6045 S. Kenwood Ave., Chicago, IL
shtoshni at ttic dot edu
I am a final year graduate student working with Kevin Gimpel and Karen Livescu at Toyota Technological Institute at Chicago (TTIC).
I also work closely with Sam Wiseman, and Allyson Ettinger.
Before TTIC, I spent two years as a software engineer in IBM Research, New Delhi. And even before that I did my undergrad in Computer Science from Indian Institute of Technology Kanpur.
My research is focused on natural language understanding, specifically on tracking the world state described in text, with particular focus on entities. This problem is essential to language comprehension, especially for long documents where understanding these abstractions are key to situating far apart evidences. We have used memory models for this task in recent work [1, 2]. We also worked on a fun project on language modeling for chess where the chess board is the world, and the chess pieces are the entities in this world . We're currently working on integrating ideas from these two threads to endow language models with "built-in" probing capabilities with regards to entity tracking.
I have also worked at the intersection of speech and text understanding [4, 5, 6, 7], and end-to-end speech recognition [8, 9, 10].
I am on the job market for research positions in industry.
Featured in the first episode of PyTorch Lightning Community Spotlight discussing our recent work Learning Chess Blindfolded: Evaluating Language Models on State Tracking.
Learning Chess Blindfolded: Evaluating Language Models on State Tracking
Shubham Toshniwal, Sam Wiseman, Karen Livescu, Kevin Gimpel
[Code and Data] [Colab - Play against LM] [bib]
[Pytorch Lightning Community Spotlight]
A Cross-Task Analysis of Text Span Representations
Shubham Toshniwal, Haoyue Shi, Bowen Shi, Lingyu Gao, Karen Livescu, Kevin Gimpel
[Code] [Slides] [Talk] [bib]
Pre-trained Text Embeddings for Enhanced Text-to-Speech Synthesis
Tomoki Hayashi, Shinji Watanabe, Tomoki Toda, Kazuya Takeda, Shubham Toshniwal, Karen Livescu
A Comparison of Techniques for Language Model Integration in Encoder-Decoder Speech Recognition
Shubham Toshniwal, Anjuli Kannan, Chung-Cheng Chiu, Yonghui Wu, Tara N Sainath, Karen Livescu
IEEE Workshop on Spoken Language Technology (SLT), 2018
Hierarchical Multitask Learning for CTC-based Speech Recognition
Kalpesh Krishna, Shubham Toshniwal, Karen Livescu
Parsing Speech: A Neural Approach to Integrating Lexical and Acoustic-Prosodic Information
Trang Tran*, Shubham Toshniwal*, Mohit Bansal, Kevin Gimpel, Karen Livescu, Mari Ostendorf
NAACL HLT 2018 (Oral)
Multilingual Speech Recognition With A Single End-To-End Model
Shubham Toshniwal, Tara N. Sainath, Ron J. Weiss, Bo Li, Pedro Moreno, Eugene Weinstein, Kanishka Rao
ICASSP 2018 (Oral)
[Slides] [Blog] [bib]
Multitask Learning with Low-Level Auxiliary Tasks for Encoder-Decoder Based Speech Recognition
Shubham Toshniwal, Hao Tang, Liang Lu, Karen Livescu
Interspeech 2017 (Oral)
[Slides] [Summary] [bib]
Jointly learning to align and convert graphemes to phonemes with neural attention models
Shubham Toshniwal, Karen Livescu
IEEE Workshop on Spoken Language Technology (SLT), 2016
[Code] [Poster] [bib]
System and method for cognitive filtering of audio in noisy environments
Jitendra Ajmera, Nitendra Rajput, Saurabh Srivastava, Shubham Toshniwal
U.S. Patent No. 10,187,738 B2, issued January 22, 2019
Generating natural language dialog using a questions corpus
Jitendra Ajmera, Ajay K. Gupta, Sachindra Joshi, Shubham Toshniwal
U.S. Patent No. 10,049,152 B2, issued August 14, 2018
Visual information processing allocation between a mobile device and a network
Anirban Majumder, Samik Datta, Sharad Jaiswal, Nisheeth Shrivastava, Sreedal Menon, Shubham Toshniwal
U.S. Patent No. 8,913,838 B2, issued December 16, 2014
Love to travel. Here are a selected few photos from those travels since starting PhD.
Amateur chess player - Lichess
I have a weird sense of humor and here are some catchphrases that I use and which make me tickle: