A significant number of video game reviews can be found on online platforms, such as Steam. These reviews
contain various information, including game evaluations and user experiences, which may have the potential to influence
consumer purchasing behavior. In this study, we extracted 14 review perspectives, including “game system,” “bugs,” and
“music”, from reviews and analyzed the differences in review perspectives for each game genre. The analysis employed a
dataset comprising 178,353 Japanese reviews of 300 game titles, obtained from the Steam platform. To extract review
perspectives, a large language model was employed, in conjunction with genre information defined by users. The analysis
using the IQR method revealed that certain tags were identified as outliers for each review perspective, indicating a
correlation between review perspectives and genres.
View full abstract