Analysis of Golf Application Reviews Using Text Mining: Application of Machine Learning-Based Sentiment Analysis and Topic Modeling

OBJECTIVES This study aimed to analyze golf application reviews using text-mining techniques. Specifically, the study sought to identify key review keywords, develop a sentiment dictionary for review classification, and examine major topics within positive and negative reviews.METHODS Five golf-related applications were selected for analysis. User reviews and rating data were collected from Google Play and the App Store using a Python-based automated web crawler. To identify key terms, Term Frequency and Term Frequency-Inverse Document Frequency values were calculated and compared. A Lasso model was employed to develop the sentiment dictionary, and the Latent Dirichlet Allocation algorithm was applied to identify major topics within positive and negative reviews.RESULTS The findings were as follows. First, high-frequency keywords included good, app, convenient, and bad. When document-specific weighting was considered, prominent keywords included nice, good, convenient, and app. Keywords demonstrating high volatility across frequency and weighting indices included positive action-oriented terms such as satisfied, useful, easy, fun, and grateful. These results indicate that keywords such as “convenient,” “satisfied,” “useful,” “easy,” and “fun” play a central role in review classification. Second, the sentiment dictionary achieved a classification accuracy of 87%. Positive sentiment keywords included “good,” “convenient,” “best,” and “friendly,” whereas negative sentiment keywords included “no,” “update,” and “certification.” Third, six topics were identified from positive reviews. The four topics with the highest allocation ratios accounted for 87% of all positive-review content. These topics were summarized as “customized and convenient golf reservation systems,” “satisfaction with clubhouse and caddy services,” “personalized information service,” and “satisfaction with photos and scorecards.” Fourth, seven topics were identified from negative reviews. The two dominant topics accounted for 69.1% of the allocation ratio and were “functional degradation after app updates” and “user authentication and login errors.” Additional concerns included excessive advertising exposure and dissatisfaction with golf course reservation cancellation policies.CONCLUSIONS In conclusion, this study developed a specialized sentiment dictionary for analyzing golf application reviews and identified major themes associated with positive and negative user experiences. The findings provide foundational data for guiding future application development. Although the addition of new functions remains important, minimizing login and update-related errors appears to be a critical determinant of perceived service quality among users.