Knowledge Graph
A structured representation of entities (people, places, concepts) and the relationships between them, stored as a graph for reasoning and retrieval.
What Is Knowledge Graph?
A knowledge graph is a network of interconnected facts represented as nodes (entities) and edges (relationships). For example: (Claude) —[developed by]→ (Anthropic), (Anthropic) —[founded in]→ (2021), (2021) —[is a]→ (Year). This triple-based structure allows sophisticated graph queries: "find all AI companies founded after 2020 that have released a model with a context window over 100k tokens."
In AI systems, knowledge graphs are used as a structured retrieval source alongside or instead of vector databases. Graph RAG — combining graph traversal with LLM generation — can retrieve multi-hop information that pure vector similarity search misses, because relationships between entities are explicitly encoded rather than implicitly captured in embeddings.
Why It Matters
Knowledge graphs excel at representing factual, relational knowledge where relationships between entities matter. They are particularly valuable for enterprise AI (product catalogues, organisational charts, compliance rules), scientific research (drug-protein interactions, citation networks), and recommendation systems (user-item-attribute graphs).
Real-World Examples
Google Knowledge Panel
The information box that appears when you search for a person or company on Google is powered by a knowledge graph that stores verified facts and relationships.
Drug Discovery
Pharmaceutical companies build knowledge graphs connecting drugs, proteins, diseases, and clinical trials to identify new treatment hypotheses automatically.