Applied the Fuzzy Naïve Bayes Algorithm to Word Sense Disambiguating

Abstract

In this article, the Fuzzy Naïve Bayes algorithm is presented. This algorithm integrates the classical Naïve Bayes model with fuzzy logic, in order to address the complex problem of semantic disambiguation in Arabic. This task remains particularly challenging due to the morphological richness and lexical ambiguity of the Arabic language.The approach adopted aims to model the uncertainty linked to the multiple possible interpretations of a word in context by assigning each meaning a fuzzy degree of membership rather than a strict classification.The evaluation of the algorithm was conducted on three distinct corpora, utilising lexical and syntactic features. The performance obtained was systematically compared with that of the standard Naïve Bayes model.The experimental results demonstrate a substantial enhancement in terms of accuracy and robustness, underscoring the contribution of fuzzy logic to the management of semantic uncertainties.