现代情报2025，Vol.45Issue(10)：39-50,12.DOI:10.3969/j.issn.1008-0821.2025.10.004
融合大模型与图嵌入模型的领域知识图谱补全研究
Research on Domain Knowledge Graph Completion by Integrating Large Models and Graph Embedding Models
摘要
Abstract
[Purpose/Significance]To improve the performance of domain knowledge graph completion and address the challenges of insufficient semantic understanding in existing graph embedding models and generation bias and computa-tional resource waste in large models,this paper proposes a domain knowledge graph completion framework that integrates large models and graph embedding models.[Method/Process]Firstly,deep pre-training of domain corpus was conducted on the open-source large model to enhance its understanding of domain terminology during knowledge graph completion.Secondly,the traditional graph embedding model was used to generate candidate relationships or entities based on the exist-ing structure of the knowledge graph,providing a high-quality candidate set for subsequent use of large models for knowl-edge graph completion.Thirdly,based on different prompt word strategies,the pre-trained domain model was guided to sort the candidate options,achieving efficient completion of the knowledge graph.Finally,empirical research was con-ducted on existing datasets in the biomedical field to verify its feasibility.[Result/Conclusion]The experimental results show that the method proposed in this study has significant effects on multiple evaluation indicators,and can provide new ideas and technical means for subsequent domain knowledge graph completion.关键词
知识图谱/大语言模型/知识图谱补全/图嵌入模型/Prompt提示词Key words
knowledge graph/large language model/knowledge graph completion/graph embedding model/prompt words分类
信息技术与安全科学引用本文复制引用
张君冬,严颖,王震宇,刘江峰,刘艳华,黄奇..融合大模型与图嵌入模型的领域知识图谱补全研究[J].现代情报,2025,45(10):39-50,12.基金项目
江苏省研究生科研与实践创新计划项目"面向用户响应式场景的在线医疗智慧问答服务研究"(项目编号:KYCX25_0125) （项目编号:KYCX25_0125）
江苏高校哲学社会科学研究重大项目"中医古籍文献预训练模型构建及其应用研究"(项目编号:2023SJZD084) （项目编号:2023SJZD084）
江苏智慧中医药健康服务工程研究中心开放课题项目(项目编号:ZHZYY202501). （项目编号:ZHZYY202501）
