摘要
针对军事指挥控制保障领域知识图谱构建的实际,提出了一种融合汉字多特征的BiLSTM+CRF命名实体识别模型,验证了拼音特征、五笔编码特征和分词边界特征对于模型性能的影响。对军事领域的命名实体识别相关工作和整体研究框架进行介绍;详细介绍模型中各层实现的原理和相关细节;通过在军事指挥控制保障领域命名实体识别语料库C2NER上进行实验,结果表明分词边界特征对于模型性能的提升效果较为明显,F1值从67.21%提升到了70.23%,而拼音特征和五笔特征对于模型性能提升则效果一般。
Aiming at the reality of knowledge graph construction in military command and control support field,this paper proposes a named entity recognition model of BiLSTM+CRF,which integrates multiple features of Chinese characters,and verifies the influence of Pinyin features,Wubi coding features and word segmentation boundary features on the performance of the model.Firstly,the related work and the whole research framework of named entity recognition in military field are introduced;Secondly,the implementation principle and related details of each layer in the model are introduced in detail;Finally,the experiment is carried out on C2NER,which is a named entity recognition corpus in the field of military command and control support.The results show that the segmentation boundary feature can improve the performance of the model obviously,and the F1 value is increased from 67.21%to 70.23%,while the Pinyin feature and Wubi feature can improve the performance of the model generally.
作者
袁清波
杜晓明
姚奕
杨帆
蒋祥
YUAN Qingbo;DU Xiaoming;YAO Yi;YANG Fan;JIANG Xiang(Command and Control Engineering College,Army Engineering University of PLA,Nanjing 210007,China)
出处
《火力与指挥控制》
CSCD
北大核心
2022年第9期48-53,共6页
Fire Control & Command Control
基金
全军军事类研究生资助课题(JY2019C078)。
关键词
命名实体识别
汉字特征
军事指挥控制保障
知识图谱
named entity recognition
characteristics of Chinese characters
military command and control support
knowledge graph
