Inferring User's Preferences using Ontologies

We consider recommender systems that filter information and only show the most preferred items. Good recommendations can be provided only when an accurate model of the user's preferences is available. We propose a novel technique for filling in missing elements of a user's preference model using the knowledge captured in an ontology. Furthermore, we show through experiments on the MovieLens data set that our model achieves a high prediction accuracy and personalization level when little about the user's preferences is known.