Object Synchronizations and Specializations with Silent Objects in Object-Centric Petri Nets

Processes involve interacting objects of different types, such as orders, items and machines. Object-centric event logs capture the execution of activities with the involved objects in such processes. Discovery algorithms use these logs to construct process models in standardized formalisms, such as object-centric Petri nets. Conformance checking techniques quantify how fittingly and precisely these models represent a log. Existing techniques for object-centric Petri nets only consider explicitly recorded objects. However, unobserved objects might still influence the behaviour of the observed objects in a log. Observed objects may belong to the same unobserved group and therefore remain synchronized across multiple activities. Additionally, observed objects might have an unobserved property and hence specialize on certain activities. In this paper, we propose the notion of silent objects in object-centric Petri nets, which are unobserved objects that nonetheless affect observed objects. We provide a novel algorithm to automatically detect silent objects that cause synchronizations and specializations in object-centric event logs. We show that the inclusion of silent objects in the discovery of object-centric Petri nets preserves fitness and increases precision for logs that include them. We apply our method to multiple public logs and consistently find silent objects to be present in them. Additionally, we observe our approach’s runtime to remain feasible for large input logs.