Michael Arbib

University of Southern California
Talk Title: 
Template Construction Grammar and the Description of Visual Scenes
Event Type: 
Linguistics Lecture Series
Spring 2010
Tuesday, March 2, 2010, 3:30 pm - 5:00 pm
SAC 302

The present study reports eye-movement studies inspired by our earlier computational modeling of the production of descriptions of visual scenes, using a spatially anchored layout of schema instances to provide the formal semantics for the description. We have developed Template Construction Grammar (TCG) as a new version of Construction Grammar which works through competition and cooperation of constructions to derive a verbal description from a semantic representation, SemRep, in the form of a graph-like hierarchical structure that bridges between vision and language.