Context-based Multimedia Recommender System

Competence CenterInformation Retrieval and Machine Learning
ContactProf. Dr.-Ing. Sahin Albayrak, Alan Said, M.Sc.

Partner: moviepilot GmbH


There are many Internet portals that compute collaborative movie recommendations to present users with interesting movies to watch or to buy. Recommendations are determined based on a user’s similarity to other users, or based on explicitly specified preferences (e.g. genre "crime"). They are hence only minimally personalized, and disregard the current usage context (time of day, weekend) as well as other metadata (favorite actor).

Context-based movie recommendations

The goal of the KMULE project is the implicit identification of context-related preferences based on an analysis of users’ interaction histories and current usage contexts, and the implementation of a conceptual architecture for a context-aware movie recommender system. Key contextual and metadata features are identified and used for the creation of several sets of user-specific and context-aware recommendations. The recommendations computed by the system are expected to have a higher quality and higher user acceptance rate. The resulting prototype system is evaluated and incorporated into the movie recommendation website Moviepilot. 

KMULE is a cooperation project of DAI-Labor and moviepilot GmbH and is funded by the Federal Ministry of Economics and Technology.