What is a knowledge graph?

Direct answer

A knowledge graph is a structured representation of knowledge where entities are connected by explicit, typed relationships.

Entities are things that exist: people, organisations, products, materials, concepts, or locations. Relationships describe how entities connect to each other, such as produces, used in, competes with, or located in.

Unlike documents or databases, knowledge graphs encode meaning in a form machines can traverse and reason over.

Core components

Knowledge graphs consist of two primitives:

Entities: Discrete things that exist in a domain. Each entity has a unique identifier and can carry attributes (properties that describe it).

Relationships: Typed connections between entities. The relationship type defines the nature of the connection and carries semantic meaning that machines can interpret.

These are typically represented as triples: subject-predicate-object statements such as Victrex → produces → PEEK polymer.

Simple example

Victrex → produces → PEEK polymer
PEEK polymer → used in → medical implants
Victrex → competes with → Solvay

Three entities, two relationship types, six discrete facts that machines can query and traverse.

What it is not

A knowledge graph is not:

• A database (stores records, not meaning)
• A content library (content is unstructured, graphs are explicit)
• A taxonomy (taxonomies categorise, graphs relate)
• A search optimisation technique (graphs exist independently of search engines)