基于BERT和分面分类的多标签的南海证据性数据分类研究  

Research on Multi-label Evidence Data of the South China Sea Classification Based on BERT and Faceted Classification

作　　者：彭玉芳 石进[2] 徐浩 杨海平[2] Peng Yufang;Shi Jin;Xu Hao;Yang Haiping(i School of Economics&Management,Nanjing Institute of Technology;Department of Information Management,Nanjing University)

机构地区：[1]南京工程学院经济与管理学院,江苏南京211167 [2]南京大学信息管理学院,江苏南京210046

出　　处：《图书馆杂志》2022年第5期102-108,共7页Library Journal

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建构研究”(项目编号:19ZDA347);南京工程学院校级科研基金创新基金面上项目“学术文献方法论知识驱动的跨学科知识扩散特征识别研究”(项目编号:CKJB202003);江苏省社科基金青年项目“融合在线学术社交信息的电子文献资源推荐策略研究”(项目编号:21TQC003)的研究成果之一。

摘　　要：为了能更好地组织南海证据性数据,提高南海疆智库的检索性能,便于科研人员对南海问题展开深入研究,本文融合了中国图书馆分类法的族性检索和分面分类法的特性检索,构建了南海文献分类法,应用BERT预训练模型实现细粒度的多标签南海证据性数据的自动分类,并通过数据库技术将南海证据性数据分为12个面进行组配检索。通过以上方法赋予每一条南海证据性数据唯一的分类号和类名,形式化表示为南海证据性数据的数据分类=证据来源分类+证据的数据体系分类+证据的数据分面分类。从更细粒度的视角,实现数据分类,与粗放式文献载体的分类相比,数据分类更有利于展开数据挖掘,找到数据间潜在的关联关系,发现新的研究视角,是科学研究创新的基础和保障。In order to better organize the evidence data of the South China Sea,improve the retrieval function of the South China Sea Think Tank,and facilitate researchers to conduct in-depth research on the South China Sea issue,this paper combines the ethnic retrieval of the Chinese Library Classification and the feature retrieval of the faceted classification,constructs the South China Sea document classification,applies the BERT pre-training model to realize the automatic classification of the fine-grained multi-label South China Sea evidence data,and uses the database technology to divide the evidence data of the South China Sea into 12 facets for grouping and retrieval.Accordingly,each piece of South China Sea evidence data is given a unique classification number and category name,which is formally expressed as “the data classification of South China Sea evidence data=evidence source classification + evidence data system classification + evidence data facet classification”.From a more fine-grained perspective,data classification is realized.Compared with the classification of broad-type document carriers,data classification is more conducive to data mining,to the identification of potential relationship between data,and to the discovery of new research perspectives,which are the foundation of scientific research and innovation.

关 键 词：BERT 分面分类 数据分类 多标签分类 南海证据性数据 

分 类 号：G254.1[文化科学—图书馆学]