Analysis of ChatGPT Application Service Quality Using Text Classification and Topic Modeling

This research aims to understand the sentiments and main topics of users related to the service quality of the ChatGPT Android application by analyzing User-Generated Content (UGC) on the Google Play Store based on user experiences.The research method used in this study is sentiment analysis with the RoBERTa algorithm and topic modeling using LDA.After comparing the Naive Bayes, SVM, and RoBERTa algorithms, RoBERTa was found to have the highest accuracy, with 72% for dimension classification and 94% for sentiment classification.Positive sentiment was 86.17%, and negative sentiment was 13.83%.The dimensions used to measure service quality are Content Quality, Engagement, Reliability, Usability, and Privacy.The results of the study show that the service quality of the ChatGPT Android application is mostly positive in the dimensions of Engagement, Reliability, and Content Quality.However, the dimensions of Privacy and Usability have negative views.Positive sentiment findings include satisfaction with the information provided, the dark mode feature, the user-friendly interface, user engagement, and application reliability.On the other hand, negative sentiment highlights issues with privacy and usability.This knowledge can be used to improve service quality, especially in the dimensions of Privacy and Usability.