基于ERNIE的航空情报文本信息抽取方法研究OA
Research on Aviation Information Text Extraction Methods Based on ERNIE
针对航空资料汇编(AIP)数字化过程中文本结构复杂、专业术语密集、人工处理效率低的问题,提出一种航空情报文本信息抽取方法.构建领域实体体系与标注数据集,采用 ERNIE-BiLSTM-CRF 模型进行语义表示、上下文建模与序列解码,并与传统模型对比.实验结果表明,该方法在命名实体识别任务中精确率为 91.8%、召回率为90.6%、F1 值为 91.2%,较对比模型有一定提升,在专业术语和长实体识别上更稳定.研究为 AIP 关键信息结构化提取及数字航空情报产品构建提供了可行技术路径.
To address the challenges of complex structure,dense terminology,and low manual processing efficiency in the digitization process of Aeronautical Information Publication(AIP),this paper proposes an information extraction method for aeronautical text.A domain-specific entity system and annotated dataset are constructed,and the ERNIE-BiLSTM-CRF model is employed for semantic representation,contextual modeling,and sequence decoding,in comparison with traditional models.Experimental results show that the proposed method achieves a precision of 91.8%,a recall of 90.6%,and an F1-score of 91.2%in the named entity recognition task,demonstrating an improvement over the baseline model and greater stability in recognizing technical terms and long entities.This study provides a feasible technical pathway for structured extraction of key AIP information and the development of digital aeronautical intelligence products.
唐凯;赖欣;李炎林
中国民用航空飞行学院,四川 成都 641419中国民用航空飞行学院,四川 成都 641419中国民用航空飞行学院,四川 成都 641419
信息技术与安全科学
航空情报航空资料汇编信息抽取命名实体识别ERNIE
aeronautical informationAeronautical Information Publication(AIP)information extractionnamed entity recognitionERNIE
《现代信息科技》 2026 (7)
58-61,67,5
中国民用航空飞行学院重点科研项目(25CAFUC03046)中国民用航空飞行学院研究生创新创业项目(25CAFUC10041)
评论