Reviewing Natural Language Requirements with Requirements Smells

The quality of requirements artifacts, such as software requirements specifications, is crucial for the success of a software development project, because the later a defect is found the more expensive it is to fix. However, as virtually all requirements are still written in natural language, and requirements artifacts grow often large, they are very hard to review for quality due to the imprecise nature of natural language. In contrast, it is easier to not find quality, but symptoms of violations of quality, because they often leave concrete trace in the artifacts. For example, passive sentences in requirements are said to make testing harder as they can potentially hide the actor. Here it is easier to find the symptom of violation of testability, i.e. a passive sentence, than to prove that the requirement is “easily testable”. This is a concept well known for code quality as code (bad) smells, which has been proposed by Fowler and Beck. We suggest introducing the smell concept to requirements engineering in order to find possible violations of requirements quality. Consequently, a requirements (bad) smell is a concrete symptom for a requirement artifact’s quality defect in the usage context of a certain activity. The proposed research aims at understanding whether smells can help reviewing natural language requirement artifacts by pointing out to symptoms for potential quality defects in order to improve quality reviews of requirements.