Which Feature is Unusable? Detecting Usability and User Experience Issues from User Reviews

Usability and user experience (UUX) strongly affect software quality and success. User reviews allow software users to report UUX issues. However, this information can be difficult to access due to the varying quality of the reviews, its large numbers and unstructured nature. In this work we propose an approach to automatically detect the UUX strengths and issues of software features according to user reviews. We use a collocation algorithm for extracting the features, lexical sentiment analysis for uncovering users' satisfaction about a particular feature and machine learning for detecting the specific UUX issues affecting the software application. Additionally, we present two visualizations of the results. An initial evaluation of the approach against human judgement obtained mixed results.