Application of Latent Dirichlet Allocation Algorithm on User Review Analysis: A Case Study on Online Games

The online gaming market has grown rapidly. To study online game-related issues, most researchers either relied on consumer survey questionnaire data or focused on experts' comments to evaluate consumers' perceptions of an online game. Only a few studies explored user reviews using user-generated content (UGC) of game players on social networks. UGC contains rich information about players' feelings, helping game developers understand players' gaming experiences and thoughts. Therefore, we analyzed the UGC of three role-playing games (RPG) by text mining techniques, including word identification and Latent Dirichlet Allocation (LDA) algorithms. We explored players' experiences and attitudes toward the three RPG games. The player review content was extracted from the Steam website for several topics such as game characters, game graphics, and game stories, which are regarded as key attributes of an online game. It was found that most players focused on the rules and objectives of role-playing games. Since the three games have different contents, their LDA topic models showed different topic attributes. The study results provide an applicable method for game developers to identify the topics players and suggestions for necessary improvements.