EmoLASP: Emotion Recognition with Language Models and Answer Set Programming

Researchers propose EmoLASP, a framework combining language models with Answer Set Programming to predict VAD scores in conversations, improving performance over using language models alone on the IEMOCAP benchmark across six open-source LLMs (3B-120B) and two PLMs (BERT, RoBERTa). The gains are largest for prompt-only LLMs without fine-tuning, while fine-tuned PLMs benefit little once dialogue history is available.

arXiv:2608.29035v1 Announce Type: new Abstract: Emotion recognition in conversations is increasingly tackled with language models, but these models can be unstable and expensive to fine-tune or to prompt with long dialogue histories. We propose EmoLASP, a framework that combines a language model with declarative reasoning via Answer Set Programming ASP to predict VAD scores Valence-Arousal-Dominance in conversations. Experiments on a widely used benchmark dataset IEMOCAP across six open-source LLMs 3B-120B and two PLMs BERT, RoBERTa show that EmoLASP improves prediction performance compared to using the language model alone, even when the LLMs/PLMs are given no dialogue history in their prompts or input vectors. The gains are largest for prompt-only LLMs, which EmoLASP uses without any fine-tuning. However, for fine-tuned PLMs, the reasoner adds little once dialogue history is available. EmoLASP's LLM pipeline demonstrates the potential advantages of using a reasoning approach to ensure emotion prediction consistency and to reduce both the cost of fine-tuning and the cost of prompting with long dialogue histories.