A PERFORMANCE ANALYSIS OF SUGGESTION MINING IN MOBILE APP REVIEWS USING LARGE LANGUAGE MODELS TRANSFER LEARNING TECHNIQUES
Makarand Lotan Mali, Nitin N. Patil · Proceedings on Engineering Sciences · 2025
One of the enduring issues with obtaining user feedback is to find effective ways of actionable improvement recommendations from this feedback for quality mobile app development. To address this gap, we introduce two textual syntactical approaches using NLP. The solution is a highly optimized version of the BERT, DistilBERT Large Language Models (LLMs). Both models are expected to efficiently learn decisions regarding potential modifications to the mobile application We incorporate domain action verbs and phrases of the language of the specialized field. These specific terms enable the models to successfully identify the user intent when providing change preferences regarding the application. We assess the effectiveness of the above strategies by carrying out an extensive study of 8,061 mobile app reviews, we used these strategies to get insight into the common user experiences that provide direction to the mobile app development team, hence it indirectly enhancing the quality of the mobile apps. This could adopt a more user-centric approach within the application, increasing user satisfaction. Further investigation of the collected suggestions will enhance sorting and prioritization of user problems, guiding development efforts to address biases effectively.