Model-Based Testing from Controlled Natural Language Requirements

Abstract

Model-Based Testing (MBT) techniques usually take as input models that are not available in the very beginning of a development. Therefore, its use is postponed. In this work we present an approach to MBT that takes as input requirements described in a Controlled Natural Language. Initially, the requirements are syntactically analyzed according to a domain specific language for describing system requirements, and their informal semantics is depicted based on the Case Grammar theory. Then, the requirements semantics is automatically represented as a Transition Relation, which provides formal basis for MBT, and test cases are generated with the support of a solver. Our approach was evaluated considering four examples from different domains. Within seconds, our approach generated 94 % of the test vectors manually written by specialists. Moreover, considering a mutant-based strength analysis, our approach yielded a mutation score between 54 % and 98 %. © Springer International Publishing Switzerland 2014.