安全、健康和环境2026，Vol.26Issue(4)：65-73,9.DOI:10.3969/j.issn.1672-7932.2026.04.010
基于EG-UIE的危化品事故处置措施实体识别方法研究
Research on Entity Recognition Method for Hazardous Chemical Accident Disposal Measures Based on EG-UIE
摘要
Abstract
The emergency response to hazardous chemical accidents heavily relies on the rapid matching of historical cases with disposal strategies.However,existing accident reports and disposal plans are mostly un-structured texts,making precise and efficient strategy retrieval difficult in emergency situations,which restricts the speed of emergency response.The root cause lies in the cross-dimensional and multi-entity nested character-istics of hazardous chemical incident information,which prevent traditional methods from accurately identifying and mapping key disposal elements.To address these issues,a disposal measure entity recognition model was constructed based on the Universal Information Extraction(UIE)framework,integrating the Enhanced Lan-guage Representation with Informative Entities(ERNIE)pre-trained model and a Bidirectional Gated Recurrent Unit(BiGRU),effectively achieving cross-dimensional entity relationship mapping and nested entity recogni-tion.The model was trained and tested on a dataset comprising 1,074 entries from the Material Safety Data Sheet(MSDS)database and historical accident reports,achieving an overall precision(P),recall(R)and F1-score of 96.82%,96.98%,96.9%,respectively.The accuracy of this model demonstrated a distinct improve-ment compared to traditional models.This study applies entity extraction technology to the emergency response phase of hazardous chemical safety accidents,providing a foundation for the generation of accident disposal plans and emergency decision support,and offering crucial technical support for the future development of intel-ligent emergency decision support systems.关键词
危化品事故处置/实体识别/文心预训练模型/双向门控循环神经网络/通用信息抽取Key words
hazardous chemical accident disposal/entity recognition/Enhanced Language Representation with Informative Entities(ERNIE)/Bi-directional Gated Recurrent Unit(BiGRU)/Universal Information Ex-traction(UIE)分类
信息技术与安全科学引用本文复制引用
王梓鸣..基于EG-UIE的危化品事故处置措施实体识别方法研究[J].安全、健康和环境,2026,26(4):65-73,9.基金项目
山东省重点研发计划(2024CXGC010706),危化品全流程风险管控技术与装备开发. （2024CXGC010706）
