Knowledge Graph Modelling



Knowledge Graphs (KGs) are directed labelled graphs where edges between nodes (entities) encode facts [1]. This kind of knowledge bases are widely used in industry (e.g. Google Knowledge Graph and Microsoft Academic Knowledge Graph) as well as in academia (e.g. Wikidata and YAGO). KGs are extremely useful to enable AI systems to reason (deductively and inductively) in various domains. However, due to their resources and the way (automatic or semi-automatic) that they are constructed they are far from complete. Thus, modelling knowledge graph to carry out different curating and maintaining tasks (e.g. KG completion and entity resolution) is essential [2]. A few learning/modelling paradigms have been considered to carry out the KG modelling task, including Transfer Learning [3], Logic-based Learning [4], Reinforcement Learning, Representation Learning, and Deep Learning [2,5].
In this project, we aim to investigate the different learning methods that have been proposed for KG modelling and propose a novel method that advances the current methods in at least one of the following performance measurements including explainability, accuracy, or efficiency. The project includes surveying the literature, implementing/deep understanding of the state-of-the-art methods and implementing the novel (to be proposed) method. 

[1] A. Hogan et al., “Knowledge Graphs,” arXiv Prepr., 2020,

[2] M. Nickel, K. Murphy, V. Tresp, and E. Gabrilovich, “A Review of Relational Machine Learning for Knowledge Graph,” in IEEE, 2016. 

[3] P. G. Omran, Z. Wang, K. Wang, P. Ghiasnezhad Omran, Z. Wang, and K. Wang, “Knowledge Graph Rule Mining via Transfer Learning,” in PAKDD, 2019.

[4] P. G. Omran, K. Wang, and Z. Wang, “An Embedding-based Approach to Rule Learning in Knowledge Graphs,” TKDE, vol. 33, pp. 1348--1359, 2021.

[5] Z. Chen, Y. Wang, B. Zhao, J. Cheng, X. Zhao, and Z. Duan, “Knowledge graph completion: A review,” IEEE Access, vol. 8, pp. 192435–192456, 2020.



Knowledge Graph, Representation Learning, Rule Learning, Reinforcement Learning

Updated:  10 August 2021/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing