一种面向突发事件应急要素的自动识别方法研究

An Automatic Recognition Method for Emergency Response Elements

突发事件影响广泛,涉及因素较多,当前突发事件要素识别研究存在灾害要素信息、应对流程信息、承灾载体信息等数据自动识别困难等问题。文章构建突发事件应急要素“四维”模型,确定包含保障措施、响应措施等流程性信息在内的八类应急要素类别,提出TDA-RoBERTawwm-BiLSTM-CRF模型。该模型采用基于模板的数据增强策略,融合预训练模型与双向长短期记忆网络-条件随机场方法,实现了突发事件多要素信息的自动识别。对比与消融实验结果表明,该模型在突发事件应急要素自动识别中能起到良好效果。

Current research on the identification of emergency response elements faces the challenge of automatically recognizing data on disaster factors,response process,affected objects and so on.This paper establishes a four-dimensional model of emergency response elements,which are classified into eight categories covering safeguard measures,response measures and other processes,and proposes a TDA-RoBERTa-wwm-BiLSTM-CRF model.By adopting a template-based data augmentation strategy and integrating a pre-trained model with BiLSTM-CRF,the proposed model automatically identifies multiple factors of emergency events.The comparative and ablation study shows that the proposed model performs effectively in the automatic recognition of emergency response elements.