Sentiment Analysis for Sarcasm of Video Gamers

The video game industry has been growing rapidly over the years, and video gamers have started to form a subculture where sarcasm is highly appreciated. Despite the important influence of the video gaming industry, the research focusing on the video game community is still very rare. In addition, the emerging large language models based on transformer architecture provide new tooling and perspective to study culture-related topics. Therefore, in this work, we would like to focus on studying the sarcasm culture among video gamers with the help of state-of-the-art large language models and massive game reviews. We show that a general large language model can be fine-tuned to specialize in capturing the gamer’s sarcastic sentiment in Steam game reviews. The fine-tuned model has a 16% point improvement in accuracy compared to general sentiment models.