基于依存关系的命名实体识别  

Named Entity Recognition Based on Dependency

作　　者：张雪松 郭瑞强[1,2] 黄德根[3] ZHANG Xuesong;GUO Ruiqiang;HUANG Degen(School of Computer and Cyber Security,Hebei Normal University,Shijiazhuang,Hebei 050024,China;Hebei Provincial Engineering Research Center for Supply Chain Big Data Analytics&Data Security,Hebei Normal University,Shijiazhuang,Hebei 050024,China;School of Computer Science and Technology,Dalian University of Technology,Dalian,Liaoning 116024,China)

机构地区：[1]河北师范大学计算机与网络空间安全学院,河北石家庄050024 [2]河北师范大学河北省供应链大数据分析与数据安全工程研究中心,河北石家庄050024 [3]大连理工大学计算机科学与技术学院,辽宁大连116024

出　　处：《中文信息学报》2021年第6期63-73,共11页Journal of Chinese Information Processing

基　　金：国家自然科学基金(61672127,U1936109);河北省重点研发计划项目民生专项(20375701D)。

摘　　要：现有的命名实体识别方法主要是将句子看作一个序列进行处理,忽略了句子中潜在的句法信息,存在长距离依赖问题。为此,该文提出一种基于依存关系的命名实体识别模型,通过在输入数据中增加依存树信息,改变双向长短时记忆网络的层间传播方式,以获得单词在依存树中的子节点和父节点信息,并通过注意力机制动态选择两者的特征,最后将特征输入到CRF层实现命名实体标注。实验表明,该方法较BiLSTM-CRF模型在性能上得到了提高,且在长实体识别上优势明显。在OntoNotes 5.0 English和OntoNotes 5.0 Chinese以及SemEval-2010 Task 1 Spanish上的F1值分别达到了88.94%、77.42%、84.38%。Most of the existing named entity recognition methods treat the sentence as a sequence,ignoring the syntactic information in the sentence.This paper proposed a named entity recognition model based on dependency relationship.Adding dependency tree information to the input data,the child and parent node information of words in the dependency tree are obtained by changing the inter layer propagation mode in Bi-LSTM.The features are dynamically selected by the attention mechanism.Finally,the CRF layer is adopted to realize named entity annotation.Experimental results show that the proposed method is better than BiLSTM-CRF model,especially for long entity recognition,achieving 88.94%,77.42%and 84.38%F1 values on OntoNotes 5.0 English,OntoNotes 5.0 Chinese and semeval-2010 Task 1 Spanish respectively.

关 键 词：命名实体识别 依存树 有向图 注意力机制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]