Enhancing Sentiment Analysis with GPT—A Comparison of Large Language Models and Traditional Machine Learning Techniques

Abstract

Sentiment analysis is the process of extracting and analyzing opinions, attitudes, and emotions expressed in text data. Due to the increased amount of user-generated content on the Internet, sentiment analysis has become an important research area, which has successfully been tackled with machine learning models such as SVMs and Random Forests. Against the backdrop of rapidly growing popularity of Generative Pre-trained Transformers (GPT), the question arises, as to how good such models perform in sentiment analysis. This research paper analyzes and compares the performance of GPT-based models with traditional machine learning models for sentiment analysis. The results paint a clear picture: GPTs are a powerful concept that are applicable in sentiment analysis. In our study, they outperform traditional models such as SVM, Random Forests, or Naïve Bayes based on F1-score, precision, accuracy, and AUC-ROC score. As more advanced versions of GPT continue to be developed, it is likely that these models will become even more effective and popular sentiment analysis. Hence, the application and evaluation of GPTs represents a promising avenue for future research.