基于深度学习的作战文书命名实体识别  

Named Entity Recognition for Combat Documents Based on Deep Learning

作　　者：张晓海 操新文 高源 ZHANG Xiao-Hai;CAO Xin-Wen;GAO Yuan(Joint Operation College, NDU of PLA, Shijiazhuang 050084, China)

机构地区：[1]国防大学联合作战学院

出　　处：《指挥控制与仿真》2019年第4期22-26,共5页Command Control & Simulation

摘　　要：随着深度学习技术的不断发展,基于双向长短期记忆网络和条件随机场的神经网络模型被广泛应用到作战文书以及军事命名实体识别的研究中。提出了一种基于CNN-BiLSTM-CRF的作战文书命名实体识别方法,首先利用卷积神经网络(CNN)提取字符级特征向量,而后与词向量、词性特征向量进行拼接作为输入,从而达到提高识别率的目的。同时,分析命名实体在不同分类标注情况下,对模型性能所产生的影响,并提出一种针对作战文书命名实体识别的细分类标注策略。实验结果证明,该模型相对于其他方法表现出了更好的性能,且细分类的标注体系对于特征表达有一定的帮助。With the development of deep learning, BiLSTM-CRF model are widely used in combat documents and military named entity recognition.This paper proposes a method based on CNN-BiLSTM-CRF for identifying the named entity of combat documents.In this paper, the character-level feature is extracted by convolutional neural network (CNN).Then we have stitched character vectors and word vectors and part-of-speech(POS) vectors together.At the same time, we analyze the impact of different annotations on the performance of the models.A labeling strategy of fine classification is proposed.Finally, the result has shown that, the performance of our method is better than others without relying on artificial features.And the labeling strategy of fine classification is helpful for the expression of features.

关 键 词：深度学习 作战文书 命名实体识别 神经网络 

分 类 号：E94[军事] TP391[自动化与计算机技术—计算机应用技术]