A stereotypes-based hybrid recommender system for media items

Many Recommender Systems use either Collaborative Filtering (CF) or Content-Based (CB) techniques to receive recommendations for products. Both approaches have advantages and weaknesses. Combining the two approaches together can overcome most weaknesses. However, most hybrid systems combine the two methods in an ad-hoc manner. In this paper we present an hybrid approach for recommendations, where a user profile is a weighted combination of user stereotypes, created automatically through a clustering process. Each stereotype is defined by an ontology of item attributes. Our approach provides good recommendations for items that were rated in the past and is also able to handle new items that were never observed by the system. Our algorithm is implemented in a commercial system for recommending media items. The system is envisioned to function as personalized media (audio, video, print) service within mobile phones, online media portals, sling boxes, etc. It is currently under development within Deutsche Telekom Laboratories - Innovations of Integrated Communication projects.