Sentiment Analysis and Emotion Detection of ChatGPT Users on Google Play Store Using the Naïve Bayes Algorithm

Abstract

With the rapid adoption of artificial intelligence applications such as ChatGPT, understanding both users' sentiments and emotional expressions has become increasingly important for evaluating user experience and improving service quality. This study aims to analyze user sentiment and identify emotional tendencies expressed in Google Play Store reviews of the ChatGPT application. Methods: A total of 10,000 user reviews were collected through web scraping from the Google Play Store. The collected data underwent several preprocessing stages, including data cleaning, case folding, tokenization, normalization, stopword removal, and stemming. Sentiment analysis was performed using the Multinomial Naïve Bayes algorithm to classify reviews into positive, neutral, and negative categories. Emotion detection was subsequently conducted by interpreting dominant lexical patterns and frequently occurring terms within each sentiment category through Word Cloud visualization, enabling the identification of users' emotional tendencies rather than direct emotion classification. Model performance was evaluated using accuracy, precision, recall, and F1-score. Results: The sentiment distribution showed that 84.7% of the reviews were classified as positive, 3.9% as neutral, and 11.4% as negative. The best-performing model, using an 80:20 train–test split, achieved an accuracy of 86.8%, precision of 83.8%, recall of 86.8%, and an F1-score of 82.0%. The emotion detection results revealed that positive reviews were dominated by words reflecting satisfaction, usefulness, and productivity, whereas negative reviews mainly expressed frustration related to subscription costs, system errors, and application performance. Novelty: Unlike previous studies that primarily focused on sentiment classification, this research combines machine learning-based sentiment analysis with lexical interpretation of emotional tendencies derived from user-generated reviews. This integrated approach provides a more comprehensive understanding of users' perceptions and emotional responses toward ChatGPT, offering practical insights for improving AI-based application services.