无线电通信技术2025，Vol.51Issue(3)：440-446,7.DOI:10.3969/j.issn.1003-3114.2025.03.002
基于动态超图嵌入的时域知识图谱推理模型
Temporal Knowledge Graph Reasoning Model Based on Dynamic Hyper-graph Embedding
摘要
Abstract
Reasoning over the Temporal Knowledge Graph(TKG)for predicting future facts has received much attention.Most research works attempt to model temporal dynamic using knowledge graphs and Graph Convolutional Network(GCN).How-ever,these methods fail to capture high-order interactions between objects in the TKG,which is an important factor for predicting future facts.To handle these problems,a TKG reasoning model based on dynamic hyper-graph embedding is proposed.High-or-der interactions are obtained by constructing hyper-graphs based on the TKG at different timestamps.In addition,the difference brought by timestamps is integrated into the hyper-graph representation to better adapt to the TKG.Then,dynamic meta-embed-ding is adopted for the representation of the temporal hyper-graph,allowing the model to select appropriate high-order interactions for downstream reasoning.Experiments are conducted on public TKG datasets,and the results show that the proposed model out-performs other baseline models.Moreover,the analysis demonstrates that the proposed method brings good interpretability to the predicted results.关键词
时域知识图谱/图卷积网络/动态超图嵌入/动态元嵌入/高阶交互Key words
TKG/GCN/dynamic hyper-graph embedding/dynamic meta-embedding/high-order interactions分类
信息技术与安全科学引用本文复制引用
高秀东,陈阳,张静..基于动态超图嵌入的时域知识图谱推理模型[J].无线电通信技术,2025,51(3):440-446,7.基金项目
四川省科技计划资助(2025YFHZ0265) Sichuan Science and Technology Program(2025YFHZ0265) （2025YFHZ0265）
