Emotions based movie recommender systems

Abstract

In this study we focus on development of emotions based movie recommender system. Emotions based movie recommender systems have the potential to provide personalized recommendations by matching movie affective content with user emotional needs and preferences. The user emotional preferences can be affected by such factors as mood, social context and personality, which are hard to infer in unobtrusive manner. We propose a new approach to infer user emotional preferences based on the affective content of the previously watched movies. We explore the potential of emotions mined from the review text to to resemble the actual emotions the movie induces from the viewers. Unsupervised emotion detection methods based on vector space model are used to extract emotions and build emotion profile for each movie which we further incorporate into state-of-the-art model for rating prediction and elicitation of user emotional preferences. We show promising results from application of emotions in movie recommender systems and propose possible way to analyze obtained movie and user emotional profiles with the aim to get more insights on the user preferences.