Detecting Ambiguities in Requirements Documents Using Inspections

Natural language is the most used representation for stating requirements on computer-based systems in industry. However, natural language is inherently ambiguous. Customers and software developers can disagree on the interpretation of a requirement without being aware of this fact. This disagreement can result in disastrous software failures. We argue that ambiguity is a more complex phenomenon than is often recognized in the literature. While inconsistencies and some types of incompleteness can be mechanically detected in formal specifications, ambiguities in informal specifications often result in incorrect formal specifications. These misinterpretations can be detected only by execution or inspection of the formal specification. We suggest an inspection technique for detecting ambiguities in informal requirement documents before formal specifications are produced in order to avoid misinterpretations, rather than searching for them afterwards. We show how this technique can be tailored to different project contexts with the help of metamodels to increase its effectiveness. Finally, we report on experimental validation of the technique.