What About Emotions? Guiding Fine-Grained Emotion Extraction from Mobile App Reviews

This paper explores the underexamined area of fine-grained emotion classification in app reviews, extending beyond the traditional focus on sentiment polarity (positive, negative, neutral). To capture the complexity of users’ affective responses, the study adapts Plutchik’s emotion taxonomy and introduces a structured annotation framework and dataset tailored to app reviews. Through an iterative human annotation process, the authors establish clear guidelines, highlight challenges in interpreting emotions, and assess the feasibility of automation with large language models (LLMs). The results show that LLMs substantially reduce manual annotation effort and achieve notable agreement with human annotators, though full automation remains difficult due to the nuanced nature of emotions. This work provides structured guidelines, an annotated dataset, and insights for building semi-automated pipelines, offering valuable contributions to opinion mining, requirements engineering, and user feedback analysis.