CombinE: A Fusion Method Enhanced Model for Knowledge Graph Completion

Knowledge Graph Completion (KGC) task aims at predicting absent links between entities by leveraging existing information. Conventional neural network (CNN)-based KGC models have gained substantial attention due to their efficacy and computational efficiency. The effectiveness of these models greatly hinges on the fusion methods employed for combining entity and relation embeddings. In this paper, we proposed a novel approach for fusing entity and relation embeddings, ensuring a evener combination of elements in terms of indices. This technique serves as the foundation for a new CNN-based KGC model called CombinE. Experimental results on five datasets show that CombinE achieves better performance than the baseline models.