Knowledge Graph

A knowledge graph is a structured representation of facts as a network of entities (nodes) and the relations between them (edges). The term entered wide usage when Google deployed its Knowledge Graph in 2012 to enrich search results with semantic information drawn from structured databases. In academic contexts, knowledge graphs are studied as instances of formal ontologies — explicit specifications of what kinds of entities a domain recognizes and what kinds of relations hold between them.

The structural properties of knowledge graphs are a branch of graph theory applied to epistemology. A concept with many incoming edges — referenced by many other nodes — is foundational in the sense that many claims depend on it. A concept with many outgoing edges — that itself references many others — is synthetic in the sense that it integrates disparate claims. The balance between foundational depth and synthetic breadth is a topological property of the knowledge structure, not merely a logical one.

Knowledge graphs have become important in artificial intelligence for knowledge representation and question answering. Their relationship to the distributed representations of large language models remains an open problem in cognitive architecture: whether neural language models implicitly learn graph-like structures, or whether they represent something fundamentally different, is contested.

The most significant knowledge graph is, arguably, mathematics itself — a graph so dense with interdependencies that its foundational structure took centuries to make explicit.