Extracting Cross-organization Emergency Response Process Models from Chinese Plans

Abstract

Emergency plans are used as effective instructions of hazard emergency response and they describe the overall emergency response process in natural language. In this paper, we propose an approach to extract a BPMN process model of cross-organization emergency response from plan text. It comprises three components: model elements identification, plan text decomposition and process model generation. First, a CRF (Conditional random field) network is combined with Bi-LSTM (a bidirectional long short-term memory) network (Bi-LSTMCRF) to identify model elements. Then, plan text is decomposed into subtexts about executive departments. Finally, inner-process models of all departments are generated from these subtexts and a complete collaborative process model is integrated by message flows. In addition, a case study is introduced and the precision of model elements identification is showed to illustrate that the proposed extraction approach is feasible and available.