Time sequence knowledge graph prediction method fusing historical rules and curriculum enhanced embedding
By integrating historical rules and curriculum-intensive embedding methods, a new timing knowledge graph prediction method is proposed, which solves the shortcomings of existing methods in capturing rules and handling complex timing relationships, and achieves higher prediction accuracy and efficiency.
Patent Information
- Application Number
- CN202510238071.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-28
- Publication Date
- 2025-06-17
AI Technical Summary
The existing timing knowledge graph prediction methods have shortcomings in capturing rules in the graph and dealing with complex timing relationships, and how to reasonably introduce course learning to improve embedded prediction performance is still a difficult point.
A time sequence knowledge graph prediction method that integrates historical rules and course-strengthening embedding is proposed. The rules are extracted through the rule module and traversed the quadruple related to the rules for reasoning. The embedding module adopts course learning to adjust the training strategy, and gradually learns the information in the time sequence knowledge graph from simple to complex, and represents entities, relationships and time as vectors for prediction.
The prediction accuracy and efficiency in the timing prediction task were improved. The experimental results showed that the existing mainstream methods were better than the existing mainstream methods on the four public timing knowledge graph benchmark data sets, verifying the effectiveness and innovation of this method.