基于SG-CIM模型的实体自动对齐框架研究  

An Automatic Entity Alignment Framework for SG-CIM Model

作　　者：杨诗语 王洁 任小伟 李雨霏 YANG Shiyu;WANG Jie;REN Xiaowei;LI Yufei(Big Data Center,State Grid Corporation of China,Xicheng District,Beijing 100032,China)

机构地区：[1]国家电网有限公司大数据中心,北京市西城区100032

出　　处：《电力信息与通信技术》2023年第10期66-71,共6页Electric Power Information and Communication Technology

摘　　要：SG-CIM模型是一种基于国家电网业务场景和数据需求的电网统一数据模型。该模型主要由3种不同类型的模型表组成,主要是根据网格的业务需求手工设计的。因此,要将SG-CIM模型应用于国家电网业务场景,需要实现SG-CIM公共数据模型中的物理表、逻辑表和标准表的统一。为了实现上述目标,需要找到一种有效的实体对齐统一方法来实现不同知识图谱的统一任务。文章针对SG-CIM模型实体,提出了一种模型表实体自动对齐框架。该框架能够有效计算模型表实体之间的相似度和关联度,有助于实现不同知识图之间的实体对齐。实体对齐模型基础框架还通过对三元实体使用推理规则丰富三元实体对齐的数量,提高了文章框架对三元实体的学习效果。通过设计相关实验,证明该模型表实体自动对齐框架在真实图数据上优于传统的实体对齐方法。SG-CIM model is a kind of unified data model for the power grid based on the business scenarios and data requirements of the state grid.The model mainly consists of three different types of model tables,which are mainly designed manually based on the business requirements of the grid.Therefore,to apply the SG-CIM model to the state grid business scenario,it is necessary to achieve the unification of physical,logical and standard tables in the SG-CIM.In order to reach the above goal,an effective entity alignment unified method should be used to achieve the unified task of different knowledge graphs.In this paper,we propose an automatic model-table-entity alignment framework for SG-CIM entities.This entity alignment framework can effectively calculate the similarity and association degree of model-table-entities,which helps to achieve entity alignment among different knowledge graphs.The entity alignment framework also enriches the number of triple entities used for alignment by using inference rules for triple entities,which improves the learning effect of the framework proposed in this paper for triple entities.Relevant experiments are designed in this paper to demonstrate that this model-table-entity alignment framework outperforms traditional entity alignment methods on real graph data.

关 键 词：SG-CIM 知识图谱 自然语言处理 实体对齐 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]