NLARE, a natural language processing tool for automatic requirements evaluation

Most research works have found that an important root cause of software project failure comes from the requirements; their quality has an important impact over other artifacts. As the requirements are expressed in natural language, they can be an important source of defects. Aspects such as non ambiguity, completeness, and atomicity can be affected due the characteristics of natural language. Traditional practices focus on finding software bugs, as a corrective approach, until the project has been coded already, instead assuring quality since the beginning. By other hand, evaluating such quality attributes can be a difficult task. In this paper we propose some guidelines for a disciplined sentence structure for expressing the requirements, which allows natural language processing techniques to evaluate quality. We also propose a tool for automatic requirement evaluation based on the grammar structure of sentences expressed in natural language. With this tool we have a huge speed increase over manual evaluation. In order to validate our proposal we have implemented a set of experiments with real projects, assessing the impact of requirements quality over project results.