Research

Research Interests:

My research area is Natural Language Processing (NLP). In particular, I have worked on a unification-based parser called LINK. LINK is a syntax-semantics integrated unification-based system that achieves average-case linear-time performance for limited domain texts. LINK gains this efficiency by encoding syntax, semantics and domain knowledge in a uniform way, and by utilizing all those information during parsing to prune unsuccessful parses. Because of this, LINK gives an elegant and efficient solution to parsing with unification grammars. In the current implementation, we developed an algorithm called LC, which is a variant of left-corner parsing, and further improved the performance by an optimization technique called lazy unification. For my PhD thesis, I formalized the algorithm in logic and proved the correctness of the algorithm. Here is the link to my thesis "Left-corner parsing algorithm for unification grammars".

I have also been involved in a project called FAQ Finder, a joint research by the AI groups at Northwestern University, University of California, Irvine and DePaul University. FAQ Finder (online at Irvine or at U of C) is a web-based question-answering system which retrieves answers to users' questions from FAQ files (thus it is a kind of Information Retrieval system). I mainly worked on an NL parser which uses a question grammar, and an automatic Q&A tagging module.  Information on recent progress is found here.

Recently I have also been interested in the area of Lexical Semantics, in particular a phenomena called systematic polysemy: senses of a word that are related in systematic and predictable ways. Using WordNet as the lexical resource, I used systematic polysemy to collapse similar word senses and derive an appropriate level of sense granularity for semantic lexicon. 

Publications

Refereed Journal Articles

Refereed Conference & Workshop Papers

Tech Reports