匾烈右棕邑.PDFVIP

Validating Inconsistent Requirements Mo dels using Graph-based Ab duction TIM MENZIES University of British Columbia, Canada STEVE EASTERBROOK University of To ronto, Canada BASHAR NUSEIBEH Imp erial College of Science Technology and Medicine, UK and SAM WA UGH Defence Science and Technology Organisation Air Op erations Division, Australia Multiple viewp oints are often used in requirements engineering to facilitate traceability to stake- holders, to structure the requirements pro cess, and to provide richer mo deling by incorp orating multiple con icting descriptions. Toleration of inconsistency is a key advantage. However, in- consistency intro duces considerable extra complexity when reasoning ab out requirements. This complexity may limit our ability to maintain a lazy approach to consistency management when we want to validate requirements mo dels. In this pap er, we describ e a series of exp eriments with graph-based ab duction for multiple world reasoning over inconsistent requirements mo dels. Our ab ductive algorithm, HT4, sorts an inconsistent mo del into a numb er of consistent worlds that supp ort answers to sp ecic queries ab out the mo del. This approach avoids the trivialization that o ccurs in classical deductive inference in the presence of inconsistency. Exp eriments with this approach reveal that surprisingly few worlds are generated. This result is robust over a range of dierent mo dels, dierent amounts of data available from the domain, and dierent mo deling primitives for representing time. To explore the implications of this result, we develop ed a second algorithm, HT0, that extracts a single world at random from an inconsistent mo del. Exp erimen- tally, HT0 runs fast even for very large mo dels, and supp orts most of the queries a