A fuzzy radial basis model for arabic documents classification

Abstract

In this paper, we bring an improvement to the classical fuzzy model of classification by implementing a new approach which based on radial basis functions for the Arabic documents classification. This approach takes into account the concept of semantic vicinity by calculating of the similarity degree between terms in relation to the documents. We combine the calculation of the relevance of these terms (using NEAR operator) with a radial basis function to identify the relevant documents to the query. The use of linguistic resources namely semantic graphs and semantic dictionaries (specifically created for the studied domain) significantly improves the process of classification. Preliminary and promising results are shown on a Arabic press database which show very good performance compared to the literature.