Abstract

Purpose–The purpose of this thesis is to investigate the game reviews and to see how the natural language processing can be used to analyze game reviews. Using Steam platform as our data source and Latent Dirichlet Allocation as one of the main methods of data analysis in this research there will be attempt to figure out what kind of topics will emerge within and how interpretations of these can be used to see the correlation between game updates and game reviews. All these results will be looked throughout the timeframes of the life of the game, from alpha/beta testing to the most current reviews. Design/Methodology/Approach–Within this research Latent Dirichlet Allocation will be used on five different games from multiple different genres to test our hypothesis. Findings–The findings of this research could be used by the fields of both data analysis and game studies as this research contributes towards these areas. It is believed that this research will show how the game reviews can be interpreted and give ideas to further research. Practical implications–In theory this research will heavily be based on opinions of both players/people who left the reviews and a researcher hence these should not be treated as a fact. As there are many factors that could affect the quality of the reviews, there will be taken careful measures to eliminate these one by one (such as length of the reviews, duplicates. Etc.)Originality/value–This research should stimulate the game producers and developers to see a different aspect of the games and allow for better decision making. Key Questions–How natural language processing can be used to analyze this data? How will LDA and LDA Sequential contribute to analysis of Steam reviews? How does pre-processing change the results of the LDA and LDA Sequential? What is the correlation between the game updates and game review analysis results?