Knowledge graphs represent structural relations between entities as a central research problem in artificial intelligence.

A knowledge graph formalizes entities and their relations so that representation learning, acquisition, completion, temporal reasoning, and logical inference can operate over connected structure. The survey by Ji and colleagues classifies embedding methods by representation space, scoring function, encoding model, and auxiliary information. This framework explains why graph models can preserve non-linear context: paths and relation types are computational objects, not incidental links reconstructed from unrelated tables.