Toward Discourse-guided Theta-grid Chart Parsing for Madarin Chinese － A Preliminary Report

An attempt for this work is to show a way of combining word identification, syntactic processing, semantic processing, and discourse processing into a cohesive framework. We utilize thematic information as a media to integrate these processing modules. In this work, the thematic information is assumed to reside in lexicons based on the theta grid theory. For determining the main verb(s) of a sentence with Serial Verb Constructions (SVCs), we propose an algorithm which evaluates a scoring function; examples showing how Chinese texts with SVCs in the legal domain can be parsed by our parser are presented. We also show how the previously acquired anaphoric knowledge can be used to guide the chart parser.