Data processing method and device and electronic equipment
By generating a knowledge graph visual subgraph and combining it with large-scale model processing, the problems of insufficient expression of multiple hop relationships and loss of structural information in knowledge graph questions and answers are solved, and the accuracy and completeness of answer generation are improved.
Patent Information
- Application Number
- CN202510228222.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-27
- Publication Date
- 2025-07-18
AI Technical Summary
In the knowledge graph question and answer tasks, the existing technology has problems such as insufficient expression of multiple hop relationships, insufficient utilization of semantic correlation prior information, and loss of graph structure information, which affects the accuracy and completeness of answer generation.
By recalling triple data from the knowledge graph subgraph, a knowledge graph visual subgraph is generated, and the problem data, text sequence and visual subgraph are input to the big model at the same time, an answer or query statement is generated, the knowledge graph structure information is retained, and the multi-hop relationship is intuitively expressed.
It improves the answer generation accuracy of the knowledge graph question and answer system, effectively integrates semantic correlation information, and improves the accuracy and completeness of the answers.
Smart Images
Figure CN120336501A_ABST