emotion modeling

Beyond Sentiment Classification: A Generative Framework for Emotion Intensity Evaluation in Text

This work introduces a novel approach to emotion modeling, shifting from discrete classification to continuous emotion intensity evaluation in text. The authors constructed a dataset of emotional intensity scores and fine-tuned generative language models to output continuous values from 0-100, outperforming classification baselines and demonstrating generalization capabilities.