Leveraging Unstructured Data to Analyze Implicit Process Context

Abstract

Adapting a business process to different context requires identifying various situations and evolving the process to support such situations. Previous work focused on modeling, observing and collecting contextual information. Furthermore, impact of context on process or resource performance has been studied. However, much of the work considers explicit contextual information that is defined by domain experts. There are several implicit contextual dimensions, that are difficult to model as all situations cannot be anticipated a priori. Context mining involves analysis of process logs to identify context and correlate with process performance indicators or outcomes. In this work, we leverage unstructured data available in user comments or mails to discover implicit context of the process. We automatically analyze textual data and group process instances by applying information extraction and text clustering techniques. Groups of process instances are correlated to their process outcomes to filter irrelevant information. We apply the approach on real-world process logs to identify contextual information.