Concept
Knowledge Graph
A connected representation of entities, relationships, context, and evidence for reasoning and discovery.
Position in the Trusted Data Framework
Knowledge Graph
Living Graph View
Knowledge Graph in the concept network
Drag nodes to adjust the view, double-click a node to expand its neighbourhood, and right-click a node to hide it while preserving the exploration context.
Relationship Matrix
Knowledge Graph as a network node
Ontology defines the graph schema, including entity classes, relationship types, and valid patterns of meaning.
The semantic layer supplies governed definitions and mappings that make graph data consistent across sources.
Reality Mapping anchors graph nodes to real-world objects, events, and responsibilities.
A knowledge graph gives AI systems structured context, traceable relationships, and paths for explanation.
Knowledge graphs connect evidence, dependencies, and consequences so decisions can be evaluated in context.
Concept Relationships
How Knowledge Graph works with other concepts
Ontology defines the graph schema, including entity classes, relationship types, and valid patterns of meaning.
An infrastructure graph uses the ontology to distinguish Asset, Component, Location, Sensor, Event, and Control.
What schema is needed before graph connections can be trusted?
The semantic layer supplies governed definitions and mappings that make graph data consistent across sources.
Multiple source-system customer identifiers are mapped into one governed Customer concept before entering the graph.
Which semantic mappings are needed to prevent duplicate or conflicting graph nodes?
Reality Mapping anchors graph nodes to real-world objects, events, and responsibilities.
A physical asset, its digital twin, maintenance record, inspection event, and risk decision become connected evidence.
Which graph nodes represent real things, and what evidence proves that mapping?
A knowledge graph gives AI systems structured context, traceable relationships, and paths for explanation.
A risk assistant can explain which supplier, component, location, and control evidence influenced a recommendation.
What connected context must AI retrieve before making or explaining a decision?
Knowledge graphs connect evidence, dependencies, and consequences so decisions can be evaluated in context.
A capital planning decision can trace affected assets, service risks, dependencies, and previous interventions.
Which decisions require relationship-based evidence rather than flat reports?
Definition
A knowledge graph represents domain knowledge as connected entities and relationships. It combines data, meaning, context, and evidence in a structure that people and machines can navigate.
Why It Matters
Complex organisations rarely need one more isolated dataset. They need connected knowledge that can answer relationship-based questions and support explainable AI.
Role in the Trusted Data Framework
Knowledge graphs sit in the Knowledge layer. They turn semantic models and trusted data into a navigable structure for reasoning, discovery, and AI augmentation.
Practical Examples
- Connecting assets, locations, events, documents, risks, and decisions.
- Supporting retrieval-augmented generation with governed context.
- Mapping dependencies across infrastructure, suppliers, systems, and controls.
Case Evidence
GBlocks deliberately does not build a knowledge
graph, matching this framework's own guidance not to make the graph the
centre of the architecture prematurely. But the relationships a graph
would formalise already exist as materialised, queryable traversal
views, and one endpoint (GET /scene/uprn/{uprn}) assembles identity,
every source's representation, the full boundary hierarchy, an event
timeline, and per-link confidence for a single property in one call --
exactly the question a property-centred graph node would answer,
answered today without graph infrastructure. This is itself a useful
answer to the Knowledge Graph Readiness Accelerator's central question:
not yet, because the questions asked so far are one-hop. A graph would
earn its complexity the moment a question becomes genuinely multi-hop
-- e.g. "every property within 500m of an Outstanding-rated school,
built before 1950, outside any flood zone, sold in the last two years."
A concrete example of the gap a graph's entity-resolution layer exists to close showed up later in the same case: a new schools dataset from a second government department (DfE) turned out to describe the same real-world entity type as a schools dataset already in the model (Ofsted), using a different identifier scheme, with no link between them. That's recorded today as an honest open gap rather than presented as two unrelated tables -- exactly the kind of duplication a knowledge graph's entity-resolution step exists to close, and exactly the kind of finding a Knowledge Graph Readiness engagement should surface before recommending graph infrastructure to fix it.
Consulting Questions
- Which entity relationships matter most for decisions?
- What evidence should be attached to claims and recommendations?
- Which graph questions cannot be answered by current data platforms?
- How will the graph remain aligned with source systems?
- Which users need graph navigation, analytics, or AI access?