摘要
针对军事领域的命名实体识别问题,提出一种基于BiLSTM-CRF的实体识别方法,旨在识别军事文本中的人名、军用地名、军事机构名、武器装备、设施目标、部队番号等军事命名实体。使用词嵌入方法自动学习中文字符的分布式表示作为模型输入;利用双向长短时记忆(Bi-directional Long-Short Term Memory,BiLSTM)神经网络处理输入的字符向量序列,统筹上下文语义学习任务特征;将学习到的特征接入线性链式条件随机场(CRF)进行军事命名实体标注,获得命名实体识别结果并输出。在人工构建数据集上的实验结果表明,提出的方法能够很好地完成军事命名实体识别任务。
To deal with the problem of named entity recognition in the military field,a method of entity recognition based on BiLSTM-CRF is proposed,which aims to recognize character names,military place names,military institutions,equipment names,military facilities and military code designation.Firstly,the model uses word embedding method to automatically learn the distributed representation of Chinese characters as its input.Secondly,a Bi-directional Long-Short Term Memory(BiLSTM)neural network is used to learn the text features from the character vector sequence output by the previous step.Finally,a linear chain Conditional Random Field(CRF)is used to tag the military named entities according to the learned feature vector sequence,and the results of military named entity recognition are obtained and output.Experiment results with artificial dataset show that the proposed method has good effect on the named entity recognition of the military domain.
作者
高学攀
杜楚
吴金亮
GAO Xuepan;DU Chu;WU Jinliang(The 54th Research Institute of CETC,Shijiazhuang 050081,China)
出处
《无线电工程》
2020年第12期1050-1054,共5页
Radio Engineering