Movies Recommendation Based on Opinion Mining in Twitter

A traditional way for movie recommendation in a real scenario is by word of mouth. People ask their friends or relatives their opinion about a movie and then make their own judgment about whether to go to see the movie. In this article, we take this paradigm to evaluate Twitter as a source of information for movie recommendation. We built a balanced dataset consisting of 3036 tweets expressing opinions regarding movies. Then, we evaluated different tokenization strategies, pre-processing techniques and algorithms to build classification models that are able to determine the sentiment (opinion + polarity) expressed in the short texts published in Twitter. Finally, the best classifier is used to extract the main sentiment of Twitter users regarding a target movie in order to help users to decide to see the movie or not, obtaining promising results.