A Public Talk presented by Dr. Sandiway Fong

(Please Note - This talk will be delivered via Skype)

Over the past two decades, there has been increasing use of linguistically annotated sentence collections, such as the Penn Treebank (PTB), for constructing statisticallybased parsers across a variety of languages. Such Treebank-based parsers are trained on and exploit syntactic regularities in (manually-annotated) phrase structures - statistically interpolating where data is missing - in order to synthesize a most-likely parse when presented with novel (and pre-existing) sentences. Within this framework, one must rely on the promissory note that all necessary grammatical knowledge is encapsulated and statistically extractable from the Treebank corpus. In this talk we discuss the practical limits of this approach, cognitive implications for the problem of language acquisition, and finally, on-going Treebank work for a bilingual English/Arabic roboceptionist (Hala), a joint project with Carnegie Mellon University
(Qatar and Pittsburgh).

Co-sponsored by the School of Modern Languages, Ivan Allen College Dean's Office and the School of Interactive Computing.

Biography of Dr. Fong:  Sandiway Fong received his B.Sc. in Computing Science, at Imperial College of Science and Technology, University of London. He received an S.M. in 1986 at MIT, where he worked in the Artificial Intelligence Laboratory.  After working at IBM's Watson Research Center, he returned to MIT  for his Ph.D. which was awarded in 1991.  Upon graduation Dr. Fong joined the NEC Research Institute to work on natural language processing, and machine translation.  In 2003, he moved to the University of Arizona, where he is now an Associate Professor.  His research interests are at the intersection of computer science and formal linguistics, with a focus on multilingual parsing, ontolinguistics, computational lexical semantics and computational morphology.