Associating Semantic features with Intersective Levin classes

点击翻译按钮获取中文摘要

The difficulty of achieving hand-crafted semantic representations has limited the field of natural language processing. The field has yet to develop a clear consensus on guidelines for a computational lexicon that could provide a springboard for automated methods of lexical acquisition. One of the most controversial areas has to do with polysemy. This is an effort to address polysemy with respect to verbs, and how regular extensions of meaning can be achieved through the adjunction of particular syntactic phrases. Verb classes are seen as the key to making generalizations about regular extensions of meanings. This is an effort to continue development of a set of intersective classes compatible with the verb synonym set in WordNet, with more syntactic information and a more explicit mapping of semantic features to particular classes. The outcome of this effort will be an explicit association of semantic components and sets of syntactic frames to the 4000 verbs of Levin, and a mapping of those verbs into existing WordNet senses. Small transfer lexicons for English, French, and Portuguese using tree-adjoin grammars as the framework will be developed.