The School of EECS is hosting the following PhD Progress Review 1 Confirmation Seminar
Entity Alignment for Evolving Temporal Knowledge Graphs
Speaker: Jiayun Li
Host: Prof. Xue Li
Abstract: Entity Alignment (EA) is crucial for integrating heterogeneous knowledge graphs (KGs) into a unified knowledge base by identifying equivalent entities across them. Traditional models focus mainly on structural characteristics, utilizing entity embeddings for alignment. However, they often neglect the distinct temporal properties and challenges of heterogeneity in Temporal Knowledge Graphs (TKGs). We enhance the embedding approach to better utilize temporal data by initializing dual temporal adjacency matrices using temporal statistics and employing a temporal attention mechanism to optimize weight distribution among neighbouring temporal entities. Our approach also manages temporal heterogeneity by leveraging intermediate model outputs to adapt to varying temporal intervals. Extensive evaluations demonstrate our method's effectiveness in improving the use of temporal information for entity alignment.
Bio: Jiayun Li is a second-year Ph.D. student at the University of Queensland, Australia. Prior to his candidacy, he earned a Bachelor of Engineering from Tianjin University and Master of data science from UQ. Jiayun is currently working on the research topic of temporal entity alignment, advised by Prof. Xue Li and A/Prof. Wen Hua.
About Data Science Seminar
This seminar series is hosted by EECS Data Science.