面向学术全文本的南海维权证据知识元自动识别研究  

Research on Automatic Recognition of Evidence Knowledge Element of South China Sea Rights Protection for Full Academic Texts

作　　者：程为 郑轩昂 郑德俊[1] 杨海平[3] 王燕红 Cheng Wei;Zheng Xuanang;Zheng Dejun;Yang Haiping;Wang Yanhong(College of Information Management,Nanjing Agriculture University,Nanjing 210031;School of Data Science and Engineering,East China Normal University,Shanghai 200062;School of Information Management,Nanjing University,Nanjing 210023)

机构地区：[1]南京农业大学信息管理学院,南京210031 [2]华东师范大学数据科学与工程学院,上海200062 [3]南京大学信息管理学院,南京210023

出　　处：《情报杂志》2023年第9期141-148,共8页Journal of Intelligence

基　　金：国家社会科学基金重大项目“南海疆文献资料整理中的知识发现与维权证据链构建研究”(编号:19ZDA347)。

摘　　要：[研究目的]自动识别出潜藏在非结构化南海维权学术全文本中的证据知识元,是完整、全面、多角度地重组织证据知识元并构建证据链、厘清我国南海维权历史过程的基础。[研究方法]根据证据的内涵及知识元语义描述模型理论,提出证据知识元的概念;分析证据知识元的描述需求,以结构化的形式定义了面向南海维权学术全文本的证据知识元表示模型,并在此基础上提出证据知识元自动识别方法;以南海维权证据知识元自动识别的实证研究验证了该方法的可行性。[研究结论]证据知识元表示模型在面向南海维权学术全文本时具有适用性,结合证据知识元识别规则与深度学习方法达到较好的自动识别效果,其中BERT模型和ERNIE模型表现最佳,在南海维权证据知识元自动识别的对比实验中,微观F1值分别达到了96.75%和96.64%,明显领先其他模型,可以满足南海维权证据知识元自动识别的要求。[Research purpose]Automatic identification of evidence knowledge element hidden in the unstructured full academic texts of South China Sea rights protection is the basis for reorganizing evidence knowledge element,constructing evidence chain and clarifying the historical context of South China Sea rights protection in a complete,comprehensive and multi-angle way.[Research method]Firstly,this paper puts forward the concept of evidence knowledge element according to the connotation of evidence and the theory of knowledge element semantic description model.Then,the description requirement of evidence knowledge element is discussed,the evidence knowledge element representation model for the full academic text of South China Sea rights protection is defined in a structured form,and the automatic recognition method of evidence knowledge element is proposed on this basis.Finally,the feasibility of the method is verified by an empirical study on the automatic recognition of evidence knowledge element.[Research conclusion]The evidence knowledge element representation model has applicability when facing the full academic texts of South China Sea rights protection.The combination of evidence knowledge element recognition rules and deep learning method achieves good automatic recognition effect,among which BERT model and ERNIE model perform best.In the comparative experiment of automatic recognition of evidence knowledge element,The Micro_F1 values are 96.75%and 96.64%respectively,which are obviously ahead of other models and can meet the requirements of automatic recognition of evidence knowledge element for South China Sea rights protection.

关 键 词：学术全文本 知识元 南海维权 证据知识元 深度学习 文本分类 

分 类 号：G353.1[文化科学—情报学] TP391.1[自动化与计算机技术—计算机应用技术]