Detecting Role Inconsistencies in Process Models

Abstract

Business process models capture crucial information about business operations. To overcome the challenge of maintaining process definitions in large process repositories, researchers have suggested methods to discover errors in the functional and the behavioral perspectives of process models. However, there is a gap in the literature on the detection of problems on the organizational perspective of process models, which is critical to manage the resources and the responsibilities within organizations. In this paper, we introduce an approach to automatically detect inconsistencies between activities and roles in process models. Our approach implements natural language processing techniques and enterprise semantics to identify ambiguous, redundant, and missing roles in textual descriptions. We applied our approach on the process model repository of a major telecommunication company. A quantitative evaluation of our approach with 282 real-life activities displayed that this approach can accurately discover role inconsistencies. Practitioners can achieve significant quality improvements in their process model repositories by applying the approach on process models complemented with textual descriptions.