Many large-scale knowledge bases simultaneously represent two views of knowledge graphs (KGs): an ontology view for abstract and commonsense concepts, and an instance view for specific entities that are instantiated from ontological concepts. Existing KG embedding models merely focus on representing one of the two views alone. However, simultaneous learning from both views will likely produce better knowledge embedding models and enable new applications that rely on multi-view knowledge. 
In this seminar, we introduce our recent research on a novel two-view KG embedding model, JOIE. It employs both cross-view and intra-view models that learn on multiple facets of the knowledge base. The cross-view association model is learned to bridge between the embeddings of ontological concepts and their corresponding instance-view entities. The intra-view models are trained to capture the structured knowledge of instance and ontology views in separate embedding spaces. Experimental results show JOIE can perform better on knowledge graph completion and entity typing tasks.
In the second part of this talk, we will introduce one application of the two-view product graph on product complementary recommendation in e-commerce. By using similar techniques, we can make a high-quality and diverse product recommendation for online shopping customers to purchase together.