MSDA: Wordsense Discrimination Using Context Vectors and Attributes
Abdulrahman Almuhareb, Massimo Poesio · 2006
Abstract. We present MSDA (Major Senses Discovery Algorithm) – a development over the context vector approach to (noun) sense discrimination [20, 24] that uses attributes and values instead of word features to cluster contexts, and does not require for the number of senses to be fixed beforehand. The algorithm achieves a precision of 89% on a dataset including both ambiguous and non-ambiguous nouns, twice that of previous algorithms. 1