Unifying Lexical Resources

This paper describes our efforts to create a Unified Lexicon by extracting information from a variety of external resources, namely our XLE syntactic lexicon, WordNet, Cyc, and VerbNet. The UL is built in several steps: first the information is extracted from the resources; then it is merged into lexical entries based on word stem, syntactic subcategorization frame, meaning concept, and WordNet class; finally, patch files are run over the UL to create a cleaner version. The patched version of the UL is used to extract semantics to KR mapping rules, including default rules for where gaps occur in the external resources. This paper focuses on unifying lexical resources for verbs.