QLogicE: Quantum Logic Empowered Embedding for Knowledge Graph Completion
作者：
Highlights：
• A combining model for knowledge graph completion is proposed and it achieves surprising performance results.
• The embedding dimension of the proposed model is only 8 which means very low computing cost.
• The low embedding dimension is even lower than the lower-bound of the baseline model and inspired us the idea of Dense Feature Model.
• Research results show that neural networks may be not the best model for KGC task and the logical rules based one may be a good alternative.
• The metric Hits@1 is up to 94.84% on dataset FB15k237 seems get very close to the solutions of KGC problem.
摘要
•A combining model for knowledge graph completion is proposed and it achieves surprising performance results.•The embedding dimension of the proposed model is only 8 which means very low computing cost.•The low embedding dimension is even lower than the lower-bound of the baseline model and inspired us the idea of Dense Feature Model.•Research results show that neural networks may be not the best model for KGC task and the logical rules based one may be a good alternative.•The metric Hits@1 is up to 94.84% on dataset FB15k237 seems get very close to the solutions of KGC problem.
论文关键词：Quantum logic,Knowledge graph,Knowledge graph completion,Link prediction,Combination model
论文评审过程：Received 14 July 2021, Revised 8 December 2021, Accepted 11 December 2021, Available online 17 December 2021, Version of Record 10 January 2022.
论文官网地址：https://doi.org/10.1016/j.knosys.2021.107963
