17th International Conference on Principles of Knowledge Representation and Reasoning

Knowledge Graph Inference using Tensor Embedding

, , , and

Axiom based inference provides a clear and consistent way of reasoning to add more information to a knowledge graph. However, constructing a set of axioms is expensive as it requires domain expertise, time, and money. It is also difficult to use or adapt the same set of axioms to a knowledge graph in a different domain. We present a family of four novel methods for embedding knowledge graphs into real-valued tensors that capture the ordered relations found in simple knowledge graphs like RDF. We demonstrate our models on the task of predicting new facts on eight different knowledge graphs, achieving a 5% to 50% improvement over existing systems. Through experiments, we derived recommendations for selecting the best model based on knowledge graph characteristics. We also give a provably-convergent, linear tensor factorization algorithm.

graph embeddings, knowledge graph

InProceedings

AAAI Press

Recently Published Research Track

UMBC ebiquity