基于企业知识图谱的多源数据融合分析  

Multi-source Data Fusion Analysis Based on Enterprise Knowledge Map

作　　者：杨椋 柯枫 刘新明 信希涛 YANG Liang;KE Feng;LIU Xinming;XIN Xitao(Institute of Network Technology,ICT(YANTAI),Yantai 264003,China)

机构地区：[1]烟台中科网络技术研究所,山东烟台264003

出　　处：《现代信息科技》2020年第23期100-102,共3页Modern Information Technology

摘　　要：面向企业知识图谱的构建与分析,通过获取企业信息、企业高管信息、企业新闻资讯信息等,自动构建企业—高管—资讯知识图谱。在构建过程中,对实体的正确引用是推导、分析、构建知识库的基础和保障。构建过程中出现的部分实体重名的问题,对应于实体指称项的歧义性,会造成图谱中部分节点重合、链接错误,进而对整体知识库的构建产生负影响。为此,提出一种基于随机游走的方法解决知识图谱中的实体消歧问题,取得很好的消歧效果。For the construction and analysis of enterprise knowledge map,the enterprise-enterprise executives-information knowledge map is automatically constructed through acquiring enterprise information,enterprise executive information and enterprise’s news information.In the construction process,the correct reference to the entity is the basis and guarantee of derivation,analysis and construction of knowledge base.The problem of duplicate name of some entities appeared in the construction process,which corresponds to the ambiguity of entity references,will cause some nodes overlap and link errors in the map,and thus has negative effects on the construction of the whole knowledge base.To this end,a method based on random walk is proposed to solve the problem of entity disambiguation in knowledge map,and get better disambiguation effect.

关 键 词：知识图谱 实体消歧 随机游走 文本向量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]