Generating Functional Requirements Based on Classification of Mobile Application User Reviews

User reviews are important resources for mobile developers for maintaining and evolving mobile applications that have been released. Since there can be a lot of user reviews, it is cumbersome for the mobile development team to go through all user reviews to identify which ones contain useful information for further maintenance and evolution. This paper proposes an initial attempt to facilitate the maintenance and evolution process of a mobile development team by automating the generation of functional requirements from mobile application user reviews on the App Store and Play Store. Specifically, these functional requirements serve as change requirements for maintaining existing functions of the application or enhancing the application with new functions. The proposed approach is based on machine learning and natural language processing and consists of three steps. Firstly, user reviews that identify functional requirements are classified by text classification algorithms. Secondly, since some user reviews may address the same issues, distinct user reviews are identified by clustering techniques and text similarity. Finally, relevant information is extracted from the user reviews to generate functional requirements based on user review patterns and requirement boilerplates. In an evaluation, the generated functional requirements obtained moderate to high scores in terms of readability, unambiguity, completeness, and validity. The proposed approach can help the development team identify functional requirements from direct feedback of the users which should be considered and further refined in the maintenance and evolution of the mobile application.