Knowledge graph embedding method based on self-interleaving attention mechanism and diffusion aggregation
The invention discloses a knowledge graph embedding method based on a self-interleaving attention mechanism and diffusion aggregation, which comprises the following steps: taking entities and relationships in a knowledge graph as nodes, and stretching the knowledge graph into a special heterogeneous...
Saved in:
Main Authors | , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
12.03.2024
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention discloses a knowledge graph embedding method based on a self-interleaving attention mechanism and diffusion aggregation, which comprises the following steps: taking entities and relationships in a knowledge graph as nodes, and stretching the knowledge graph into a special heterogeneous directed graph; in the heterogeneous directed graph, neighbor matrixes of all nodes in the graph are obtained based on preset hyper-parameter analysis. For a neighbor list under each diffusion distance, the method respectively learns attention vector representations of three levels of diffusion attention, node attention and semantic distance attention. Wherein in the learning process of distraction, the entity node performs attention calculation by using the relation node characteristics contained in the path, and the relation node performs calculation by using the characteristics of the entity node. According to the method, feature aggregation is carried out on a target node on diffusion paths with different leng |
---|---|
Bibliography: | Application Number: CN202311680783 |