A team from UW CSE’s Natural Language Processing group has earned Best Paper accolades at EMNLP 2016, the conference on empirical methods in natural language processing. CSE Ph.D. student Kenton Lee, postdoc Mike Lewis, and professor Luke Zettlemoyer won for their paper Global Neural CCG Parsing with Optimality Guarantees.
The winning paper describes an approach for learning a recursive neural network for CCG parsing — a core subproblem in broad coverage semantic analysis of text. The model is not only state of the art in terms of accuracy but also provides optimality certificates, nearly always proving a proof that the best parse was found under the learned model. It is the first neural parser of any kind to provide such guarantees.
The paper is one of only two selected for Best Paper recognition from more than 900 submissions. The team will present its findings at the EMNLP conference in Austin, Texas early next month. This is the second year in a row that the UW NLP group has won at EMNLP, with Lee and Zettlemoyer repeating their feat from 2015.