SUKE: Embedding Model for Prediction in Uncertain Knowledge Graph

Jingbin Wang, Kuan Nie, Xinyuan Chen, Jing Lei
2020 IEEE Access  
Graph embedding models are widely used in knowledge graph completion (KGC) task. However, most models are based on the assumption that knowledge is completely certain, and this is inconsistent with real-world situations. Although there are multiple studies on uncertain knowledge embedding tasks, they often use knowledge confidence to learn embedding and cannot make full use the structural and uncertain information of knowledge. This paper presents a new embedding model named Structural and
more » ... tain Knowledge Embedding (SUKE), which comprises two components: an evaluator and a confidence generator. For unknown triples, the evaluator learns the structural and uncertain information to evaluate its rationality and obtain a candidate set. The confidence generator then determines the confidence of the candidate set to achieve KGC. To verify the effectiveness of the proposed model, confidence prediction, triple evaluation, and fact classification tasks are performed on three data sets. Experimental results show that SUKE performs better than mainstream embedding methods. The model proposed in this paper can help advance the research on the embedding of uncertain knowledge graphs. INDEX TERMS Artificial intelligence, knowledge representation, uncertain knowledge graph. VOLUME 9, 2021 This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
doi:10.1109/access.2020.3047086 fatcat:4nlgpj524bedtkizz2oom4hpzm