Sentiment Analysis with ChatGPT Across Domains

Abstract

This paper explores ChatGPT’s applications in sentiment analysis across various  domains, including business, Reddit data, and scientific citations. It demonstrates  ChatGPT’s utility in understanding customer needs and public opinion, as well  as its ability to identify nuanced sentiment and biases in scholarly research  evaluation. Analysis of app reviews on the Google Play Store shows predominantly  positive sentiments, with models like Random Forest and SVM achieving high  effectiveness. The research evaluates ChatGPT and other large language models  such as Gemini and LLaMA for multilingual sentiment analysis, revealing their  proficiency alongside biases and inconsistencies across languages. The study  emphasizes the importance of standardized evaluation methodologies and the  need for data and algorithm improvements to enhance ChatGPT’s performance  and applicability in sentiment analysis.