Requirement Acquisition from Social Q&A Sites

Abstract

Social Q&A sites have changed the way of knowledge sharing in software communities. Comparing to the traditional mail-list, bug/change repositories, software forums and software marketplaces, users and developers are more active in social Q&A sites, and social Q&A sites are more open and free. The feedbacks from users have much potential valuable information, such as feature requests, bugs or sentiment, but there also exists lots of noise especially for social Q&A sites. How to mine the useful information from the feedbacks in social Q&A sites has become a problem. This paper focuses on the feature requirements in requirement acquisition, which can be used to assist software development. We propose an effective approach, which combines Support Vector Machine (SVM) with requirement dictionary to find the questions about feature requests from the posts in social Q&A sites. We evaluate the approach on available dataset, and compare it to the other different approaches. The results show that the automatically requirement acquisition through improved SVM approach is useful and can significantly decreases the manual effort.