Système de recommandation à base de connaissances : Résolution de problèmes à base de modèle ontologique.

Abstract

Recommender systems are a specific form of information filtering (IS) aimed at presenting the pieces of information (movies, music, books, news, images, Web pages, etc.) that are likely to be of interest to the user. Generally, a recommendation system makes it possible to compare the profile of a user with certain reference characteristics, and seeks to predict the "opinion" that a user would give. Major challenges constrain recommender systems, such as the problem of diversity of recommender systems, stability with respect to the dynamic profile of users, as well as a generic problem of these systems which is cold start. Classic recommendation approaches such as collaborative filtering, content-based filtering, hy- brid filtering, etc. have shown their effectiveness in this context. Knowledge engineering, and more specifically knowledge-based systems have played an important role in improving the performance of such systems. This is the framework of our work which suggests ways to improve the RS. A first contribution is focused on knowledge where we built an ontological model allowing case-based rea- soning in order to extract cases similar to a given problem, a second contribution is focused on the evaluation of recommendations by similarity metrics in order to validate the reasoning.