CHATGPT THROUGH INDIAN LENSES: A SENTIMENT ANALYSIS OF INDIAN TWEETS ON CHATGPT

Abstract

The importance of using sentiment analysis in various aspects of study and research has been increasing significantly across the globe. By utilising the benefits of the same, an extensive study has been proposed on the diversity of sentiments of Indian users about ChatGPT through the data gathered from Twitter. The purpose of this study is to understand the perspectives of early users and identify the major topics of discussion regarding ChatGPT in India specifically. To achieve the same, 27,275 tweets has been segregated from pre-existing Kaggle dataset which encompasses several tweets containing the hashtag ‘ChatGPT’ by the twitter users across the world, covering from the launch date i.e. 30th November, 2022 till 24th February, 2023. In order to understand them, topic classification has been performed using Latent Dirichlet allocation (LDA). Sentiment analysis has been conducted using three techniques VADER, ROBERTA and TEXTBLOB, through which the dataset is labelled. Further, five different classifiers have been applied on the dataset named as: Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), XGBoost Classifier and Decision Tree, to get the performance parameters. According to the results, Logistic Regression classifier with GridSearchCV has the highest accuracy of 86.52% using Count Vectorization.