Text2EL: Exploiting Unstructured Text for Event Log Enrichment

Abstract

Process mining provides a range of methods and techniques to analyse business processes through information stored in so-called event logs. The richer these event logs and the higher quality they are, the more insights we can obtain. Till now, information in the form of unstructured text, e.g. notes, comments, reviews, and posts, is not fully and systematically exploited for the purposes of log enrichment. In this paper, we introduce Text2EL, a two-phase event log enrichment approach based on unstructured text. In Phase 1, events, case attributes, and event attributes are extracted from unstructured text associated with organisational processes. In Phase 2, the extracted events and attributes are semantically and contextually validated before enriching the event log. Our approach applies techniques from natural language processing, sentence embeddings, and contextual and expression validation. We evaluated the completeness, concordance, and correctness of an enriched event log through experiments with a real-life healthcare data set. The experiments showed the feasibility and applicability of our approach.