Abstract
Mobile application (app) reviews are feedback about experiences, requirements, and issues raised after users have used the app. The iteration of an app is driven by bug reports and user requirements analyzed and extracted from app reviews, which is a problem that app designers and developers are committed to solving. However, a great number of app reviews vary in quality and reliability. It is a difficult and time-consuming challenge to analyze app reviews using manual methods. To address this, a novel approach is proposed as an automated method to predict high priority user requests with fourteen extracted features. A semi-automated approach is applied to annotate requirements with high or low priority with the help of app changelogs. Reviews from six apps were retrieved from the Apple App Store to evaluate the feasibility of the approach and interpret the principles. The performance comparison results of the approach greatly exceed the IDEA method, with an average precision of 75.4% and recall of 70.4%. Our approach can be applied to specific app development to assist app developers in quickly locating user requirements and implement app maintenance and evolution.
Highlights
The growth of the app market has been boosted by the maturity of smartphones, allowing users to conveniently browse and download apps from app stores (e.g., AppleApp Store and Google Play) and leave reviews for apps they have used, including star ratings and text-based feedback
We focus our attention on the user requirements in reviews and propose a novel approach to (1) extract requirement phrases for the app functionalities; (2) calculate features such as the occurrence frequency and rating for the requirement phrases; (3)
We focus on the following research questions (RQ):
Summary
The growth of the app market has been boosted by the maturity of smartphones, allowing users to conveniently browse and download apps from app stores (e.g., AppleApp Store and Google Play) and leave reviews for apps they have used, including star ratings and text-based feedback. Many studies have proven that app reviews, which contain problem feedback, feature requests, and other suggestions, can be regarded as references for the iterative design and development of the app [1,2,3]. By analyzing user data from the new versions, iteration strategies can be conducted by app designers and developers. The language expression of reviews is relatively informal, containing lots of noise, such as misspellings, casual grammatical structures, and non-English words [8]. To address these issues, there are many studies dedicated to automatically filtering out non-informative reviews [9], categorizing reviews
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
