Analyzing users’ sentiment towards video games based on reviews from microblog

Abstract

This project proposes a new model of sentiment analysis for video game’s reviews. In these days people tend to check reviews and ratings of video games before spending money and time for a game. In the proposed model, ratings for video game will be generated by doing sentiment analysis on public opinion. As Twitter is one of the most popular micro-blogging sites, for public opinion we collected data from Twitter. Before fitting the algorithms we preprocessed the gathered data to a supervised form. In the model Naïve Bayes, Support Vector Machine, Logistic Regression and Stochastic Gradient Descent algorithm were used for performance comparison. They were trained on a training set and to validate the performance the algorithms were tested several times on a test set to get better accuracy. After that a new classifier was used which acted as a voting classifier for the algorithms. This classifier was used for sentiment analysis on the data to get polarity. To validate the model, we generated rating from calculating polarity for each attribute which contains gameplay, graphics, sound, multiplayer and plotted in a graph where results are shown.