Recommending in Context: A Spreading Activation Model that is Independent of the Type of Recommender System and Its Contents

Current recommender systems make recommendations mostly independent of context. The recommender systems field is starting to acknowledge that systems need to incorporate the context in or for which they make recommendations to improve results. We show how recommender systems can make use of CASAN nets, our extension of the classical associative spreading activation network formalism. In CASAN nets, content nodes are not just connected by plain directed weighted links but also have associated link type and context nodes. Our spreading activation algorithm incorporates mechanisms to handle these two new functions of nodes. Contextual behavior is achieved because the activation values of link type and context nodes modulate the strengths of the links they are associated with. Thus, the activation pattern of the nodes associated with links establishes a context. This changes the network topology on whose basis a system chooses recommendations.