Knowledge Lean Word–Sense Disambiguation

Wepresent a corpus-based approach to word-sense disambiguation that only requires information that can be automatically extracted from untagged text. Weuse unsupervised techniques to estimate the parameters of a modeldescribing the conditional distribution of the sense group given the knowncontextual features. Both the EMalgorithm and Gibbs Sampling are evaluated to determine which is most appropriate for our data. Wecompare their disambiguation accuracy in an experiment with thirteen different words and three feature sets. Gibbs Samplingresults in small but consistent improvementin disambiguation accuracy over the EMalgorithm.

Authors

• Ted Pedersen Richard Maclin

Keywords

No keywords are indexed for this paper.