空中交通管制中的危险源关键特征提取  

Feature Extraction of Air Traffic Control Hazard Based on Text

作　　者：蔡志鹏 曾维理 郭子逸 CAI Zhi-peng;ZENG Wei-li;GUO Zi-yi(Nanjing University of Aeronautics and Astronautics,Nanjing 211000,China;State Key Laboratory of Air Traffic Management System,Nanjing 211000,China)

机构地区：[1]南京航空航天大学,江苏南京211000 [2]空中交通管理系统全国重点实验室,江苏南京211000

出　　处：《航空计算技术》2023年第6期35-39,共5页Aeronautical Computing Technique

基　　金：国家重点研发计划项目资助(2022YFB2602401);空中交通管理系统与技术国家重点实验室资助(SKLATM202007)。

摘　　要：空中交通管制危险源特征提取是针对空中交通危险源记录的关键信息检索任务。基于文本特征的提取方法仅依据统计结果来对术语的重要性进行排序。通过引入专家知识来提取符合危险源定义的特征,提出了一种基于专家知识的特征提取方法来定义危险源特征并构建危险源分析框架。算法通过迭代对专家知识进行提炼和积累。实验结果表明,基于专家知识的特定语言处理后得到的数据集可以提取更多信息量的候选特征,以便通过k-means算法构建危险源记录的分析语境。该特征提取模型比其他四种算法获得了更高的精度,在空中交通管制运行数据集和通导数据集中分别达到82%和86%。此外,信息丰富的危险特征为安全管理部门的决策提供了支持,降低了隐患排查的成本。ATC(Air Traffic Control)hazard feature extraction is a key information retrieval task for air traffic hazard records.Text-based feature extraction relies purely on textual statistical results to rank the term′s importance,but we attempt to use external knowledge to extract hazard features that satisfy the definition of hazards.This paper proposes a feature extraction method based on expert knowledge to define hazard features and construct a hazard analysis framework.The algorithm refines and accumulates expert knowledge through iteration.The experiment results demonstrate that the dataset obtained after specific linguistic processing based on expert knowledge can extract more informative candidate features to construct analysis context by k-means.The feature extraction model achieves greater accuracy than the other four algorithms,reaching 82%and 86%in the ATCO(Air Traffic Control Operation)dataset and CMS(Communication Navigation and Surveillance)dataset,respectively.Additionally,the information-rich hazard features provide support for the decision-making of the safety management department,reducing the cost of hidden hazard investigations.

关 键 词：文本特征提取 空中交通管制 规范化 危险源 领域知识 

分 类 号：V355[航空宇航科学与技术—人机与环境工程]