Knowledge graph embedding via entity and relationship attributes

The translation rule-based TransE model is considered the most promising method due to its low complexity and high computational efficiency. However, there are limitations in dealing with complex relationships such as reflexive, 1-to-N, N-to-1, and N-to-N. Therefore, we propose a knowledge graph embedding model TransP based on entity and relationship attributes. We introduce the idea of hyperplane projection to map the head entity and tail entity to the plane of a specific relationship to enhance the model's ability to handle complex relationships. Furthermore, we propose the strategy of using the attribute characteristics of entities and relationships to improve distinction between different entities or relationships. Finally, we conduct link prediction and triple classification experiments on WN11, WN18, FB13 and FB15K datasets. Experimental results verify that the proposed method outperforms the baseline models and achieves the best results.