Time: 215pm-330pm, Friday, April 8, 2011
Place: Room 4102, CUNY Graduate Center, 365 Fifth Ave (34str&35str).
Speaker: Prashant Parikh (Penn)
Title: Computing Meaning

Abstract:

I will apply ideas from philosophy of language and semantics to the task 
of natural language understanding and related tasks like parsing and 
speech recognition.

I will first describe my framework of equilibrium semantics that enables 
one to derive the meaning of an utterance from first principles by 
modeling it as a system of interdependent games. The model reduces the 
problem of natural language understanding to the numerical task of 
checking which system of linear inequalitiesamong multiple systemsis 
valid. This presupposes that certain key inputs can be made available.

If certain simplifying assumptions are made, the linear inequalities 
further reduce to a generalization of a Bayesian form for disambiguation 
commonly used in NLP. This suggests that there is a deeper connection 
between my framework and work in statistical NLP. Since the game theory 
adds a dimension of efficient action to the standard assumption of a 
probability distribution underlying an utterance, it may have greater 
power than mainstream approaches.

It is possible to extend these ideas to syntax and phonology in principle 
and I conjecture that if such a program were carried out, it would enable 
one to identify the expressions in an utterance and parse it as well.

In the talk, my focus will be on the frameworks computational 
implications, but I should mention in passing that there are several 
theoretical implications for the science of language as well. Four of 
these are a generalization of Fregean compositionality, a derivation of 
the Gricean conversational maxims, a definition of communication, and a 
universality principle for games of partial information.

Speaker's Bio:

Prashant Parikh is a Senior Research Scholar at IRCS at Penn and is the 
Founder and CEO of Noema, Inc., a natural language software company in New 
York. He pioneered game-theoretic semantics and pragmatics in the 
mid-eighties. He is the author of The Use of Language (CSLI Publications, 
2001) and Language and Equilibrium (MIT Press, 2010). He did his SB and SM 
at MIT in the seventies and his PhD at Stanford University with Ken Arrow 
and Jon Barwise in the eighties.