Convolutional 2D Knowledge Graph Embeddings

In this work, we introduce a convolutional neural network model, ConvE, for the task of link prediction where we apply 2D convolution directly on embeddings, thus inducing spatial structure in embedding space. To scale to large knowledge graphs and prevent overfitting due to over-parametrization, previous work seeks to reduce parameters by performing simple transformations in embedding space. We take inspiration from computer vision, where convolution is able to learn multiple layers of non-linear features while reducing the number of parameters through weight sharing. Applied naively, convolutional models for link prediction are computationally costly. However, by predicting all links simultaneously we improve test time performance by more than 300x on FB15k. We report state-of-the-art results for numerous previously introduced link prediction benchmarks, but concerns have been raised regarding the quality of the well-established FB15k and WN18 datasets. Previous work noted that these datasets contain many reversible triples, but the severity of this issue was not quantified. To investigate this, we design a simple model that uses a single rule which reverses relations and achieves state-of-the-art results. We introduce WN18RR, a subset of WN18 which was constructed the same way as the previously proposed FB15k-237, to alleviate this problem and report results for our own and previously proposed models for all datasets. Analysis of our convolutional model suggests that it is particularly good at modelling nodes with high indegree and nodes with high PageRank and that 2D convolution applied on embeddings seems to induce contrasting pixel-level structures.