Improving sentiment rating of movie review comments for recommendation
Jenq‐Haur Wang, Ting‐Wei Liu · 2017
People usually ask for advices before making decisions, for example, watching movies. It's convenient if user opinions can be automatically aggregated and analyzed. To obtain an exact rating of a movie, sentiment rating can be formulated as a regression problem. Since our goal is only an overall suggestion of worthwhile or not, lexicon-based sentiment classification is used in this paper. To facilitate efficient movie recommendation, we propose to adjust sentiment lexicons for improving the sentiment classification accuracy in movie reviews. Also, we compare several methods of opinion rating aggregation for movie recommendation. In our experiments on Chinese movies, we can obtain high accuracy for the top-rated movies using our proposed approach. Further investigation is needed to evaluate the performance in larger scale.