A comparison of audiovisual content recommender systems performance: Collaborative vs. semantic approaches

In this paper we compare the efficiency of the semantic approach with the traditional collaborative filtering in recommender systems applied to the field of audiovisual contents. We use actual data from MovieLens database to train the recommenders' algorithms and to compare results. An ontology for the audiovisual content was developed importing data from IMDB. For the sake of reproducibility the novel LensKit platform has been utilized.