Events Calendar

CS Colloq: Dr. Liang Huang

Talk Title: Packed Forest for Parsing and Translation Speaker: Dr. Liang Huang Host: Prof. Paul Rosenbloom Abstract: A major challenge in natural language research is how to deal with ambiguity. This is because ambiguity grows *exponentially* with sentence length (for example, consider the number of interpretations of a 60-word long sentence). This suggests that the common practice of a k-best list largely under-represents the whole search space. So is there a better way to efficiently represent and exploit the vast amount of ambiguities in human languages? The answer is "packed forest", a polynomial-space data-structure for representing exponentially many trees in a compact form by sharing common substructures. In this talk, I will apply this forest idea to both machine translation and syntactic parsing. In both tasks, we show that working with a forest encoding millions of trees improves the state-of-the-art accuracies by considering many more alternatives. This results in the best parsing performance reported on the Penn Treebank. More interestingly, translating a forest of 10 millions
trees is even faster than translating 30 individual trees, thanks to dynamic programming. (This talk is intended for a general CS audience.)
Bio: Liang Huang is currently a Research Scientist at ISI. He received his PhD from the University of Pennsylvania in 2008, and worked as a Research Scientist at Google before moving back to ISI where he had two internships. He is mainly interested in the theoretical aspects of computational linguistics, in particular, efficient algorithms in parsing and machine translation, generic dynamic programming, and formal properties of synchronous grammars. His work received an Outstanding Paper Award at ACL 2008, and Best Paper Nominations at ACL 2007 and EMNLP 2008. He also loves teaching, and won a University
Teaching Award at Penn in 2005. He is currently teaching CS 562, Empirical Methods in Natural Language Processing. http://www.isi.edu/~lhuang
http://www.cis.upenn.edu/~lhuang3

Location: Seaver Science Library (SSL) - 150

Audiences: Everyone Is Invited

Contact: CS Front Desk