Combining Independent Knowledge Sources for Word Sense Disambiguation

Disambiguation Yorick Wilks and Mark Stevenson Department of Computer Science, University of She eld, Regent Court, 211 Portobello Street, She eld S1 4DP, UK fyorick, [email protected] Abstract Sense tagging, the automatic assignment of the appropriate sense from some lexicon to each of the words in a text, is a specialised instance of the general problem of word sense disambiguation. We discuss which recent word sense disambiguation algorithms are appropriate for sense tagging. It is our belief that sense tagging can be carried out effectively by combining several simple, independent, methods and we include the design of such a tagger. A prototype of this system has been implemented, correctly tagging 88% of polysemous word tokens in a small test set, providing evidence that our hypothesis is correct. 1 Sense Tagging There has been a tendency in word sense disambiguation (WSD hereafter) literature to carry out di erent disambiguation tasks and classify them all as WSD procedures. There are however, at least, three di erent levels of algorithm in WSD. The most speci c of these are sense tagging procedures, these assign, to each word1 in a text, its particular sense from some lexicon and each word type having a set of senses speci c to it. This di ers from the more general case of semantic tagging, where the tags for each word (type) need not be speci c to that type and do not correspond to word senses in a lexicon. These tags may be broad semantic categories such as HUMAN or ANIMATE or WordNet synsets which may apply to any word in the text (similar to part of speech tagging in the sense that there is a class of tags which may apply to any token in the text). The most general class of algorithms are semantic disambiguation algorithms. These algorithms are procedures which carry out semantic disambiguation on words, these may not necessarily be tagging algorithms, in that they do not attempt to mark every token in a text but may be restricted to disambiguating small sets of word types. The class of sense tagging algorithms is a proper subset of the class of semantic tagging algorithms, which is, in turn, a proper subset of the class of semantic disambiguation algorithms. This hierarchical relationship is represented in Figure 1. 1This is often loosened to each content word. Semantic Tagging Tagging Sense Semantic Disambiguation