Sentiment Analysis in Software Engineering: Evaluating Generative Pre-trained Transformers

TL;DR

This paper evaluates the performance of generative pre-trained transformers like GPT-4o-mini against traditional models like BERT for sentiment analysis in software engineering, highlighting the impact of dataset complexity and fine-tuning.

Contribution

It provides a systematic benchmark of GPT-4o-mini and BERT in SE sentiment analysis, revealing conditions where generative models outperform or match traditional models.

Findings

Fine-tuned GPT-4o-mini achieves high F1-scores on structured datasets.

Default GPT-4o-mini generalizes better on complex, imbalanced datasets.

Trade-offs exist between fine-tuning and using pre-trained models for SE sentiment analysis.

Abstract

Sentiment analysis plays a crucial role in understanding developer interactions, issue resolutions, and project dynamics within software engineering (SE). While traditional SE-specific sentiment analysis tools have made significant strides, they often fail to account for the nuanced and context-dependent language inherent to the domain. This study systematically evaluates the performance of bidirectional transformers, such as BERT, against generative pre-trained transformers, specifically GPT-4o-mini, in SE sentiment analysis. Using datasets from GitHub, Stack Overflow, and Jira, we benchmark the models' capabilities with fine-tuned and default configurations. The results reveal that fine-tuned GPT-4o-mini performs comparable to BERT and other bidirectional models on structured and balanced datasets like GitHub and Jira, achieving macro-averaged F1-scores of 0.93 and 0.98, respectively. However, on linguistically complex datasets with imbalanced sentiment distributions, such as Stack Overflow, the default GPT-4o-mini model exhibits superior generalization, achieving an accuracy of 85.3\% compared to the fine-tuned model's 13.1\%. These findings highlight the trade-offs between fine-tuning and leveraging pre-trained models for SE tasks. The study underscores the importance of aligning model architectures with dataset characteristics to optimize performance and proposes directions for future research in refining sentiment analysis tools tailored to the SE domain.