CHEER-Ekman: Fine-grained Embodied Emotion Classification

TL;DR

This paper introduces CHEER-Ekman, a new dataset for fine-grained embodied emotion classification in text, and demonstrates that prompt engineering with large language models enhances emotion recognition accuracy, even with smaller models.

Contribution

The paper presents a novel dataset extending embodied emotion classification to Ekman's six basic emotions and shows improved performance using prompting techniques with large language models.

Findings

Prompting with simplified instructions boosts accuracy.

Chain-of-thought reasoning improves emotion recognition.

Smaller models can perform competitively with larger models.

Abstract

Emotions manifest through physical experiences and bodily reactions, yet identifying such embodied emotions in text remains understudied. We present an embodied emotion classification dataset, CHEER-Ekman, extending the existing binary embodied emotion dataset with Ekman's six basic emotion categories. Using automatic best-worst scaling with large language models, we achieve performance superior to supervised approaches on our new dataset. Our investigation reveals that simplified prompting instructions and chain-of-thought reasoning significantly improve emotion recognition accuracy, enabling smaller models to achieve competitive performance with larger ones. Our dataset is publicly available at: https://github.com/menamerai/cheer-ekman.