Fine-Grained Emotion Classification from Mobile App Reviews: An Empirical Study with Large Language Models

The paper investigates how large language models can be used for automatic multi-label emotion classification of mobile app reviews under severe class imbalance. It compares encoder-only fine-tuning, decoder-only prompting, and imbalance mitigation strategies using a synthetic-review generator and prompting strategy selected via an intrinsic augmentation-utility ranking. The study builds on a previously published annotation framework and human-labelled ground truth adapted from Plutchik's taxonomy.