摘要
为充分发挥中文航空安全短文本信息在由数据驱动的安全管理中的作用,提出基于语义角色标注和依存句法分析以及K-means++算法的中文航空安全信息核心内容获取和主题分析方法。以2017-2020年“偏离姿态/高度”事件类型信息为样本进行实例分析,结果表明,利用语义角色标注和依存句法分析可获取样本的核心内容,即飞行机组操作特征,通过K-means++聚类可挖掘核心内容中潜在的主题,即违规、遗漏、动作目标差错、动作幅度差错、动作时间差错和注意力分配六个主题,通过社会网络对主题进行展示和分析可发现动作目标差错、违规和注意力分配三个主题处于核心地位,尤其违反交叉检查程序是实际安全管理工作的重点,是减少偏离姿态/高度事件的核心环节。
In order to make full use of Chinese aviation safety short text information in data-driven safety management,a method of Chinese aviation safety information core content acquisition and topic analysis based on semantic role labelling,dependency parsing and K-means++algorithm is proposed.Taking the event type"deviation attitude/altitude"from 2017 to 2020 as an example,the results show that the core content of the sample,namely flight crew operation characteristics,can be obtained by semantic role tagging and dependency parsing,and the potential topics in the core content can be mined by K-means++text clustering.The potential topics in the core content can be mined through K-means++text clustering,that is,violations,omissions,action target errors,and so on Through the social network to display and analyze the topics,it can be found that the three topics of action target error,violation and attention allocation are in the core position,especially the violation of cross check procedure is the focus of the actual safety management work,and the core link to reduce the deviation attitude/altitude events.
作者
刘俊杰
叶英豪
LIU Junjie;YE Yinghao(School of Safety Science and Engineering,Civil Aviation University of China,Tianjin 300300,China)
出处
《综合运输》
2022年第5期47-52,共6页
China Transportation Review
关键词
航空安全信息
安全管理
语义角色标注
依存句法分析
文本聚类
Aviation safety information
Safety management
Semantic role tagging
Dependency parsing
Text clustering
