Possible Extensions to FrameNet: Formulation of a General Syntax-Semantics Interface

L arge-scale lexical semantic resources that provide relational information about words have recently received much focus in Natural Language Processing (NLP). In particular, corpora with predicate-argument structure annotation have been employed as the backbone for the development of shallow semantic parsing systems that automatically identify the semantic relationships, or semantic roles, conveyed by sentential constituents (Gildea and Jurafsky, 2002). Given an input sentence and a predicator the system labels constituents with general roles like Agent, Patient, Theme, etc. or more specific roles, as in (1). As a first step towards text understanding semantic role labeling has proved useful for a variety of NLP applications (including Information Extraction, Question Answering, Machine Translation, Summarization).