Exploiting label semantics for rule-based activity recommendation in business process modeling. Issue 108 (September 2022)

• Record Type: Journal Article
• Authors: Sola, Diana; van der Aa, Han; Meilicke, Christian; Stuckenschmidt, Heiner
• Is Part Of: Information systems. Issue 108(2022)
• Publication Date: 2022-09
• Subjects: Process modeling -- Activity recommendation -- Rule learning -- Semantic analysis; Database management -- Periodicals; Electronic data processing -- Periodicals; Bases de données -- Gestion -- Périodiques; Informatique -- Périodiques; Database management; Electronic data processing; Periodicals; 005.7
• DOI: 10.1016/j.is.2022.102049
• Languages: English
• ISSNs: 0306-4379

Abstract

Abstract: Business process modeling is a crucial task in organizations. Yet, the creation of consistent and complete process models is challenging and necessitates the support of process modelers with their task. In previous work, we presented a rule-based activity-recommendation approach, which recommends appropriate labels for a new activity inserted by a modeler in a process model under development. While our method has shown to work well, it is limited by the fact that it only learns rules that describe the inter-relations between complete activity labels. In the case that the model's activities and the ones in the training repository are disjoint, the existing approach will thus not be able to provide any recommendations. In this paper, we overcome this restriction by additionally considering the natural language-based semantics of the process models. In particular, we propose a semantics-aware recommendation approach that extends the existing approach in both central phases, i.e., in the rule-learning phase and in the rule-application phase. We equip the rule learning with novel rule types, which capture action and business-object patterns in process models. For the rule application, we developed an optional similarity extension that allows rules to make recommendations even if the bodies of the rules are not exactly true for the given model. Through an evaluation on a large set of real-world process models, we demonstrate that the semantic extensions can improve the quality of recommendations. Highlights: A rule-based activity-recommendation approach to support business process modeling. Improved recommendations through the exploitation of natural language-based semantics. Learning and applying action and business-object patterns in the use of activity labels. Considering semantic similarity of actions and business objects. Experiments providing insights on the impact of different rule types.