Computational learning of subcategorization for nominals and pronominals: Portuguese and Turkish

Seminário LISP / DI 

Computational learning of subcategorization for nominals andpronominals: Portuguese and Turkish

 

 

Cem Bozahin is professor of cognitive science at the Informatics Institute, ODTU, Ankara. He holds a PhD in computer science from Arizona State University. His research topics are the nature of constraints on cognition, and investigation of natural grammars. He has conducted research and/or taught at ASU, Ohio University, ODTU, Bogazici University and Edinburgh University. He is visiting University of Lisbon for the period September 2015-May 2016.

 

In languages such as Turkish, there seems to be no difference in the word-order of nominal arguments and pronominal arguments.

However in Portuguese there are several restrictions along this dimension, although the two languages appear to be similar in freedom of word order.

These differences are explainable, in fact predicted, by a radically lexicalized and computationally efficient grammar, where reference is also lexicalized.

We aim to show that exposure to different kinds of referential distributions does not require revision in a theory of grammar, or parametric grammars. An effective algorithm for learning any natural grammar as a hidden variable is feasible.

 

Joint work with Mark Steedman

 

Host:

 

Teresa Gonçalves / LISP

Organization: LIPS
At 17.02.2016
CLV-Anf 1 l 16h:30m