Knowledge Graph Courses & Tutorials (Free and Paid)

Q: Is Stanford CS 520 still available?

The course site is still online with slides, lecture notes and YouTube recordings from the Spring 2021 edition, and the Spring 2020 edition is archived alongside it. It is not an enrollable course and there is no assessment. The content predates current retrieval-augmented generation work, so it covers data models, acquisition and inference rather than LLM integration.

Structured knowledge graph courses are scarcer than blog posts about knowledge graphs, and the good ones cluster around three sources: university material left online, vendor academies attached to a product, and community-maintained tutorials from Wikidata and the Apache projects. This page lists material confirmed to be online, grouped by what a learner is trying to do, with a note on which entries are dated and a section on what none of them teach. A course that runs on a schedule can close and an archived one can be taken down, so check the provider’s page before planning around it.

Semantic web and RDF fundamentals

CS 520: Knowledge Graphs (Stanford University). Organized by Vinay K. Chaudhri, Naren Chittar and Michael Genesereth, this was a seminar rather than a lecture series, built around invited talks from people running production graphs in industry and research. The Spring 2021 site is still up with slides, HTML lecture notes and YouTube recordings, and the Spring 2020 edition is archived alongside it. The format is self-study, with no assessment. It suits someone who already knows what a graph is and wants to hear how the problem gets framed. It is dated: the sessions predate the current wave of retrieval-augmented generation work, so nothing in it covers LLM integration. https://web.stanford.edu/class/cs520/

Knowledge Graphs (openHPI, Hasso Plattner Institute). Taught by Harald Sack and Mehwish Alam, this ran in late 2020 and has been in self-study mode since. It is a proper MOOC: video units, quizzes, and a structured progression through RDF, RDFS, OWL, SPARQL, graph embeddings, and applications. Graded assignments aren’t available in self-study mode, so the only outcome is a certificate of participation. It suits learners who want a sequenced curriculum rather than a set of talks, and it is the closest thing to a free, full university course on the topic. https://open.hpi.de/courses/knowledgegraphs2020

Semantic Web Technologies (openHPI). Also taught by Harald Sack, available in self-study mode since 2013. The RDF and OWL core it teaches is still accurate, and its treatment of Linked Data principles is careful, but it is thirteen years old, and the examples and tooling show it. Worth it for grounding, not for anything practical about current stacks. https://open.hpi.de/courses/semanticweb

The W3C’s own primers are free self-study material and cover the same ground with more precision, which is the argument for reading them alongside any of the above. They are indexed in specs and standards. For the concepts these courses assume, start with the “What is a knowledge graph” and “linked data” pages.

SPARQL

Wikidata:SPARQL tutorial (Wikidata community). A hands-on tutorial written against a live endpoint, so every example runs in the Wikidata Query Service without setup. It works up from a single triple pattern to the constructs a practitioner needs daily. A companion page, “A gentle introduction to the Wikidata Query Service”, covers the same ground more slowly. It suits anyone who wants to query within an hour, and it is the fastest route into SPARQL. https://www.wikidata.org/wiki/Wikidata:SPARQL_tutorial

Apache Jena SPARQL Tutorial (Apache Software Foundation). The vendor-neutral counterpart: short pages on basic patterns, filters, OPTIONAL, UNION, named graphs and result formats, with local data files rather than a hosted endpoint. It is written by the people who implement the query engine, which shows in the accuracy. It suits a developer who will run SPARQL against their own store. https://jena.apache.org/tutorials/sparql.html

Introduction to Wikidata (Library Carpentry). A lesson in the Carpentries format, released in 2026 and still marked beta, covering what Wikidata is, how items and statements are structured, how to edit and how to query across five episodes. It is designed for a workshop but readable solo, and each episode carries explicit objectives and exercises. It suits librarians, researchers, and anyone whose interest in graphs starts with a collection to describe rather than an application to build. https://librarycarpentry.github.io/lc-wikidata/

Wikidata:Tours is a set of short interactive walkthroughs inside the Wikidata interface itself, covering items, statements and references. It is an orientation rather than a course, and the right first step before either tutorial above. For deeper coverage, see Wikidata and the Wikidata Query Service; for query syntax, see SPARQL.

Ontology engineering

Protégé Short Course (Stanford University). Taught by members of the Protégé team, this is a two-and-a-half-day hands-on course in ontology development, OWL 2 and the Protégé toolset, run in person at Stanford, with no prior Protégé experience required. It is instructor-led and paid rather than free; dates are announced on the course page. It suits someone who has to model a domain properly and has been getting by on trial and error. https://protege.stanford.edu/short-courses/

Beyond that, online courses serve formal ontology engineering poorly. The openHPI courses above cover OWL as one module among many, which is enough to read an ontology and not enough to design one. A realistic substitute is to read existing ontologies (SKOS, PROV-O, schema.org) closely and rebuild a small one from scratch, with the OWL 2 Primer open. Background is at OWL, ontologies, and knowledge graph vs ontology.

Graph databases and Cypher

Neo4j GraphAcademy (Neo4j). Free, self-paced, browser-based courses with a sandbox database attached, so exercises run against a live instance rather than in a text box. Categories include Foundational, Cypher, Development, Analytics, Generative AI and GraphRAG, and MCP, with curated paths for application development and for certification, and titles such as “Neo4j & GenerativeAI Fundamentals” and “Building Knowledge Graphs with LLMs”. It suits developers, and it is the most complete free training on this page.

The obvious caveat: it teaches Neo4j and the labeled property graph (LPG) model, not RDF. Cypher is not SPARQL, and a graduate of GraphAcademy will know one half of the field well and the other not at all. Pair it with the SPARQL tutorials above. The model difference is covered in property graph vs RDF and graph databases. https://graphacademy.neo4j.com/

Knowledge graph construction and NLP

Knowledge Graphs for RAG (DeepLearning.AI, built with Neo4j). A short course on using a graph as the retrieval layer for a language model: loading text into a graph, building vector indexes over it, and writing Cypher that an LLM generates. Format is video with in-browser notebooks. It suits an engineer who already builds with LLMs and wants the graph half, and it is the most current material here. Its scope is narrow, and it assumes rather than teaches the graph modeling underneath. https://www.deeplearning.ai/courses/knowledge-graphs-rag

Advanced NLP with spaCy (Explosion). A free interactive course covering the spaCy pipeline, rule-based and statistical entity recognition, and training a custom model. It is not a knowledge graph course, but named entity recognition is the first stage of almost every text-to-graph pipeline, and this is the best structured teaching of that stage available for free. The format is browser-based lessons with executable exercises. It suits Python developers doing extraction. https://course.spacy.io/en/

What follows extraction is entity resolution, and that is where courses stop. See entity linking, NER and extraction, and GraphRAG for the parts that have to be assembled from documentation.

Structured data and the SEO side

There is no equivalent of GraphAcademy for entity SEO. The material that exists is documentation rather than instruction, and it is good documentation.

Google Search Central structured data documentation (Google) is the reference for which types produce which search features, what each type requires, and the policies. It’s a gallery of supported features, each with required and recommended properties and a worked example. Read the general guidelines page first, because most audit findings are policy problems rather than syntax problems.

schema.org’s Getting Started guide teaches the vocabulary itself rather than one search engine’s subset, which matters when marking up an entity nobody generates a rich result for.

Treating those two as a course, in that order, with the Rich Results Test open in another tab, gets an SEO further than any video series currently on offer. The applied guidance covers entity SEO, structured data, and JSON-LD.

What no course teaches well

You have to learn four things by building, because no course on this page covers them properly.

Entity resolution at scale. Deciding that two records describe the same thing is the hard part of every real project, and courses stop at a worked example with clean data. Blocking strategies, similarity thresholds, transitive merge errors and the human review queue are all learned from the wreckage of a first attempt.

Schema decisions that survive contact with data. Whether something is a class or a property, when to reify a relationship, how much to model up front: courses present finished ontologies, which teaches recognition rather than judgment. The only way through is to model a domain, load real data, find the model wrong, and change it.

Operations. Nothing here covers backup and restore for a triple store, incremental reload without downtime, query performance on a graph too large for memory, or what to do when an inference rule makes materialization explode. That knowledge sits in vendor documentation, mailing lists, and experience.

Provenance and maintenance. Courses build a graph once. Production graphs are rebuilt continuously from sources that disagree, and the questions that dominate are which source wins, how to record where a triple came from, and how to retract a fact without breaking everything that referenced it.

The practical route is to take one small domain, work through the pipeline end to end using how to build a knowledge graph, and come back to the courses for the pieces that turned out to be missing. Books cover some of this ground better than any course does; see books and papers.

FAQ

Is there a free knowledge graph course?

Yes, several. openHPI’s Knowledge Graphs course by Harald Sack and Mehwish Alam is free in self-study mode and covers RDF, OWL, SPARQL, and embeddings in sequence. Neo4j GraphAcademy is free and hands-on for the property graph side. Stanford’s CS 520 seminar materials, including slides and recordings, remain online at no cost.

What is the fastest way to learn SPARQL?

The Wikidata SPARQL tutorial, because every example runs against a live endpoint with no setup and uses real data. Working through it takes an afternoon and ends with the ability to write queries against any endpoint. The Apache Jena SPARQL tutorial covers the same syntax with local files, which suits a developer running their own store.

Is Stanford CS 520 still available?

The course site is still online with slides, lecture notes, and YouTube recordings from the Spring 2021 edition, and the Spring 2020 edition is archived alongside it. It is not an enrollable course, and there is no assessment. The content predates current retrieval-augmented generation work, so it covers data models, acquisition, and inference rather than LLM integration.

Which course should an SEO take?

None of the ones aimed at developers, at first. Google Search Central’s structured data documentation and schema.org’s Getting Started guide cover what an SEO actually needs, in more detail than any video course. Take a graph course later, when the question shifts from marking up a page to building an entity model across a site.

Sources and further reading

Stanford CS 520: Knowledge Graphs: https://web.stanford.edu/class/cs520/
openHPI, Knowledge Graphs: https://open.hpi.de/courses/knowledgegraphs2020
Wikidata SPARQL tutorial: https://www.wikidata.org/wiki/Wikidata:SPARQL_tutorial
Library Carpentry, Introduction to Wikidata: https://librarycarpentry.org/lc-wikidata/
Apache Jena SPARQL Tutorial: https://jena.apache.org/tutorials/sparql.html
Neo4j GraphAcademy: https://graphacademy.neo4j.com/
DeepLearning.AI, Knowledge Graphs for RAG: https://www.deeplearning.ai/courses/knowledge-graphs-rag
Advanced NLP with spaCy: https://course.spacy.io/en/
Protégé Short Courses: https://protege.stanford.edu/short-courses/
Google Search Central, Intro to structured data markup: https://developers.google.com/search/docs/appearance/structured-data/intro-structured-data