Detection User Needs: LDA-Based Analysis of Arabic Reviews for Governmental Mobile Applications

User reviews in app stores are considered very rich maintenance information texts that developers need to know. Many apps’ reviews contain requirements details such as bugs or problems, evaluation of user experience with some features, suggestions for improvements, and ideas for new features. The previous literature has illustrated different techniques and approaches to reduce the work needed to analyze and extract valuable content from mobile app reviews. However, no attention has been paid to analyzing and studying Arabic user reviews. This research explores user reviews of some Saudi governmental apps in the Google Play Store as a source of Arabic reviews dataset to aid software maintenance and improvement tasks in governmental applications. We adopted a seven-phase approach to analyze Arabic app reviews on the Google Play Store by applying natural language processing (NLP) techniques and Latent Dirichlet Allocation (LDA), which is one of the most used topic modeling algorithms to extract requirements issues from Arabic user feedback expressed in the app’s reviews. NLP techniques combined with the LDA model enable us to identify the types of requirements issues that users are complaining about. According to the finding results, governmental applications’ most frequent requirements issues are related to functional requirements issues (authentication and operational issues), after-update errors, nonfunctional requirements issues (usability issues), and user needs/user requirements. The proposed methodology can provide insight into the main requirements issues in governmental services apps to aid software engineering maintenance tasks.