Modeling Relational Data with Graph Convolutional Networks
This addresses the issue of missing facts in knowledge graphs for applications like question answering, representing an incremental advancement by adapting graph neural networks to multi-relational data.
The paper tackles the problem of incomplete knowledge graphs by introducing Relational Graph Convolutional Networks (R-GCNs) for knowledge base completion tasks, achieving a 29.8% improvement on FB15k-237 for link prediction over a baseline.
Knowledge graphs enable a wide variety of applications, including question answering and information retrieval. Despite the great effort invested in their creation and maintenance, even the largest (e.g., Yago, DBPedia or Wikidata) remain incomplete. We introduce Relational Graph Convolutional Networks (R-GCNs) and apply them to two standard knowledge base completion tasks: Link prediction (recovery of missing facts, i.e. subject-predicate-object triples) and entity classification (recovery of missing entity attributes). R-GCNs are related to a recent class of neural networks operating on graphs, and are developed specifically to deal with the highly multi-relational data characteristic of realistic knowledge bases. We demonstrate the effectiveness of R-GCNs as a stand-alone model for entity classification. We further show that factorization models for link prediction such as DistMult can be significantly improved by enriching them with an encoder model to accumulate evidence over multiple inference steps in the relational graph, demonstrating a large improvement of 29.8% on FB15k-237 over a decoder-only baseline.