KNOWLEDGE DOSSIER · Données structurées

Knowledge graphs and JSON-LD

A knowledge graph explicitly represents entities and their relationships. On the Web, vocabularies, identifiers and linked-data formats support exchange between systems. A graph is not a mechanism that automatically guarantees its relationships are true.

Edition: 13 November 2026 Extended dossier Scientific review pending
AT A GLANCE
TypeKnowledge graphs and the semantic Web
AudienceArchitects, information specialists and researchers
01

Entities, identifiers and relationships

Resources may be people, publications, museums or artworks. Graphs connect them through identified properties. Stable URIs reduce confusion between identically named things but do not replace checking attribution and the actual relationship.

02

RDF, JSON-LD and context

RDF describes subject–predicate–object triples. JSON-LD expresses linked data in JSON syntax using contexts that define term interpretation. A syntactically valid JSON-LD document may still claim a false relationship: structural validation and factual review remain distinct.

03

Align vocabularies

Systems may use different properties for similar concepts. Alignment requires precise definitions and context, especially for multilingual or heterogeneous collections. Equivalence should be justified rather than assumed.

04

Search and document assistants

A portal may use a graph to navigate between authors, works and institutions or enrich RAG retrieval. Generated answers still need source evidence and a distinction between graph traversal and model interpretation.

05

Quality, provenance and updates

Store source, date, version and possibly uncertainty for important relationships. Graphs can accumulate obsolete identifiers, contradictory relations and duplicates. Documented detection and correction processes improve usefulness.

06

Exercise: a museum collection

Describe three works with creators, dates, locations and licenses using stable identifiers. Confirm information against original catalogs. Then offer navigation by artist and institution without inventing unsupported relationships.

HOW IT FITS TOGETHER

Core steps

  1. 01Define resources and their URIs
  2. 02Choose documented properties
  3. 03Create a small JSON-LD graph
  4. 04Validate each relation against source
  5. 05Check duplicates and contradictions
  6. 06Test navigation and maintenance
CRITICAL THINKING

Three questions to ask

  • Is JSON-LD automatically true?

  • Do identifiers distinguish namesakes?

  • Is every relation sourced?

  • Does the relation remain valid after updates?

PRIMARY SOURCES

Read the source documents

Versions, reuse terms and document status can change. Check them on the original sites.