Abstract. Online reviews that manifest user feedback have become an available resource for eliciting requirements to design future releases. However, due to complex and diverse opinion expressions, it is challenging to utilize automated analysis for deriving constructive feedback from these reviews. What's more, determining important changes in requirements based on user feedback is also challenging. To address these two problems, this paper proposes a systematic approach for transforming online reviews to evolutionary requirements. According to the characteristics of reviews, we first adapt opinion mining techniques to automatically extract opinion expressions about common software features. To provide meaningful feedback, we then present an optimized method of clustering opinion expressions in terms of a macro network topology. Based on this feedback, we finally combine user satisfaction analysis with the inherent economic attributes associated with the software's revenue to determine evolutionary requirements. Experimental results show that our approach achieves good performance for obtaining constructive feedback even with large amounts of review data, and furthermore discovers the evolutionary requirements that tend to be ignored by developers from a technology perspective.

No abstract