Broad Area Colloquium For AI-Geometry-Graphics-Robotics-Vision

Domains of Locality: Starting With Complex Primitives Pays Off

Aravind K. Joshi
Department of Computer and Information Science
Institute for Research in Cognitive Science
University of Pennsylvania

Monday, May 14, 2001, 4:15PM


Each grammar formalism specifies a domain of locality, i.e., a domain over which various dependencies (for example, syntactic and semantic) can be specified. It turns out that the various properties of a formalism (syntactic, semantic, computational, statistical, and even psycholinguistic), follow, to a large extent, from the initial specification of the domain of locality. In this talk, I will briefly explore a domain of locality specified by structured objects (trees or acyclic graphs) instead of strings, in the context of some linguistic, computational, statistical and psycholinguistic properties. Such studies provide insights into many aspects of strong generative capacity which is relevant to characterizing structural descriptions. Recently, they have also found some applications to the description of secondary and higher structures of some biological sequences.

About the Speaker

Aravind Joshi received his undergraduate education in Electrical Engineering at the University of Pune and the Indian Institute of Science, Bangalore, India and his Ph. D. Degree in Electrical Engineering at the University of Pennsylvania. At present he is the Henry Salvatori Professor of Computer and Cognitive Science and Co-Director of the Institute for Research in Cognitive Science, with secondary appointments in the Departments of Linguistics and Psychology. He He is a former Guggenheim Fellow, Fellow of IEEE and ACM, and Founding Fellow of AAAI. He is a member of the National Academy of Engineering. His research interests are in the various areas of natural language processing--computational modeling of syntax, semantics and pragmatics and discourse, mathematical linguistics, psycholinguistic implications of processing models, and applications to machine translation, question-answer systems and information extraction.


Back to the Colloquium Page