Knowledge Graph
person concept tool org talk claim — click a node to jump to its page; hover an arrow for the relation
A Knowledge Graph is a structured representation of entities and their relationships to other entities, typically implemented using graph databases and knowledge representation techniques. As described by Frank Coyle (UC Berkeley), they serve as a concrete realization of Ontology|ontological structures in modern AI and data systems.
Definition and Structure
Coyle defines knowledge graphs succinctly: "They're basically a representation of entities and their relationships to other entities" — tying them directly to the ontological framework he advocates for agentic systems. → This makes knowledge graphs the practical, graph-database instantiation of an ontology: where an ontology provides the schema or conceptual model, a knowledge graph populates it with actual entity data and relational assertions.
Role in Agentic Systems
Coyle situates knowledge graphs within the contemporary landscape of graph databases and knowledge representation, noting that this kind of structured relational modeling "relates to what people are doing now" in the field. → The implication is that knowledge graphs are the modern, scalable answer to problems of knowledge representation that Agentic Systems need to reason about their environments, goals, and constraints.
Relationship to Symbolic AI
Coyle positions knowledge graphs as an alternative to Symbolic AI — rather than encoding knowledge through rigid logical rules and symbol manipulation as classical symbolic systems did, knowledge graphs offer a more flexible, graph-structured approach to representing the same kinds of relational information. → This framing suggests knowledge graphs occupy a middle ground: more structured than raw neural representations, but more scalable and flexible than traditional symbolic systems.
Relationship to Ontology
The ontology defines the knowledge graph — providing the type system, relationship vocabulary, and constraints that give a knowledge graph its semantic coherence. → Without an ontological backbone, a knowledge graph is merely a property graph; with one, it becomes a semantically grounded knowledge representation capable of supporting inference and interoperability across agentic components.