Knowledge graph
A way of storing information as a network of things and the relationships between them: a customer, the products they bought, the rep who handles them, drawn as connected dots and lines. Each dot is a thing, each line is a named relationship. AI tools use one to answer from checked facts instead of guessing from text.
In plain English
Picture a corkboard with photos and string. Every photo is a thing: a client, a product, a person. Every string is a relationship: bought, manages, supplies. That board, stored so software can follow the strings, is a knowledge graph.2
In the real world
- The box on the right of a Google search for a famous person, with birth date, spouse, and films. Search engines have run on knowledge graphs for years.1
- A customer database (CRM) that can answer “which clients share a supplier with us.”
- You don’t buy one. You meet one inside tools you already use, or in an AI vendor’s pitch.
- The “related” links on a product page that actually make sense.
When to use it
- Your questions are about relationships. Who knows whom, what depends on what, what touched what.
- You want AI answers based on facts you control, not on what the model happens to remember. See hallucination.
When not to use it
- Your data is a handful of documents. A project with the files in it does the job.
- Nobody owns the upkeep. A graph that isn’t maintained becomes a confident map of last year.
Good, bad, ugly
- Good: answers come with a trail. You can see which facts and relationships the answer used.
- Bad: building one is real work. Someone decides what counts as a thing and what counts as a relationship.
- Ugly: “we need a knowledge graph” said about a problem a spreadsheet filter would solve.
Key facts
- Made of
- Dots for things, lines for relationships, and a label on each saying what kind
- Classic example
- The info box Google shows when you search a famous person or place
- Why AI likes it
- A fact stored as a clear link is harder to misread than one buried in a paragraph
Common questions
How is this different from a database?
A normal database is a set of spreadsheets that reference each other. A knowledge graph stores the connections themselves, so 'who else works with this client's supplier' is one step, not a hunt across five sheets.
Do I need one?
Only if your questions are mostly about how things relate. If your data is a few spreadsheets, a project with the files in it is enough.
Is this the same as a context graph?
No. A context graph builds on a knowledge graph by adding decisions, policies, and when each fact was true. See the context graph entry.
References
- Knowledge graphWikipediaView source
- What is a knowledge graph?IBMView source