Mingda Chen PhD Candidate Toyota Technological Institute at Chicago 6045 S Kenwood Ave, Chicago, IL, 60637 Email: mchen [at] ttic [dot] edu [CV (as of Sept. 2021)] [Google Scholar]

About

I am a 6th-year PhD student at Toyota Technological Institute at Chicago advised by Prof. Kevin Gimpel. Previously, I have interned at Google AI and Facebook AI. I received my BS in Applied Math from Zhejiang University in 2015.

I am supported by a Google PhD Fellowship.

Research Summary

How can we build computational models capable of understanding and learning from the traces left by humans? To tackle this problem, my research seeks to tailor naturally-occurring textual data to serve as training corpora for various purposes and to establish challenging evaluation tasks. In particular,

• I constructed datasets to introduce new challenges: e.g., long-form text generation [ACL-Findings’21], long-form text summarization [ACL’22], and story generation [arXiv’21];

• I discovered training signals for existing natural language processing problems: e.g., large-scale language pretraining for generic representation learning [ICLR’20], and mining knowledge from Wikipedia categories [EMNLP-Findings’20];

• I designed learning objectives to improve interpretebility and controllability of neural models: e.g., learning disentangled sentence representations [NAACL’19] and syntactic exemplar-controllable generation [ACL’19].

Publications and Preprints