Knowledge graph concepts are a small set of ideas that every other topic on this site depends on: what an entity is, how a triple states a fact, how triples form a graph, how an ontology gives the graph meaning, and how graphs relate to the Semantic Web, to databases, and to the vector representations used in machine learning. This hub arranges those ideas in the order they build on one another. Read it top to bottom as a course, or jump to the comparison that answers the question you arrived with.
The learning path
Each step below introduces one idea and assumes only the steps before it.
1. What is an entity? Start with the unit everything else is made of. An entity is a uniquely identifiable thing or concept, independent of the words used to name it. The page covers identifiers (IRIs, Wikidata Q-IDs, Google machine IDs), the difference between an entity and a keyword, and why disambiguation is the central problem.
2. RDF triples: subject, predicate, object A fact about an entity is written as a triple. This page shows how subjects, predicates, and objects are formed, how literals carry datatypes and language tags, and how the same triple looks in Turtle, N-Triples, and JSON-LD.
3. What is a knowledge graph? Many triples about many entities form a graph. The pillar page defines the term, walks through a worked example, explains how graphs are built and enriched, and surveys public and enterprise examples.
4. Knowledge graph examples and use cases Concrete instances: open graphs such as Wikidata and DBpedia, the search-engine graphs at Google and Microsoft, consumer recommendation graphs, and domain graphs in biomedicine and finance, plus a small graph a reader could build.
5. Ontologies The graph needs a vocabulary. This page explains classes, properties, hierarchies, and constraints, distinguishes a taxonomy from a thesaurus from an ontology, introduces RDFS, OWL, and SKOS, and shows how a subclass declaration lets a reasoner infer new facts.
6. Linked data Identifiers should be resolvable and graphs should connect to one another. The four Linked Data principles and the 5-star scheme explain how, and the page shows why a sameAs link on an ordinary website is linked data in practice.
7. The Semantic Web The historical and standards context: the W3C stack from IRIs through RDF, RDFS, OWL, SPARQL, and SHACL, what parts of the original vision succeeded, and how the term relates to “knowledge graph.”
8. Entity extraction, NER, and entity linking How graphs are populated from text: mention detection, named entity recognition, disambiguation, linking to identifiers, and relation extraction, with the tools used at each step.
9. Knowledge graph embeddings How graphs become vectors: translational and bilinear models, link prediction, evaluation metrics, and the libraries that implement them. This is the bridge from symbolic graphs to machine learning.
Comparisons
These pages answer the “X vs Y” questions that appear most often in search. Each has a comparison table and a section on when to use which.
| Page | The one-line answer |
|---|---|
| Knowledge graph vs ontology | An ontology is the vocabulary and rules; a knowledge graph is the instance data organized by them. |
| Knowledge graph vs graph database | A graph database is storage and query technology; a knowledge graph is the data-plus-semantics artifact that may live in one. |
| Knowledge graph vs vector database | Vectors find things that are similar; graphs find things that are connected. Retrieval systems increasingly use both. |
| Property graph vs RDF (and Cypher vs SPARQL) | Two data models for the same graphs: labeled property graphs favor developer ergonomics, RDF favors standards and interoperability. |
A note on terminology
The field has accumulated several names for overlapping ideas. “Semantic Web” describes the W3C standards program that began around 2001. “Linked data” describes the publishing practice that grew out of it. “Knowledge graph” became the common term after Google adopted it in 2012 and is now used for any large graph of typed entities, whether or not it uses W3C standards. “Ontology,” “schema,” “vocabulary,” and “taxonomy” are related but not interchangeable; the ontologies page draws the lines. The glossary holds short definitions for all of these.
For developers
The concept pages deliberately stay above any single product. Once the ideas are clear, the Languages & Formats section covers syntax (RDF, JSON-LD, SPARQL, OWL, SHACL), and the Building Knowledge Graphs section covers tooling, graph databases, and GraphRAG.
For SEOs
The same concepts underlie every structured data decision. Entities and identifiers explain why sameAs matters. Triples explain what a schema.org block actually asserts. Ontologies explain what schema.org is. The Knowledge Graphs for SEO section applies each idea, starting with entity SEO and the Google Knowledge Graph.
Frequently asked questions
In what order should I learn knowledge graph concepts?
Start with entities, then triples, then the knowledge graph itself, then ontologies and linked data. The Semantic Web page supplies historical context once the pieces are familiar. Entity extraction and embeddings come last because they describe how graphs are populated and used, which is easier to follow once the structure is clear.
Do I need to know RDF to understand knowledge graphs?
No. RDF is one data model for knowledge graphs, and labeled property graphs are another. The concept pages explain ideas in plain terms first and use RDF syntax only in worked examples, because it is the most compact way to show a fact. The property graph vs RDF page covers the alternative model.
What is the difference between a knowledge graph and a knowledge base?
The terms overlap. “Knowledge base” is the older, broader term for any structured store of facts and rules, including those built for expert systems. “Knowledge graph” specifically describes a knowledge base organized as a graph of entities and relationships. Most modern knowledge bases, such as Wikidata, are knowledge graphs.
Sources and further reading
- Hogan, A. et al. Knowledge Graphs. Springer, 2021. Free edition: https://kgbook.org/
- W3C. RDF 1.1 Primer. https://www.w3.org/TR/rdf11-primer/
- W3C. Semantic Web standards overview. https://www.w3.org/standards/semanticweb/
- Google. Introducing the Knowledge Graph: things, not strings (2012). https://blog.google/products/search/introducing-knowledge-graph-things-not/
- Stanford CS520: Knowledge Graphs course notes. https://web.stanford.edu/class/cs520/