May Social Behavior Reveal Preferences on Different Contexts? Recommending Movie Titles Based on Tweets

Abstract

Recommendation systems allow users to deal with information overload by generating personalized recommendations that guide them through the universe of available options. These systems have been successfully used for over a decade in various application domains. However, most recommendation techniques do not consider context, generating recommendations which do not consider user's daily generated information. The increasing use of social networks and microblogs has created a valuable source to extract information that might help improve the results of recommender systems. Specifically, microblogging messages may help on contextualized movie recommendation. In this paper, a hybrid recommender is proposed merging a collaborative filtering recommender with a content based recommender. The results confirm the usefulness of the proposed recommender, where the user's exchanged messages define a bias to insert context in the recommendations.