Pdf.pdf

Researchers James Pustejovsky, Patrick Hanks, and Anna Rumshisky introduce a model for sense assignment that relies on assigning senses to contexts within which words appear, rather than to the words themselves. They argue that word senses are not directly encoded in the lexicon, but rather each word is associated with stereotypical syntagmatic patterns, called selection contexts, which are associated with meanings. The researchers present a formalism for encoding contexts and a methodology, called Corpus Pattern Analysis (CPA), for identifying these contexts from large corpora and structuring them in a selection context dictionary. The CPA technique uses a semi-automatic bootstrapping process to produce a dictionary of selection contexts for predicates in a language, distinguishing word senses for verbs through corpus-derived syntagmatic patterns mapped to Generative Lexicon Theory. The procedure consists of manual discovery of selection context patterns, automatic recognition of instances of these patterns, and automatic acquisition of patterns for unanalyzed cases. The researchers also discuss the Brandeis Shallow Ontology (BSO), a shallow hierarchy of types selected for their prevalence in manually identified selection context patterns, and present preliminary results.