一种多信息感知的知识图谱实体对齐方法
By integrating various types of information from knowledge graphs, optimizing entity embedding representations using graph attention networks and graph convolutional networks, and combining them with delayed acceptance algorithms, the problems of insufficient information utilization and alignment conflicts caused by local alignment strategies in existing technologies are solved, achieving higher accuracy in entity alignment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHEASTERN UNIV CHINA
- Filing Date
- 2023-08-10
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies fail to fully utilize entity names, structural information, attribute information, and relational information in knowledge graph entity alignment, resulting in poor alignment performance. Furthermore, local alignment strategies lead to alignment conflicts among multiple entities, affecting accuracy.
A multi-information perception approach is adopted, which utilizes entity names, structural information, relational information and attribute information in knowledge graphs, and combines graph attention network and graph convolutional network to perform global entity alignment. An improved attention mechanism and a high-level graph convolutional network are designed, Manhattan distance and regularizer are used to optimize entity embedding representation, and a delayed acceptance algorithm is used for stable matching.
It improves the accuracy of entity alignment, optimizes entity embedding representation by integrating multiple types of information, achieves stable one-to-one matching, and enhances the quality of knowledge graphs.
Smart Images
Figure QLYQS_1 
Figure QLYQS_23 
Figure QLYQS_24