Multimodal Process Mining

Process modeling has a long and successful history, greatly aided by process discovery from event logs. However, many processes leave limited digital traces. Often, manual and physical activities go unrecorded in these logs, leading to significant gaps in the data. To address these blind spots, we must manually model certain activities, which disconnects the conceptual model from actual execution traces. This paper explores how to better track and understand manual business activities that are often poorly documented digitally. We introduce a new method that converts video into detailed event logs, enhancing the way businesses monitor and improve their processes. Our research includes analysis of the Multimodal Process Mining concept, and a Vid2log tool, designed to improve the way manual activities are recorded and analyzed.