Automatic Requirements Elicitation from Social Media (ARESM)

Abstract

After software applications are installed and deployed, the software applications enter to maintenance phase. Changes during maintenance phase often occur due to software faults and user new requests. Both software faults and user new requests are user feedbacks that can be normally obtained by interviewing, distributing questionnaires, or setting up JAD meeting. Nowadays, some user feedbacks are scattered online on App Store, Play Store, and social media and normal elicitation methods are not suitable for collecting online user feedbacks. Although some online user feedback elicitation methods are used to collect App Store's reviews and Play Store's reviews, these methods are not designed for eliciting user feedbacks from some social media. Social media contains both positive and negative user feedbacks which are important for software improvement. Positive user feedbacks are tentative new software requirements. Negative user feedbacks are suggestions for fixing existing features. Lacking adequate methods on how to elicit these requirements may eventually result to inefficient software evolution. This paper proposes an approach to extract requirements automatically from user feedbacks on social media and classify user feedbacks to requirements and non-requirements using Naive Bayes's Machine Learning. The approach is evaluated with an example. A preliminary result is present.