Knowledge Graph
Also called Knowledge Graph
Facts stored as connected dots — things, and the labeled relationships between them.
Think of it like
A detective’s corkboard: photos pinned up with string labeled "works for" or "married to," so you can trace who connects to whom.
Example
"Marie Curie — won — Nobel Prize" and "Nobel Prize — awarded in — Stockholm" let a system answer "where was Curie’s prize awarded?" by walking the links.
How it actually works
A knowledge graph stores information as triples — subject, relationship, object — instead of loose text. That structure lets you answer multi-hop questions by traversing edges, something pure vector search struggles with. Paired with retrieval (sometimes called GraphRAG), it grounds answers in explicit, checkable relationships. The catch is building and maintaining the graph, which is real ongoing work.
For product teams
Unlocks precise, multi-step questions ("which of our customers use a competitor?") that flat search fumbles.
For engineers
Store facts as (subject, predicate, object) triples; answer by graph traversal, optionally feeding retrieved subgraphs into an LLM.
Related
- Entity Linking — Populated by resolving mentions to real entities.
- Semantic Search — A structured alternative to vector-only search.
- Grounding — Can supply grounding facts to a model.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome