A multi-agent open architecture for a TV recommender system: a case study using a Bayesian strategy

In this paper we present a recommender system of personalized TV contents, called AVATAR, for which we propose a modular multiagent architecture, that combines different knowledge inference strategies (such as Bayesian techniques, profiles matching and semantic reasoning). We focus on the description of one of these strategies, the naive Bayesian classifiers, explaining an example in the context of personalized digital television. In order to represent the knowledge in the television domain, we have developed TV contents ontology, to infer new data from the known information. Besides, the TV-anytime specification has been used referred to the description of contents and the management of user preferences and their activity logs. The proposed recommender system has been conceived as an application conforming to multimedia home platform (MHP) standard, to be distributed over the broadcast transport stream that is tuned by the user receiver.