Framework for Analyzing App Reviews in Software Development

This technical report presents a novel framework designed to enhance the technology value stream in software development by leveraging user feedback from app reviews. The framework employs various Natural Language Processing (NLP) techniques to systematically analyze reviews from mobile applications, enabling developers to identify critical areas for improvement. The authors detail the methodology used to create this framework, which includes literature studies, topic analysis techniques, and sentiment analysis. The report outlines the challenges faced in extracting meaningful insights from the vast volume of unstructured review data and proposes solutions to automate the continuous maintenance process within the software development pipeline. An empirical evaluation of the framework is conducted using reviews from sixteen popular Android applications, providing evidence of its effectiveness. The contributions of this research include a key phrase extraction model, a zero-shot classifier for topic recognition, and a comprehensive recommendation framework that integrates sentiment analysis to address user concerns effectively.