WebSim: A Web-based Semantic Similarity Measure

抄録

Semantic similarity measures are important for numerous tasks innatural language processing such as word sense disambiguation,automatic synonym extraction, language modelling and document clustering. We propose a method to measure semanticsimilarity between two words using information availableon the Web. We extract page counts and snippets for the AND queryof the two words from a Web search engine. We define numerous similarity scoresbased on page counts and lexico-syntactic patterns. These similarity scoresare integrated using support vector machines to form a robust semanticsimilarity measure. Proposed method outperforms all existing Web-basedsemantic similarity measures on Miller-Charles benchmark dataset achievinga high correlation coefficient of 0.834 with human ratings.