The semantic augmentation of a psycholinguistically-motivated syntactic formalism
Asad Sayeed, Vera Demberg · 2013
We augment an existing TAG-based incremental syntactic formalism, PLTAG, with a semantic component designed to support the simultaneous modeling effects of thematic fit as well as syntactic and semantic predictions. PLTAG is a psycholinguistically-motivated formalism which extends the standard TAG operations with a prediction and verification mechanism and has experimental support as a model of syntactic processing difficulty. We focus on the problem of formally modelling semantic role prediction in the context of an incremental parse and describe a flexible neo-Davidsonian formalism and composition procedure to accompany a PLTAG parse. To this end, we also provide a means of augmenting the PLTAG lexicon with semantic annotation. To illustrate this, we run through an experimentally-relevant model case, wherein the resolution of semantic role ambiguities influences the resolution of syntactic ambiguities and vice versa.