Ambiguity Measures in Requirements Engineering

摘要

In many software system development projects, the documents available for requirement analysis are couched in natural language. However, their quality is not always optimal, owing to the presence of ambiguous expressions open to different interpretations. This may be an obstacle against understanding of the problem and against the identification and subsequent modelling of requirements. It may, moreover, hamper the development process as a whole and be deleterious to the quality of the system. This paper proposes a scheme with which to define indices of ambiguity, both structural and semantic. It also investigates the feasibility of calculating these indices using the knowledge base of a natural language processing system and the outputs produced during the text parsing phase. It may thus be possible rapidly to signal the terms and phrases open to several interpretations and thereby improve the quality of the requirements analysis process.