Learning Knowledge Graph Embeddings with Type Regularizer

摘要: Learning relations based on evidence from knowledge repositories rely processing the available relation instances. Many relations, however, have clear domain and range, which we hypothesize could help learn a better, more generalizing, model. We include such information in RESCAL model form of regularization factor added to loss function that takes into account types (categories) entities appear as arguments base. Tested Freebase, frequently used benchmarking dataset for link/path predicting tasks, note increased performance compared baseline terms mean reciprocal rank hits@N, N = 1, 3, 10. Furthermore, discover scenarios significantly impact effectiveness type regularizer.