Automatic Validation and Correction of Formalized, Textual Requirements

Nowadays requirements are mostly specified in unrestricted natural language so that each stakeholder understands them. To ensure high quality and to avoid misunderstandings, the requirements have to be validated. Because of the ambiguity of natural language and the resulting absence of an automatic mechanism, this has to be done manually. Such manual validation techniques are timeconsuming, error-prone, and repetitive because hundreds or thousands of requirements must be checked. With an automatic validation the requirements engineering process can be faster and can produce requirements of higher quality. To realize an automatism, we propose a controlled natural language (CNL) for the documentation of requirements. On basis of the CNL, a concept for an automatic requirements validation is developed for the identification of inconsistencies and incomplete requirements. Additionally, automated correction operations for such defective requirements are presented. The approach improves the quality of the requirements and therefore the quality of the whole development process.