In this paper we provide a solution for recommending multimedia materials inside an e-learning platform, which considers the binding of user profile information with document annotations via the exploitation of domain ontologies. The user profile is structured on three levels – competences, interests and fingerprints – all expressed through ontological constructs, as also the multimedia materials annotations are. The user current activity (his fingerprints) is supervised in terms of conceptual ontology navigation (which ontology concepts are related to the currently accessed materials). For establishing the suitable topics for being recommended at each moment, a Markov decision process is used.