Topic Modeling and EDA for Analyzing User Sentiments for ChatGPT

This paper uses Topic Modeling and Exploratory Data Analysis (EDA) with Twitter data to investigate user emotions on ChatGPT. We find recurring themes and patterns in user thoughts by analyzing millions of tweets. While EDA investigates temporal trends, user demographics, and general sentiment landscapes, we leverage topic modeling, namely Latent Dirichlet Allocation (LDA), to discover important discussion topics. Through the integration of Topic Modeling and EDA, the research offers a comprehensive perspective on user sentiments and insights into the perceptions of ChatGPT on Twitter. Sentiment analysis, performed using VADER and Twitter-roBERTa, shows that 65% of tweets were positive, 25% neutral, and 10% negative. VADER and roBERTa differ significantly in sentiment classification, with roBERTa identifying 12% more negative sentiments. This integrated approach provides a comprehensive view of user sentiments and highlights areas of concern, such as potential misuse and reliability. The findings contribute to the ongoing discourse on AI usage and suggest the need for further research across different platforms and over extended periods to capture evolving sentiments and inform the development of AI technologies.