The emergence of new communication technologies and the latest generation of smart-phone devices, allow the user to continuously access the Web, at any time and from any location with different devices. The introduction of the user and his environment in the research process is intended to solve the information overload problem and to increase the accuracy of the retrieval system. In this paper, we present a new approach to select the best sources from different heterogeneous sources by exploiting a multidimensional contextual user profile in a mobile environment, that includes the research situation "time, location ...", the device and the user preferences. We also present our approach to define user and source profiles, using external ontologies and a learning algorithm supervised by user results satisfaction score.