Extracting Software Functional Requirements from Free Text Documents

The acquisition of requirements assets are important in software product line (SPL) engineering for it help enhancing the effectiveness of reuse. Traditional methods are heavily based on manual effort. This appears to be a barrier for many organizations which tend to launch a SPL. In this paper, we propose an approach to extract functional requirements by analyzing text-based software requirements specifications (SRSs). We analyze the linguistic characterization of SRSs. According to it we define extended functional requirements framework (EFRF) which consists of 10 semantic cases, then we generate converting rules. We introduce an NLP (natural language process) approach to build EFRFs from documents based on the concept of EFRF and the converting rules. The extracted EFRFs are suitable for expression and modeling of functional requirements variability. We apply our method to an auto-marker software product line. The result shows the approach has high accuracy and efficiency, and the approach is readily scalable and extensible.