Enhancing text clustering performance using semantic similarity

Abstract

Text documents clustering can be challenging due to complex linguistics properties of the text documents. Most of clustering techniques are based on traditional bag of words to represent the documents. In such document representation, ambiguity, synonymy and semantic similarities may not be captured using traditional text mining techniques that are based on words and/or phrases frequencies in the text. In this paper, we propose a semantic similarity based model to capture the semantic of the text. The proposed model in conjunction with lexical ontology solves the synonyms and hypernyms problems. It utilizes WordNet as an ontology and uses the adapted Lesk algorithm to examine and extract the relationships between terms. The proposed model reflects the relationships by the semantic weighs added to the term frequency weight to represent the semantic similarity between terms. Experiments using the proposed semantic similarity based model in text clustering are conducted. The obtained results show promising performance improvements compared to the traditional vector space model as well as other existing methods that include semantic similarity measures in text clustering. © 2009 Springer Berlin Heidelberg.