摘要
有效地利用实体识别技术识别出航空不安全信息中的风险要素,提升安全风险识别和评价能力,对于实现航空安全风险量化分析具有重要意义。为精准识别非结构化航空不安全信息中的风险要素,提出了一种融合先验规则和知识增强语义表示模型的实体识别模型,以大量航空不安全事件案例文本报告为分析对象,在自建标注语料上训练ERNIE模型,获取动态词向量,同时引入能够表达风险要素结构的规则进行优化。实验结果表明,模型在测试集上的精确率、召回率和F1值分别达到了92.5%、92.4%和92.4%,优于实验对比的其他模型。模型能够有效识别航空不安全信息中隐含的风险要素,为航空安全风险量化分析提供数据支持。
Efficient use of entity identification techniques to identify risk elements in aviation security information is important to improve the ability of security risk identification and assessment for quantitative analysis of aviation security risks.In order to accurately identify the risk elements in unstructured aviation insecurity information,an entity recognition model combining Prior Rule and ERNIE(PR-ERNIE)was proposed,and a large number of aviation insecurity incident case text reports were taken as the analysis object.The ERNIE model is trained on a self-constructed labeled corpus to obtain dynamic word vectors,and rules that can express the structure of risk elements are introduced for optimization.Experimental results show that the model achieves accuracy,recall and F1 values of 92.5%,92.4%and 92.4%,respectively,on the test set,which is better than the other models compared in the experiments.The proposed model can effectively identify hidden risk elements in aviation insecurity information and provide data support for quantitative analysis of aviation security risks.
作者
程童
高振兴
王浩锋
CHENG Tong;GAO Zhen-xing;WANG Hao-feng(Nanjing University of Aeronautics and Astronautics,Nanjing 211000,China;China Academy of Civil Aviation Science and Technology,Beijing 100000,China)
出处
《航空计算技术》
2023年第6期25-29,共5页
Aeronautical Computing Technique
基金
国家自然科学基金项目资助(52272351)。
关键词
风险分析
命名实体识别
规则匹配
预训练模型
risk analysis
named entity recognition
rule matching
pre-training model