Generating Mobile App Improvement Recommendations Through User Review Analysis Using BERTopic and GPT

Mobile applications often receive a vast number of user reviews, making it challenging for developers to efficiently analyze feedback and improve their apps. This paper proposes a novel framework for generating mobile app improvement recommendations by analyzing user reviews using BERTopic and GPT. Our approach integrates these advanced NLP techniques, employing BERTopic for contextually aware topic modeling of user reviews, followed by GPT-4 for identifying app issue-related topics through sentiment analysis and relevance detection. GPT-4 then groups these topics into main problem areas and generates improvement recommendations. We demonstrate the framework's effectiveness through a case study on LinkAja, an Indonesian e-wallet application, analyzing 5,627 user reviews. The results showcase the framework's ability to extract meaningful insights and generate relevant recommendations, as validated through interviews with professionals from LinkAja. This research contributes to the field of app development and user feedback analysis by providing an efficient, scalable tool for extracting insights from user reviews, potentially leading to more targeted app improvements and enhanced user satisfaction in the competitive mobile app market.