Intelligent proposition method based on knowledge graph and pre-training generation model

By judging the effectiveness of knowledge graph transformation and detecting embedding quality, the combination of knowledge graph and generation model is dynamically optimized, which solves the problem of the loose combination between knowledge graph and generation model and realizes high-quality proposition generation.

CN121279406BActive Publication Date: 2026-06-09GUANGZHOU EVERBRIGHT EDUCATION TECH CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202511375616.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-09-24
Publication Date
2026-06-09
Estimated Expiration
2045-09-24

AI Technical Summary

Technical Problem

In existing technologies, the integration of knowledge graphs and generative models is not tight, which leads to the omission of connections between key knowledge points when generating questions, making it difficult to meet the needs of high-quality question creation.

Method used

By monitoring the structured information transformation process of knowledge graphs, obtaining the results of transformation effectiveness judgment, determining whether to embed pre-trained generative models, and performing graph embedding initialization adjustment or transformation optimization when necessary, we can ensure the accurate representation and capture of entity, relationship and structural information.

Benefits of technology

It improves the accuracy and diversity of proposition generation, ensures that the structural information of the knowledge graph is not lost or distorted during the conversion process, enhances the generative model's ability to understand the knowledge graph, and strengthens the quality and reliability of generated propositions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121279406B_ABST
    Figure CN121279406B_ABST
Patent Text Reader

Abstract

The application discloses an intelligent proposition method based on a knowledge graph and a pre-training generation model, and relates to the technical field of intelligent proposition.The method comprises the following steps: graph conversion effective determination, graph embedding quality detection and generation model intelligent proposition.The application monitors the structured information conversion process of the knowledge graph and obtains a conversion effectiveness determination result, which helps to more accurately determine whether the hierarchical and correlation information of the knowledge graph is completely retained, and provides a reliable basis for subsequent embedding generation models.On this basis, it is determined whether graph embedding initialization adjustment is needed to enhance the representation and capture ability of the pre-training generation model for entity, relationship and structure information, and improve the learning effect of the model.Further, based on the graph embedding detection result, it is determined whether the pre-training generation model is qualified, the quality of the graph embedding can be timely evaluated, the information in the embedding process is ensured not to be distorted or missed, and the depth and breadth of proposition generation are helped to improve.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • A method for constructing knowledge graph based on large language model and vector library

    CN119129722B

  • AI intelligent proposition method

    CN119849614A

  • Knowledge proposition error correction method and system based on knowledge graph

    CN118709695A

  • Knowledge proposition error correction method and system based on knowledge graph optimization and upgrading

    CN120373298A