A Multimodal Translation-Based Approach for Knowledge Graph Representation Learning

25534

Knowledge Triples