Mining Affective Context in Short Films for Emotion-Aware Recommendation

Emotion is fundamental to human experience and impactsour daily activities and decision-making processes where,e.g., the affective state of a user influences whether or notshe decides to consume a recommended item movie, book,product or service. However, information retrieval and recommendationtasks have largely ignored emotion as a sourceof user context, in part because emotion is difficult to measureand easy to misunderstand. In this paper we explore therole of emotions in short films and propose an approach thatautomatically extracts affective context from user commentsassociated to short films available in YouTube, as an alternativeto explicit human annotations. We go beyond the traditionalpolarity detection (i.e., positive/negative), and extractfor each film four opposing pairs of primary emotions:joysadness, angerfear, trustdisgust, and anticipationsurprise. Finally, in our empirical evaluation, we show howthe affective context extracted automatically can be leveragedfor emotion-aware film recommendation.