Incremental Tree Adjoining Grammar (TAG) for Modelling Human Sentence Processing
Speaker: Vera Demberg
Institution: University of Edinburgh
Abstract:
I'll present a psycholinguistically motivated version of TAG which is designed to model key properties of human sentence processing, viz., incrementality, connectedness, and prediction. The decision to build an incremental parser is based on recent findings from psycholinguistic experiments; I define a variant of TAG that makes it possible to build fully connected structures on a word-by-word basis, and thus supports incrementality in its strict interpretation.
A key idea of the approach is to explicitly model the prediction of upcoming material and the subsequent verification and integration processes. The linking theory, which "translates" parser states into an estimate of sentence processing difficulty (specifically, predicted reading times), is based on modeling these prediction and verification processes.
