The paper introduces Relational Graph Convolutional Networks (R-GCNs) and applies them to two standard knowledge base completion tasks: link prediction and entity classification. R-GCNs are designed to handle highly multi-relational data, which is common in realistic knowledge bases. The authors demonstrate the effectiveness of R-GCNs as a standalone model for entity classification and show that factorization models for link prediction, such as DistMult, can be significantly improved by using an R-GCN encoder to accumulate evidence over multiple inference steps in the graph. The results show a 29.8% improvement on the FB15k-237 dataset over a decoder-only baseline. The paper also discusses techniques for parameter sharing and enforcing sparsity constraints to apply R-GCNs to multigraphs with large numbers of relations.The paper introduces Relational Graph Convolutional Networks (R-GCNs) and applies them to two standard knowledge base completion tasks: link prediction and entity classification. R-GCNs are designed to handle highly multi-relational data, which is common in realistic knowledge bases. The authors demonstrate the effectiveness of R-GCNs as a standalone model for entity classification and show that factorization models for link prediction, such as DistMult, can be significantly improved by using an R-GCN encoder to accumulate evidence over multiple inference steps in the graph. The results show a 29.8% improvement on the FB15k-237 dataset over a decoder-only baseline. The paper also discusses techniques for parameter sharing and enforcing sparsity constraints to apply R-GCNs to multigraphs with large numbers of relations.