Knowledge graph
A knowledge graph represents entities and the relationships between them as nodes and edges, so a system can query how things connect.
Category
Industry concept
See all terms
What is a knowledge graph?
A knowledge graph stores information as entities and the relationships between them, nodes for things and edges for how those things connect. Instead of a document that only mentions a person and a company, a knowledge graph records the person as a node, the company as a node, and "works at" as the edge between them, so the relationship becomes something a system can query directly.
This solves a limit of plain text: a document holds facts, but the connections between those facts stay implicit, left for a reader to infer. A knowledge graph makes those connections explicit, so a system can move from one entity to related ones without re-reading the surrounding prose each time.
How does a knowledge graph differ from a context repository?
A knowledge graph is a data structure: entities and relationships, typically without the surrounding document, permissions, or version history attached to that fact. A context repository holds both, actual files with interconnected references between them, plus who can access each file and how it has changed over time. The graph tells you how things relate, the repository is the governed source those relationships get drawn from.
How does a knowledge graph work?
Building a knowledge graph starts with extracting entities and relationships from source material, then storing them as triples: a subject, a relationship, and an object, such as "Mount Fuji, isLocatedIn, Japan." A query engine can then traverse these connections, finding every deal tied to a company tied to a given industry in one pass, instead of searching separate documents and manually joining what it finds.
Qontext's context repository works this way: files reference other files directly, so an agent can directly follow a connection, from a person to the deals and conversations tied to them.