Resolving polysemy in verbs: Contextualized distributional approach to argument semantics
Anna Rumshisky · 2008
Natural language is characterized by a high degree of polysemy, and the majority of content words accept multiple interpretations. Native speakers rely on context to assign the correct sense to each word in an utterance. Natural language processing (NLP) applications, such as the automated word sense disambiguation, require the ability to identify correctly context elements that activate each sense. Our goal in this work is to address the problem of contrasting semantics of the arguments as the source of meaning differentiation for the predicate. We show how the use of insight about the data can help design a targeted distributional approach to this problem. We consider the bidirectional nature of selection processes between the predicate and its arguments and the related problem of overlapping senses. The same sense of a polysemous predicate is often activated by semantically diverse arguments. We introduce the notion of contextualized distributional similarity