Unraveling Emotions with Pre-Trained Models

2025-10-22

AI Summary

Open-domain sentiment recognition faces challenges including contextual ambiguity, linguistic variation, and difficulty modeling complex, implicit, or mixed emotions. Method: This paper systematically compares fine-tuning versus prompt engineering on large language models (LLMs), proposing a structured prompt design—incorporating role specification and stepwise reasoning—and a fine-grained sentiment category aggregation strategy that groups semantically similar emotions into higher-order dimensions. Experiments are conducted across multiple scenarios using pretrained Transformer models. Contribution/Results: Structured prompting and semantic emotion grouping significantly enhance LLMs’ ability to discern implicit and compound sentiments. The optimized prompt-based approach achieves >70% accuracy without fine-tuning—approaching fine-tuned model performance—while demonstrating superior generalizability. Key findings: (1) principled emotion grouping mitigates label sparsity; (2) structured prompts effectively elicit sentiment reasoning chains in LLMs. This work provides a reproducible, lightweight adaptation methodology for LLMs in low-resource, high-ambiguity sentiment analysis tasks.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Prompt Engineering / Prompting
• Humans and AI: Emotional Intelligence

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Transformer models have significantly advanced the field of emotion recognition. However, there are still open challenges when exploring open-ended queries for Large Language Models (LLMs). Although current models offer good results, automatic emotion analysis in open texts presents significant challenges, such as contextual ambiguity, linguistic variability, and difficulty interpreting complex emotional expressions. These limitations make the direct application of generalist models difficult. Accordingly, this work compares the effectiveness of fine-tuning and prompt engineering in emotion detection in three distinct scenarios: (i) performance of fine-tuned pre-trained models and general-purpose LLMs using simple prompts; (ii) effectiveness of different emotion prompt designs with LLMs; and (iii) impact of emotion grouping techniques on these models. Experimental tests attain metrics above 70% with a fine-tuned pre-trained model for emotion recognition. Moreover, the findings highlight that LLMs require structured prompt engineering and emotion grouping to enhance their performance. These advancements improve sentiment analysis, human-computer interaction, and understanding of user behavior across various domains.

Problem

• Addresses emotion recognition challenges in open-ended text analysis
• Compares fine-tuning versus prompt engineering for emotion detection
• Evaluates emotion grouping techniques for large language models

Innovation

• Fine-tuned pre-trained models for emotion recognition
• Structured prompt engineering with Large Language Models
• Emotion grouping techniques to enhance model performance