Path-RotatE: Knowledge Graph Embedding by Relational Rotation of Path in Complex Space

Abstract

We study the problem of learning knowledge representations of entities and relations in knowledge graphs to predict missing links. The key to precisely accomplish a such task is modeling and inferring the diverse patterns of the relations. In this paper, we present a new rotation-based knowledge representation learning model named Path-RotatE, which considers additional paths to model rich inference patterns between entities. In addition, this paper considers the correlation between the path and the direct relation. In this way, we improve reliability of the path, making it more suitable to train. Finally, this paper conducts entity prediction experiments on datasets such as FB15k, FB15-237, WN18 and WN18RR. The results show that the Path-RotatE model has a certain improvement in MR, MRR and Hits@N compared to RotatE, PTransE and other baseline models.