Analysis Hybrid Metrics for Emotion Detection (Case Study: Gaming Context)

Abstract

The rising popularity of online gaming has positioned Steam as a leading platform for accessing diverse games. Beyond gameplay, Steam enables users to submit reviews, offering valuable data for analyzing emotional tone and classifying feedback as positive or negative. This study analyzed 8,000 reviews from Steam across four games: The Sims 4, Counter-Strike 2, FIFA 23, and Dead by Daylight. Plutchik’s emotion theory, with its eight basic emotions served as the foundation for classification, utilizing the NRC Lexicon as an emotional dictionary. A hybrid algorithm combining cosine similarity (70%) and Euclidean distance (30%) with a threshold mechanism was employed to label emotions. Reviews exceeding the threshold received specific emotion labels, while others were classified as "unknown." Positive reviews, associated with joy, trust, fear, and surprise, were predominant for The Sims 4. Conversely, Counter-Strike 2, FIFA 23, and Dead by Daylight garnered largely negative reviews, highlighting the utility of emotional analysis in evaluating user feedback.