Enhancing Mobile App Quality: A Data-Driven Approach to User Feedback

Abstract

In the very competitive mobile application arena, customer reviews are essential for success. Timely responses to these reviews can substantially enhance an app’s rating and prominence. The proliferation of user-generated content has rendered the extraction of useful insights increasingly challenging for developers. To enhance user experience and enable prompt changes, it is essential to swiftly and properly identify the primary issues encountered by users. This research presents a dual-phase hybrid framework. Phase 1 computes the mean sentiment rating for user reviews of Zoom Cloud Meeting, Microsoft Teams, and Google Meet. According to these averages, Phase 2 offers pragmatic recommendations to developers. A text data augmentation strategy utilizing the advanced language comprehension capabilities of big language models such as ChatGPT was implemented. The framework uses a Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) to calculate ratings from review datasets and BERT-base for labeling and analyzing feelings. The suggested hybrid model exhibited robust performance on datasets containing a substantial number of reviews, suggesting a tendency for positive ratings to surpass negative ones. The study utilized a benchmark dataset comprising 44,767 app reviews to compare its findings with actual ratings, yielding significant insights for enhancing app development. According to average evaluations, Microsoft Teams (4.52) and Zoom Cloud Meeting (4.16) outperformed Google Meet (3.12) and all other applications (3.40), indicating that users of Teams and Zoom will derive the most advantages from the latest versions. We anticipate that the study’s recommendations will aid app makers in enhancing their offerings.