The existence of unstructured information that describes processes represents a challenge in organizations, mainly because this data cannot be directly referred into process-aware ecosystems due to ambiguities. Still, this information is important, since it encompasses aspects of a process that are left out when formalizing it on a particular modelling notation.
This paper picks up this challenge and faces the problem of ambiguities by acknowledging its existence and mitigating it. Specifically, we propose a framework to partially automate the elicitation of a formal representation of a textual process description, via text annotation techniques on top of natural language processing. The result is the ATDP language, whose syntax and semantics are described in this paper. ATDP allows to explicitly cope with several interpretations of the same textual description of a process model. Moreover, we link the ATDP language to a formal reasoning engine and show several use cases. A prototype tool enabling the complete methodology has been implemented, and several examples using the tool are provided.
Published in Proceedings of BPM 2019; Vienna, Austria; September 2019.