Journal of E-Technology Omar NOUALI, Amokrane BELLOUI, Nadia Taboudjemat 1 3 2010 http://www.dline.info/jet/fulltext/v1n3/3.pdf Collaborative filtering systems have some limitations such as cold-start problems for a new user, a new resource or both. In this paper, we show that using semantic information describing users and resources can reduce the problems and lead to a better precision, coverage and quality for the recommendation engine. Semantic web is the infrastructure used for managing such semantic descriptions. We have evaluated several strategies of filtering hybridization: weighting, adaptive change or switching and feature combination.