Classification des réponses d'un moteur de recherche et évaluation de leur pertinence

In this paper, we propose a method for evaluation of relevance in Web pages. This work joins in the general framework of Information Retrieval (IR) and more precisely, with the aim of constructing an automatic summary in encyclopaedic style. This summary type allows a new approach of Web page classification. In this paper, we present our classification approach and we detail our method for Web page relevance evaluation. This method is based on form criteria allowing to estimate the global form of the page and its capacity to produce a summary, statistical criteria verifying the correspondence between the request and the general topic of pages and linguistic criteria that guarantee a good linguistic form of pages. MOTS-CLÉS : recherche d’information, classification des pages Web, évaluation de la pertinence.