Different Sense Granularities for Different Applications

This paper describes an hierarchical approach to WordNet sense distinctions that provides different types of automatic Word Sense Disambiguation (WSD) systems, which perform at varying levels of accuracy. For tasks where fine-grained sense distinctions may not be essential, an accurate coarse-grained WSD system may be sufficient. The paper discusses the criteria behind the three different levels of sense granularity, as well as the machine learning approach used by the WSD system.