摘要
本文尝试解决中文学者命名实体与外部知识库的实体对齐短缺的问题。通过SPARQL语义查询抽取维基数据子图——中国科学院院士的知识图谱子图,初步构建国内知识库的中文院士实体与Wikidata实体的对齐以及与外部ID对应的知识库的实体对齐。对院士实体的三个数量型特征对齐的外部ID个数(ids)、不同语种的Wikipedia站点个数(sites)、实体的全部陈述个数(states)与目标分类(有无VIAF实体对齐)的相关分析发现,目标分类与ids特征正向相关最强,直接VIAF实体对齐只存在ids高区的院士,占比偏低。因此,提出利用LC、ISNI等外部ID,应用VIAF对重要来源库的重定向功能,构建间接的VIAF实体对齐的方法。本文为中文知识库进行外部实体对齐提供了可行的初步方案,提出的实验方法显著地提高了较小ids值(1-7)的院士拥有VIAF实体对齐的个数,最终通过实体对齐的VIAF信息集成增加了院士实体的ids数量,丰富了中文学者与外部知识库的实体对齐信息。图4。表5。参考文献19。
This paper attempts to solve the shortage of Chinese named entity alignment with external knowledge base.By selecting Wikidata subgraph about academician of Chinese Academy of Sciences,initially mapping the Chinese academician entity in knowledge base in our country with the entity in the Wikidata,and mapping with the entity of external-ID in other knowledge bases,the paper analyses the distribution of three quantitative features which include quantity of external ID(ids),quantity of Wikipedia sites for different languages(sites),quantity of all entity statement(states),and then present visual graph of the relation between features and target classification VIAF ID.The paper initially finds that target classification is most positive correlated with the ids and the ratio of the academician with direct VIAF ID only existing in ids high-level area is relatively low.Therefore,this paper proposes that it is necessary to further exploit LC,ISNI and other external-ID,take advantage of the redirection function of VIAF to important source database,and find out a way to build indirect VIAF entity alignment.This paper not only provides an initial and useful method for external entity alignment of Chinese knowledge base,but also improves markedly the number of academicians with small ids(1-7)in having VIAF entity alignment with an experiment method.What s more,through integration with VIAF,we can add the quantity of ids of academician entity and enrich the entity alignment information of Chinese scholar with external knowledge base.4 figs.5 tabs.19 refs.
作者
王瑞云
贾君枝
Wang Ruiyun;Jia Junzhi
出处
《国家图书馆学刊》
CSSCI
北大核心
2020年第2期102-112,F0003,共12页
Journal of The National Library of China
基金
国家社会科学基金项目“中文学术领域命名实体的知识图谱构建研究”(项目编号:18BTQ072)的研究成果之一。