计算机技术与发展2025，Vol.35Issue(8)：10-18,9.DOI:10.20165/j.cnki.ISSN1673-629X.2025.0055
多模态知识图谱补全研究综述
Review of Research on Multimodal Knowledge Graph Completion
摘要
Abstract
Multi-modal knowledge graph completion(MMKGC)aims to automatically discover new knowledge in a given multi-modal knowledge graph(MMKG)by collaboratively modeling the structural and multi-modal information of the entity ternary.However,the main challenges of real-world MMKGCs lie in the diversity of data and the unevenness of data distribution.Specifically,modal information can span multiple types,e.g,images,text,numbers,audio,video,etc.,but the distribution of such information among different entities is extremely uneven,resulting in the lack of corresponding modal data for some entities,and this status quo is precisely the core driving force for the development of multimodal knowledge graph complementation technology.We mainly synthesize the relevant research contents of multimodal knowledge graph complementation.Firstly,the research status at home and abroad is summarized and organized,and the relevant knowledge of multimodal knowledge graph complementation is explained.Secondly,the methods of multimodal knowledge graph complementation are introduced in detail.Finally,the current challenges and application status in the field of multimodal knowledge graph complementation are summarized and outlooked.关键词
多模态知识图谱/多模态知识图谱补全/模态融合/负采样/综述Key words
multimodal knowledge graph/multimodal knowledge graph completion/modal fusion/negative sampling/review分类
信息技术与安全科学引用本文复制引用
崔好丽,赵震..多模态知识图谱补全研究综述[J].计算机技术与发展,2025,35(8):10-18,9.基金项目
国家自然科学基金项目(61976027) （61976027）
辽宁省教育厅基本科研项目(LJ212410167018) （LJ212410167018）
渤海大学2021年研究生教育教学改革项目(YJG20210022) （YJG20210022）
