An intelligent film recommender system based on emotional analysis

Abstract

The existing personalized film recommendation methods take the user’s historical rating as an important basis for recommendation. However, the user’s rating standards are different, so it is difficult to mine the user’s real preferences and form accurate push. Therefore, to achieve high-quality personalized recommendation of films, it is particularly important to mine the emotion of user reviews. In this article, a personalized recommendation method based on sentiment analysis of film reviews is proposed, where natural language processing technology is used to mine the emotional tendency of user reviews. The multi-modal emotional features are weighted and the weighted fusion feature vector after PSO is taken as the overall emotion vector, then the emotional similarity of weighted fusion is calculated by considering the time factor of content publishing and the average emotional tendency of users. By calculating the matching degree of emotional value between users and films, the top-N film recommendation for target users is given. The test results show that the effect of the personalized film recommendation system based on multimodality is superior to that of the comparison method, which effectively solves the problem of different user rating scales, and really increases users’ interest in watching films.