Verification of Quantitative Temporal Compliance Requirements in Process Descriptions Over Event Logs

Process compliance verification ensures that processes adhere to a set of given regulatory requirements which are typically assumed to be available in a formalized way using, e.g., LTL. However, formalized requirements are rarely available in practice, but rather embedded in regulatory documents such as the GDPR, requiring extraction and formalization by experts. Due to the vast amount and frequent changes in regulatory documents, it is almost impossible to keep formalized requirements up to date in a manual way. Therefore, this paper presents an approach towards compliance verification between natural language text and event logs without the need for requirements formalization. This enables humans to cope with an increasingly complex environment. The approach focuses on quantitative temporal requirements (QTCR) and consists of multiple steps. First, we identify clauses with temporal expressions from process descriptions. Second, we generate a set of QTCR by mapping the retrieved clauses to event log activities. Finally, in the third step, we verify that the event log is compliant with the QTCR. The approach is evaluated based on process descriptions and synthesized event logs. For the latter, we implement time shifting as a concept for simulating real-life logs with varying temporal challenges.