Automated requirement sentences extraction from software requirement specification document

Abstract

In the requirement reuse and natural language document-based Software Product Line (SPL) domain analysis, requirement sentences of the requirement document are the primary concern. Most studies conducted in this research area have document preprocessing stage in their methods that is a manual process to separate requirement sentences and non-requirement sentences from the document. This manual labor process might be tedious and error-prone since it will need much time and expert intervention to make this process completely done. In this paper, we present a method to automate requirement sentence extraction from the Software Requirement Specification (SRS) document by leveraging Natural Language Processing (NLP) approach and requirement boilerplate sentence patterns. Conducted experiments in this research show this method has such accuracy from 64% to 100% on precision value and recall value in the range of 64% to 89%.