Abstract—The explosion of textual documents on the Web (journals, social networks, blogs) generated a need to treat this mass of data to extract knowledge. We are interested in this paper in a particular treatment which is text categorization. We propose a conceptual representation of texts by using Wordnet for documents categorization. This representation is based on terms disambiguation by using Wordnet concepts. Disambiguated concepts are extracted from representative terms of a document, and three representations (terms, concepts, terms+ concepts) are applied with three training algorithms: SVM, Decision trees, KNN for the categorization. Experiments were applied on two corpora: 11 categories of reuters-21578 articles and 7 categories of 20 newsgroup discussion documents. The use of (terms+ concepts) gave better results for the three training algorithms and especially for the decision trees.

Index Terms—Text mining, categorization, classification, wordnet.

The authors are with the “LSSD” Laboratory and the Computer Science Department, Faculty of Mathematics and Computer Science of the University of Sciences and Technology of Oran Mohammed Boudiaf, Algeria (e-mail: nezreg.houaria@univ-usto.dz, lehbabhayet@yahoo.fr, h_belbach@yahoo.fr).

Cite:H. Nezreg, H. Lehbab, and H. Belbachir, "Conceptual Representation Using WordNet for Text Categorization," International Journal of Computer and Communication Engineering vol. 3, no. 1, pp. 27-30, 2014.