Sentiment Analysis of ChatGPT Tweets Using Transformer Algorithms

Abstract

This study explores the application of the Transformer model in sentiment analysis of tweets generated by ChatGPT. We used a Kaggle dataset consisting of 217,623 instances labeled as "Good", "Bad", and "Neutral". The Transformer model demonstrated high accuracy (90%) in classifying sentiments, particularly predicting "Bad" tweets. However, it showed slightly lower performance for the "Good" and "Neutral" categories, indicating areas for future research and model refinement. Our findings contribute to the growing body of evidence supporting deep learning methods in sentiment analysis and underscore the potential of AI models like Transformers in handling complex natural language processing tasks. This study broadens the scope for AI applications in social media sentiment analysis.