Computational Linguistics Colloquium

Thursday, 8 January, 16:00
Conference Room, Building C7 4

The Influence of Discourse Information on Human Syntactic Ambiguity Resolution: A Reflection on Experimental and Computational Approaches

Amit Dubey
University of Edinburgh / Stuttgart University

This talk revisits an old debate in psycholinguistics concerning the degree of separation between the syntactic and semantic/discourse modules of the human sentence processor. Using a more sensitive methodology than previous studies, we find evidence for syntax-first behaviour in reading, whereby a (possibly incorrect) structure of a temporarily ambiguous sentence is first suggested by the syntactic module, and later updated by semantic and discourse information. This is in contrast to earlier findings which suggest a much closer cooperation between syntactic and semantic processing.
We present preliminary results of a computational simulation of this syntax-first behaviour using a statistical model of human sentence processing. Unlike earlier work, we use statistics not only for the syntactic modules, but also for the discourse and semantics using a novel approach for probabilistic logic.

If you would like to meet with the speaker, please contact Berry Claus.