mohit iyyer // CV // Scholar // github // twitter

In Fall 2018, I'll be joining UMass Amherst as an assistant professor. Currently, I'm a Young Investigator at AI2 working with Luke Zettlemoyer. I received my Ph.D. in June 2017 from the Department of Computer Science at the University of Maryland, College Park, advised by Jordan Boyd-Graber and Hal Daumé III. My main research interest is in designing deep neural networks for both traditional NLP tasks as well as new problems that involve understanding creative language.

Previously, I received a BS from the Department of Computer Science at Washington University in St. Louis under the guidance of Yixin Chen. I was a research intern at MetaMind in Spring 2015 and Microsoft Research in Summer 2016.

  • Oct. 2018: talk at UVM Symposium on the Science of Stories
  • Jul. 2018: talk at TTIC Language Generation workshop
  • Jun. 2018: ELMo won best long paper at NAACL 2018!
  • Mar. 2018: talk at USC/ISI NL Seminar
  • Feb. 2018: three papers to appear at NAACL 2018 (adversarial paraphrasing, ELMo, and image colorization)
  • Feb. 2018: talk at Ursinus College on applications of machine learning to the digital humanities
  • Jan. 2018: talk at Indian Institute of Science, Bengaluru
  • Oct. 2017: submit your QA system to our human-computer QA competition at NIPS 2017!
  • Apr. 2017: COMICS data and code released here!
  • Jan. 2017: talk at CU Boulder Stats, Optimization, and Machine Learning seminar
  • Nov. 2016: talk at UMass Machine Learning & Friends Lunch
  • Nov. 2016: new paper on understanding comic book narratives and characters.
  • Nov. 2016: new paper and associated dataset for sequential semantic parsing.
  • Jun. 2016: our paper on characterizing fictional relationships won best long paper at NAACL 2016!
  • Apr. 2016: we are organizing a workshop at NAACL 2016 on human-computer question answering with great invited speakers and accepted papers!
  • May 2015: our quiz bowl robot recently faced off against a team of four Jeopardy champions. Watch the introduction to learn how it works and then check out the actual match! If you're interested, code for the entire system is also available.