Text-based requirements preprocessing using nature language processing techniques

In a distributed environment, non-technical stakeholders are required to write down requirement statements by themselves. Nature language is the first choice for them. In order to alleviate the burden of reading free-text requirement documents by requirements engineers, we extract goals and relevant stakeholders from requirement statements automatically by a computer-assisted way. In this paper, requirements are divided into system level requirements and instance level requirements. Methods are proposed to solve two types of requirements by analyzing the characteristics of requirement expressions, and combining techniques of nature language processing with semantic web. Semantic-enhanced segment and domain sentence pattern are two novel techniques utilized in our methods. Our approach accelerates goal extraction from text-based requirements and alleviates the burden of requirements engineers significantly.