Sentiment Analysis And UTAUT2 Classification On Maxim Application User Reviews Using IndoBERT And Zero-Shot

Abstract

The rapid growth of ride-hailing services has intensified competition, making user feedback on digital platforms a critical asset for service improvement. This study addresses the challenge of managing and extracting actionable insights from large volumes of unstructured user reviews on the Google Play Store for the Maxim application. To overcome this, a comprehensive text-mining framework is proposed, integrating sentiment analysis and technology acceptance modeling. A dataset of 2.000 Indonesian-language user reviews from July to September 2025 was retrieved via web scraping. Data preprocessing was executed using case folding, filtering, and normalization. Subsequently, sentiment classification was performed using the IndoBERT model, while the mapping of user text to the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) framework was automated using a Zero-Shot Classification approach. Finally, Structural Equation Modeling–Partial Least Squares (SEM-PLS) via SmartPLS 4.0 was utilized to test the structural hypotheses. The analytical findings reveal that negative sentiments slightly dominate the dataset (48.05%), heavily driven by system stability and sudden fare adjustments. Furthermore, the structural model proves that behavioral intention, effort expectancy, facilitating conditions, habit, performance expectancy, price value, and social influence exert positive and significant effects on adoption, whereas hedonic motivation exhibits no significant influence.