 -16.0629 „ 0.2 -16.0629 ×v 0.166667 -16.063 cu-zeman 0.25 -16.063 macintosh 0.117647 -16.0652 1(a) 0.0437956 -16.0654 regard 0.25 -16.0661 rome 0.25 -16.0661 n=10,m=20 0.142857 -16.0665 pagent 0.2 -16.0679 pmi(i, 0.25 -16.0692 brbo 0.25 -16.0692 vger 0.25 -16.0723 volcanic 0.25 -16.0723 (davis 0.25 -16.0723 grenoble, 0.166667 -16.0733 la1 0.166667 -16.0733 e∈e 0.25 -16.0754 89.4% 0.25 -16.0785 2,2 0.0515464 -16.0799 liu et 0.142857 -16.0807 “visible 0.25 -16.0817 ngram-r 0.25 -16.0817 25.59 0.2 -16.0829  on 0.25 -16.0848 ω(a)j 0.2 -16.0854 k3 0.0909091 -16.0857 each tweet 0.25 -16.088 (public 0.25 -16.088 “not-before” 0.25 -16.088 1105 0.2 -16.0905 93.06 0.2 -16.0905 yaser 0.25 -16.0911 0.296 0.25 -16.0943 “big 0.166667 -16.0944 ji; 0.166667 -16.0965 allergies 0.25 -16.0975 lattice-aware 0.25 -16.0975 corpus); 0.25 -16.0975 51.51 0.0304183 -16.0981 (such 0.2 -16.0981 they] 0.056338 -16.0989 responsible 0.105263 -16.0994 authors were 0.25 -16.1007 producttextx 0.166667 -16.1029 entendre 0.25 -16.1039 iter 0.25 -16.1039 carrot 0.25 -16.1071 product) 0.166667 -16.1072 0.606 0.2 -16.1084 (han) 0.2 -16.1084 above discussion 0.25 -16.1104 yaqin 0.25 -16.1104 definethe 0.25 -16.1104 15.22 0.25 -16.1104 92.98 0.117647 -16.1113 york: 0.25 -16.1136 12.83 0.25 -16.1136 4.80 0.25 -16.1136 kottaige 0.25 -16.1136 p&r= 0.25 -16.1136 (fr), 0.25 -16.1136 repubblica 0.25 -16.1136 lapalme, 0.25 -16.1136 (basili 0.142857 -16.1136 i=2 0.25 -16.1169 1373 0.25 -16.1169 82.39 0.25 -16.1169 cristo 0.0202703 -16.1169 language. 0.0952381 -16.118 (dagan 0.25 -16.1201 kereta 0.166667 -16.1202 •s 0.0258621 -16.1208 result of 0.0909091 -16.1219 turian 0.25 -16.1234 0.309 0.25 -16.1234 −→v 0.2 -16.1266 (not) 0.166667 -16.1267 1,818 0.25 -16.1267 2111 0.25 -16.1267 rti 0.2 -16.1293 23: 0.25 -16.13 “robot 0.25 -16.13 bretonnel 0.25 -16.13 0.469 0.25 -16.13 (church, 0.142857 -16.1304 paiwan 0.142857 -16.1304 regina 0.0218688 -16.1314 acknowledgments 0.25 -16.1333 benson, 0.25 -16.1333 hydroxylation 0.2 -16.1346 event), 0.2 -16.1346 costa-jussà 0.25 -16.1366 murdered 0.25 -16.1366 24.21 0.25 -16.1366 haydn 0.25 -16.1366 lado 0.25 -16.1366 q-rate 0.2 -16.1372 ohta∗ 0.25 -16.1399 chih-chung 0.25 -16.1399 szarvas 0.25 -16.1399 night, 0.25 -16.1399 ;ij 0.25 -16.1399 stoyanov, 0.25 -16.1399 맛있다 0.25 -16.1399 countmatch(n-gram) 0.25 -16.1399 recall(%) 0.25 -16.1399 eex 0.2 -16.1399 fang 0.125 -16.1399 steyvers, 0.125 -16.1432 unzipped 0.142857 -16.1437 saudi 0.2 -16.1479 m45 0.142857 -16.1495 km04 0.142857 -16.1495 cultures. 0.2 -16.1506 "a 0.2 -16.1533 2.13% 0.166667 -16.1556 np@2null1null2 0.166667 -16.1579 ser(727) 0.2 -16.1587 actvp(derived) 0.166667 -16.1602 ra1 0.0261628 -16.1602 been used 0.057971 -16.1663 forest. 0.142857 -16.1669 (abeill´e 0.166667 -16.167 hr0011-08-c-0110. 0.047619 -16.1676 ways. 0.142857 -16.1689 ˆt 0.2 -16.1697 47.07 0.0952381 -16.1702 ,q 0.142857 -16.1709 “sentiment 0.142857 -16.1728 f-­‐ 0.105263 -16.1735 bulk 0.166667 -16.1739 k,b,c 0.0833333 -16.175 ] 0.0547945 -16.1762 is composed 0.142857 -16.1768 -0.13 0.2 -16.1807 658 0.08 -16.1818 lingual 0.111111 -16.1823 afinitor 0.2 -16.1835 46.36 0.2 -16.1863 (butt 0.2 -16.1863 dcu 0.166667 -16.1878 mz 0.2 -16.1891 “(n 0.166667 -16.1901 38123 povo 0.0681818 -16.1922 (lda) 0.0666667 -16.1929 stack. 0.142857 -16.1947 imported from 0.06 -16.1958 far. 0.166667 -16.1971 fb2(rs2 0.08 -16.1987 ,q 0.142857 -16.2028 grct 0.142857 -16.2048 sourceand 0.142857 -16.2048 fi¸@ 0.2 -16.206 pa., 0.0260586 -16.2071 large number 0.2 -16.2089 kik 0.166667 -16.209 “war” 0.0125698 -16.21 corpus. 0.166667 -16.2113 wds+bi 0.125 -16.2125 cough 0.0214592 -16.2126 the task of 0.166667 -16.2137 svr-base, 0.142857 -16.215 (su 0.166667 -16.2161 87.36 0.2 -16.2204 41.24 0.2 -16.2291 hr0011-06-c-0022, 0.2 -16.2291 87.65 0.0769231 -16.2345 maryland 0.142857 -16.2358 ρkj(1) 0.0769231 -16.2367 4000 0.2 -16.2379 ω(c) 0.2 -16.2379 subtiwiki 0.166667 -16.2404 sex 0.0833333 -16.2416 semantic resources 0.166667 -16.2429 dik 0.142857 -16.2442 sgd 0.0227273 -16.2456 al., 2010) 0.2 -16.2467 isabel 0.2 -16.2467 recipe for 0.142857 -16.2485 action: 0.0342857 -16.2487 have not 0.0322581 -16.2498 small number 0.142857 -16.2506 liu1 0.0126394 -16.2517 performance of 0.142857 -16.2527 (uzuner 0.2 -16.2527 a potential solution 0.2 -16.2527 teg 0.0105475 -16.2534 data. 0.111111 -16.2544 es:it 0.2 -16.2557 moshe 0.0740741 -16.2599 (srl) 0.0444444 -16.261 probability is 0.2 -16.2617 component-whole 0.0740741 -16.2622 programme 0.0540541 -16.2637 (petrov 0.125 -16.2639 textrunner-r 0.2 -16.2647 =1 0.166667 -16.2653 baron 0.0697674 -16.2667 government. 0.142857 -16.2677 nrc 0.125 -16.2696 harvests 0.2 -16.2708 port-au-prince 0.125 -16.2715 yad 0.142857 -16.2721 98.6 0.125 -16.2734 j∈ji 0.1 -16.2754 transliteration: 0.0373134 -16.2767 results are shown 0.142857 -16.2786 “when 0.166667 -16.2856 generalizer 0.0191304 -16.2865 model for 0.142857 -16.2874 ninth 0.0714286 -16.2919     0.125 -16.2926 graeser, 0.111111 -16.295 4(a) 0.2 -16.2954 +) 0.166667 -16.296 _ 0.0833333 -16.2985 dragged 0.2 -16.2985 (conll) 0.0869565 -16.2985 features selected 0.2 -16.3048 tierra 0.142857 -16.3075 anr 0.142857 -16.3097 symposium, 0.0176282 -16.3112 the proposed 0.0769231 -16.313 any, 0.166667 -16.3143 fik 0.0526316 -16.3151 a robust 0.2 -16.3206 abroad 0.142857 -16.3211 mail, 0.05 -16.3254 end, we 0.08 -16.329 (crammer 0.2 -16.3302 invariance 0.2 -16.3302 martinis 0.2 -16.3367 agent(the 0.166667 -16.3384 “largest” 0.142857 -16.3395 slt 0.2 -16.34 t4, 0.2 -16.34 +lem+lem 0.166667 -16.3411 basic-sr 0.2 -16.3432 microand 0.125 -16.3444 n-dimensional 0.2 -16.3465 cqae 0.166667 -16.3521 204; 0.2 -16.3531 habe 0.2 -16.3531 rolls 0.166667 -16.3576 0.357 0.2 -16.363 ?r 0.2 -16.363 [i,j]h 0.2 -16.363 47th 0.1 -16.3647 (chiao 0.125 -16.3651 sadeep 0.166667 -16.3659 melbourne, 0.166667 -16.3659 x∈t 0.142857 -16.3678 clinton, 0.166667 -16.3687 combining three 0.0294118 -16.3715 4 shows 0.142857 -16.3774 0.372 0.166667 -16.3799 yd,s,n 0.0188324 -16.3821 consider the 0.125 -16.3862 manje 0.040404 -16.3931 same way 0.125 -16.3948 f¨or 0.125 -16.3991 (dcs 0.142857 -16.3994 (tsochantaridis 0.0182815 -16.4007 for example, the 0.142857 -16.4019 (µ 0.166667 -16.4028 del edge(nullnullnullnull 0.125 -16.4034 from this perspective, 0.0425532 -16.4065 the possibility of 0.037594 -16.4072 used in this 0.111111 -16.4085 nucleic 0.166667 -16.4086 〈“killed”, 0.1 -16.4091 �” 0.142857 -16.4093 jadavpur 0.166667 -16.4115 yoav 0.0526316 -16.4116 expense 0.142857 -16.4118 research india 0.0491803 -16.4127 smith and 0.142857 -16.4194 relations”, 0.0138889 -16.4225 consists 0.0223464 -16.4225 domain. 0.166667 -16.4291 implementations, we 0.0344828 -16.4296 regardless 0.166667 -16.4321 semanticparser 0.1 -16.4374 lisbon, 0.0113636 -16.4394 one of 0.166667 -16.4411 cream 0.0444444 -16.4416 this is an 0.166667 -16.4471 girls 0.0612245 -16.4482 0 10 20 0.111111 -16.4501 (cunningham 0.125 -16.4523 advantages: 0.0714286 -16.4553 extent. 0.0909091 -16.4584 cons 0.04 -16.4618 braceleftbigg 0.0285714 -16.4642 however, it 0.0666667 -16.4648 on the contrary, 0.0909091 -16.4684 jf 0.08 -16.4699 (wiebe et 0.111111 -16.4725 figures. 0.166667 -16.4778 fidanzato 0.0241379 -16.4825 uk, 0.1 -16.4834 holds, 0.166667 -16.484 j.c. 0.166667 -16.484 k=j+1 0.125 -16.4894 termino-ontology 0.166667 -16.4903 haar 0.166667 -16.4935 70.65 0.142857 -16.4979 [conceiver,intentional 0.033557 -16.4995 take into 0.0454545 -16.5061 are consistent 0.142857 -16.5062 ixa 0.125 -16.5085 ontario, 0.1 -16.512 tables 6 0.166667 -16.5126 (rst) 0.166667 -16.5159 l2. 0.0238095 -16.5159 from each 0.142857 -16.5172 head-finding 0.166667 -16.5224 andrei 0.00224423 -16.5248 of the 0.0320513 -16.5276 idea is 0.1 -16.5295 (rel) 0.0344828 -16.5365 describe our 0.111111 -16.5365 povo 0.166667 -16.5388 “variation 0.0517241 -16.5388 annotator. 0.125 -16.5453 ⇔ 0.166667 -16.5454 csli 0.111111 -16.5454 kproductdisplay 0.08 -16.5454 (pyysalo 0.125 -16.5478 fct 0.0408163 -16.5487 (blei 0.0909091 -16.5509 vp@2null2 0.1 -16.5514 describe, 0.142857 -16.5569 ‘that 0.125 -16.5579 varies. 0.111111 -16.5588 encountered. 0.142857 -16.5598 1ifnp 0.0392157 -16.5612 under grant 0.0714286 -16.5656 from. 0.111111 -16.5678 (argamon 0.0220126 -16.5782 in the case of 0.0428571 -16.5791 patterns for 0.0308642 -16.5819 dealing 0.0689655 -16.5819 ter. 0.142857 -16.5832 sta 0.0833333 -16.5862 texas* 0.0689655 -16.5919 this seems 0.111111 -16.5932 brechen 0.111111 -16.6002 null: 0.111111 -16.6072 (raghavan 0.015361 -16.6091 a sentence 0.0740741 -16.6096 details on the 0.111111 -16.612 under-resourced 0.125 -16.6152 greater. 0.0769231 -16.6154 option. 0.1 -16.6158 quarterback 0.142857 -16.6195 ont 0.0909091 -16.6248 prop. 0.0769231 -16.627 0.859 0.00321896 -16.6278 et 0.0769231 -16.6287 bracerightbigg 0.1 -16.6288 programme, 0.0120846 -16.6302 task. 0.125 -16.6314 [slot1,slot2,slot3,slot4], 0.0344828 -16.6325 we observed 0.125 -16.6369 t5 0.142857 -16.6381 ¯f 0.142857 -16.6381 a  0.142857 -16.6413  0.0769231 -16.6422 also defined 0.0909091 -16.6427 43rd 0.0422535 -16.6452 describes a 0.111111 -16.6458 large space 0.05 -16.6463 hierarchy. 0.125 -16.6479 isr, 0.125 -16.6479 uedin 0.125 -16.6507 emnlp, 0.142857 -16.6539 casacuberta 0.0909091 -16.6589 (finkelstein 0.142857 -16.6603 trainer 0.0207715 -16.663 refers 0.142857 -16.67 (kleinberg, 0.125 -16.6732 tier-based 0.142857 -16.6765 h(i+1) 0.0909091 -16.6794 “crf 0.0833333 -16.6827 null= 0.142857 -16.6863 np-sb 0.0909091 -16.6878 padt 0.111111 -16.691 (ctb) 0.0714286 -16.6996 franz, 0.046875 -16.701 features derived 0.125 -16.702 uned 0.0769231 -16.716 institut 0.0221402 -16.7198 work has 0.0134953 -16.7217 with respect 0.0128205 -16.724 respectively. 0.0215827 -16.7249 tree. 0.0333333 -16.7269 table. 0.047619 -16.7311 (svm) 0.0909091 -16.735 hence, we have 0.0909091 -16.7394 sad,” 0.0394737 -16.7397 mellon university 0.125 -16.7408 mapping text 0.0240385 -16.743 lot 0.0769231 -16.7494 chinese-side 0.111111 -16.752 ✓ 0.0337079 -16.7657 mooney, 0.0769231 -16.7666 implied, 0.111111 -16.7685 accuracy(%) 0.0666667 -16.7686 kong 0.0909091 -16.7754 >>> 0.028777 -16.7801 acknowledgments this work 0.125 -16.7811 87.38 0.111111 -16.7825 (biemann 0.00853728 -16.7871 proceedings 0.1 -16.7939 nullnull) 0.1 -16.8042 brothers 0.0769231 -16.8077 .0 0.1 -16.812 (culotta 0.125 -16.8198 teamingame 0.0769231 -16.8238 (blaheta 0.0179487 -16.8255 through the 0.125 -16.8297 4480 0.125 -16.833 ma: 0.0909091 -16.8367 hypertensive renal 0.111111 -16.8405 §3. 0.0625 -16.8414 found within 0.1 -16.8465 fork, 0.111111 -16.8465 acted 0.111111 -16.8465 lisbon 0.1 -16.8492 cj) 0.0416667 -16.8567 primeprime 0.1 -16.8601 lenci 0.0425532 -16.8621 precisely, 0.0228311 -16.8712 if there 0.032967 -16.883 0.00 0.00 0.0833333 -16.8855 eyebrows 0.0136054 -16.8865 we first 0.0118421 -16.8958 represented 0.0143627 -16.8963 in addition 0.00229843 -16.9027 in the 0.0125874 -16.9045 accuracy of 0.1 -16.9078 follow: 0.111111 -16.9148 pagar 0.1 -16.9194 [x 0.0540541 -16.9391 devoted 0.0666667 -16.941 meting 0.0833333 -16.9434 (x-y) 0.0625 -16.944 (marton 0.0625 -16.9496 eecs 0.111111 -16.9508 textinfer 0.0465116 -16.956 (brants 0.0454545 -16.9567 column is 0.0555556 -16.9643 to zero, 0.010846 -16.9744 model is 0.0106724 -16.9862 results of 0.0769231 -16.9868 nara 0.0714286 -16.9895 infocomm 0.0833333 -16.9943 second level 0.0769231 -16.9964 waltham, 0.0588235 -16.999 realisations 0.0714286 -17.0029  0.0909091 -17.006 these challenges 0.0555556 -17.0163 (mert) 0.00827016 -17.0195 well as 0.0666667 -17.0213 the y-axis 0.1 -17.0267 shortage 0.1 -17.0332 wiebe. 0.0265487 -17.0387 acl-hlt 0.00803751 -17.0467 using a 0.0588235 -17.0484 likelihood. 0.0272727 -17.0581 carnegie 0.0526316 -17.0632 case). 0.0714286 -17.0634 →r 0.0113798 -17.0639 given a 0.0769231 -17.0665 essence, 0.0625 -17.0667 (nguyen 0.0666667 -17.0787  0.0833333 -17.0787 advent 0.0104408 -17.1132 sentence. 0.0833333 -17.1134 lo and 0.00539169 -17.1139 with a 0.047619 -17.1167 example). 0.05 -17.1234 they then 0.015 -17.125 follows. 0.0909091 -17.1252 lightly-supervised 0.037037 -17.1278 on the one hand, 0.0909091 -17.135 subcontract 0.00739372 -17.1468 due 0.0714286 -17.1616 rid 0.0588235 -17.1747 mi/j 0.0526316 -17.1753 15213 0.047619 -17.1793 important step 0.0833333 -17.1803 oregon 0.0408163 -17.1931 2011: 0.00934579 -17.1932 is used 0.0134831 -17.199 would like 0.0769231 -17.2022 (uas) 0.030303 -17.2075 all pairs 0.0238095 -17.2272 al.’s 0.0124481 -17.2328 particular, 0.047619 -17.2558 whole, 0.0137741 -17.2603 carried 0.0625 -17.2664 ascertain 0.0625 -17.2768 t-it 0.00996264 -17.281 c©2011 association for 0.0384615 -17.2939 resort 0.030303 -17.3049 matrix. 0.0714286 -17.3065 norms, 0.0184332 -17.3082 performance is 0.0117417 -17.3157 section, 0.0555556 -17.3171 adhere 0.012605 -17.3185 tend 0.0289855 -17.3226 squaresmallsolid 0.0201342 -17.3407 be used for 0.0714286 -17.3632 ;a/ 0.0416667 -17.374 (wsj) 0.0277778 -17.3803 improvement is 0.0625 -17.3807 t-fr 0.0454545 -17.3978 2011 student 0.0666667 -17.4225 1...n 0.01875 -17.4237 all words 0.0263158 -17.4268 where there 0.0666667 -17.4419 ,t 0.0434783 -17.4423 its arguments 0.0666667 -17.4518 contain. 0.0232558 -17.4617 community. 0.0454545 -17.4731 sentence). 0.020979 -17.4897 for the first 0.0555556 -17.4899 austin 0.0344828 -17.5148 dinu 0.00662739 -17.5509 words in 0.037037 -17.5649 model-1 0.027027 -17.573 strategy for 0.00918836 -17.5752 fact that 0.0454545 -17.5771 nonregular 0.00859599 -17.5957 portland, 0.019802 -17.6086 is taken 0.04 -17.6102 we address this 0.0117994 -17.6352 with different 0.05 -17.639 redmond, 0.0194175 -17.6479 into three 0.0526316 -17.6556 1989; 0.0179641 -17.6667 c 2011 association for 0.0416667 -17.6668 analyser 0.00976562 -17.6681 corresponds 0.0434783 -17.6794 500 600 0.0253165 -17.688 a third 0.0454545 -17.7049 pros 0.0048216 -17.7056 in table 0.0416667 -17.7274 connective. 0.027027 -17.7333 number of edges 0.00264085 -17.7552 from the 0.0434783 -17.7581  0.0384615 -17.7939 (mwe 0.0384615 -17.8211 v null 0.0454545 -17.8446 contain many 0.037037 -17.8451 (crf) 0.0185185 -17.8575 article. 0.00801282 -17.8835 addition, 0.0121951 -17.9191 other words, 0.037037 -17.9213 heritage 0.027027 -17.9263 =+ 0.03125 -17.9291 200 300 0.0277778 -17.9294 mbr system 0.03125 -17.9595 they found 0.0104167 -18.0087 june 19-24, 2011. 0.00260126 -18.0158 is the 0.0117188 -18.0205 3 shows 0.0322581 -18.0335 optimized for 0.025 -18.0371 gratefully 0.0181818 -18.0447 is known 0.0238095 -18.0488 (akkaya 0.015748 -18.0645 as many 0.0357143 -18.066 main difference 0.0238095 -18.0748 d7 0.0322581 -18.0954 19-24 0.0322581 -18.1076 fifteenth 0.00503356 -18.1347 set. 0.0140845 -18.1762 al., 2003; 0.00273088 -18.2066 al. 0.030303 -18.2501 elhadad 0.0263158 -18.3424 a71a71 0.0166667 -18.3676 dept. 0.0149254 -18.3873 with less 0.0243902 -18.3996 extend this 0.00144681 -18.4162 to the 0.012987 -18.4177 (marcus 0.0204082 -18.4196 recent years, 0.00157266 -18.4509 on the 0.0227273 -18.4733 isolation. 0.016129 -18.5301 yield a 0.00854701 -18.5375 the underlying 0.02 -18.5375 3.2. 0.00892857 -18.6349 reviewers for 0.0196078 -18.6494 (based 0.00743494 -18.6558 relies 0.00784314 -18.6683 2 related 0.01 -18.6996 a detailed 0.0172414 -18.7157 york, 0.00328084 -18.7656 according 0.00164393 -18.7988 with the 0.0138889 -18.8019 such, 0.00593472 -18.8251 kinds 0.00126508 -18.835 al., 0.0147059 -18.9307 we wish 0.00474684 -18.9397 some of 0.00626959 -18.946 uses a 0.00628931 -18.9998 a subset 0.00112334 -19.2016 for the 0.00636943 -19.2055 found to 0.00389864 -19.2223 in proceedings 0.00806452 -19.2567 unable 0.0108696 -19.2655 three types 0.0060241 -19.3214 extend the 0.00502513 -19.3605 funded 0.00740741 -19.3663 acknowledgments we 0.00502513 -19.4246 our goal 0.00330579 -19.4457 follows: 0.00203252 -19.482 in order 0.0025 -19.5696 able 0.00497512 -19.7157 this research 0.00649351 -19.7482 compare our 0.00303951 -19.8046 the development 0.00348432 -19.8143 the majority 0.00512821 -19.8583 two types 0.000981547 -19.9079 in a 0.00253807 -20.0486 the value 0.00423729 -20.1405 for example, in 0.00251256 -20.1959 a71 0.0019305 -20.3759 for both 0.00140056 -20.4244 association for 0.00206612 -20.5459 are more 0.00252525 -20.7055 precision and 0.00080289 -20.8278 for a 0.00189394 -20.9214 most of 0.00165017 -21.0622 the overall 0.00152905 -21.2104 oregon, 0.000468604 -21.8675 used to 0.000308833 -22.6347 the number