As a branch of knowledge graphs, temporal knowledge graphs are designed to depict facts with time features, addressing the limitations of traditional static knowledge graphs in describing the dynamism of entities and relationships in the real world. Recently, Big Data Department proposed a temporal knowledge graph reasoning method based on temporal historical subgraphs. By integrating temporal information into the constructed unified graph, this approach models the overall temporal features of historical subgraph sequences to predict and represent future factual occurrences. Experimental results demonstrate outstanding performance of the model in transductive settings, meanwhile supporting inductive reasoning and providing further interpretable reasoning grounds.

This research has been accepted by Artificial Intelligence (AIJ, CCF A). The first author is Ph.D. student Hao Dong from Big Data Department and supervised by Prof. Yuanchun Zhou.

This research is supported by National Natural Science Foundation of China.

Hao Dong, Pengyang Wang, Meng Xiao, Zhiyuan Ning, Pengfei Wang, Yuanchun Zhou. "Temporal Inductive Path Neural Network for Temporal Knowledge Graph Reasoning." Artificial Intelligence. 2024. 

Paper Link: https://doi.org/10.1016/j.artint.2024.104085