JUMRv1: A Sentiment Analysis Dataset for Movie Recommendation

Nowadays, we can observe the applications of machine learning in every field, ranging from quality testing materials to building powerful computer vision tools. One such recent application is recommendation system, which a method that suggests products users based on their preferences. In this paper, our focus specific system called movie recommendation. Here, make use user reviews movies order establish general outlook about and then recommend other users. However, huge number available has baffled sophisticated review systems. Consequently, there need find extracting meaningful information classifying predicting sentiment each one. typical scenario, either be positive, negative, or indifferent movie. research articles field mainly consider as two-class classification problem—positive negative. The most popular work was performed Stanford Rotten Tomatoes datasets, are somewhat outdated. Our self-scraped IMDB website, have annotated into one three classes—positive, neutral. dataset JUMRv1—Jadavpur University Movie Recommendation version 1. For evaluation JUMRv1, took an exhaustive approach by various combinations word embeddings, feature selection methods, classifiers. We also analysed performance trends, if were any, attempted explain them. sets benchmark for systems newly developed using three-class classification.