Semantically-enhanced pre-filtering for context-aware recommender systems

Several research works have demonstrated that if users' ratings are truly context-dependent, then Context-Aware Recommender Systems can outperform traditional recommenders. In this paper we present a novel contextual pre-filtering approach that exploits the implicit semantic similarity of contextual situations. For determining such a similarity we rely only on the available users' ratings and we deem as similar two syntactically different contextual situations that are actually influencing in a similar way the user's rating behavior. We validate the proposed approach using two contextually tagged ratings data sets showing that it outperforms a traditional pre-filtering approach and a state-of-the-art context-aware Matrix Factorization model.