Abstract: Emergency response process model describes the execution process of emergency response in a formal way.However,most emergency response process models are manually created by emergency experts,which is a time-consuming and laborious work.An approach to automatically extract emergency response process models from response process text of emergency plan was proposed to provide technical support for expert modeling.The text data of emergency plans were preprocessed and annotated for the subsequent extraction tasks.The response task and task relation were extracted by Bi-directional Long Short Term Memory Attention Conditional Random Fields (Bi-LSTM-Attention-CRF) network and XGBoost model respectively.Then,BPMN model of emergency response process was constructed based on the extracted response task and task relation.The real-world data was collected and experimental results imply that task element identification and task relations extraction are effective.In addition,a tool for extracting BPMN model of emergency response process was also presented.

Key words: process model extraction, emergency plan, business process model, relation extraction, named-entity recognition

摘要： 鉴于应急响应流程模型多由应急专家手动构建而费时费力的问题,提出一种从应急预案响应过程文本中自动抽取应急响应过程模型的方法,为专家建模提供技术支持。对应急预案文本数据进行预处理和标注,为后续训练抽取模型做数据准备;基于Bi-LSTM-Attention-CRF网络和XGBoost模型抽取响应任务和任务关系;基于抽取出的响应任务和任务关系构建应急响应过程BPMN模型。通过搜集真实的应急预案构建数据集,并对过程元素识别和任务关系抽取进行实验对比分析,证明了抽取方法的有效性,同时给出相应的实现工具。

关键词: 过程模型抽取, 应急预案, 业务流程模型, 关系抽取, 命名实体识别