Recommended Reading: Books, Papers, Specs & Courses

No single canonical text exists on knowledge graphs. The subject was assembled from four traditions that developed separately and still use different vocabulary for the same ideas: database research (graph data models, query languages, integration), W3C standards work (RDF, OWL, SPARQL, linked data), natural language processing (entity extraction, entity linking, relation extraction), and search marketing practice (entities, schema.org markup, knowledge panels). A reading list that covers one tradition and ignores the others produces a lopsided understanding, which is why this section is split by format rather than by topic.

How to read this literature

Three habits make the material easier to work through.

Read a definition paper before a tutorial. The word “knowledge graph” is used loosely, and much of the confusion in the field comes from two people using it for different things. Fixing a working definition first (see /concepts/what-is-a-knowledge-graph/) makes the rest of the literature legible.

Treat standards as primary sources. The W3C Recommendations for RDF, SPARQL, OWL and SHACL are more precise than any book written about them, and they are freely available. Books are useful for motivation, worked examples and modeling judgment. Specs are the authority on what a construct actually means.

Date every source. The semantic web literature has a large body of work from roughly 2001 to 2012 that is conceptually sound and technically stale. Papers on knowledge graph embeddings from before about 2017 describe methods that later models outperform, but they still define the problem and evaluation setup. Anything on knowledge graphs and language models written before 2023 predates the current retrieval architectures. None of that makes older work wrong; it means readers should know which parts to treat as history.

Books

The books page is an annotated list grouped by what the reader wants from it: foundational and conceptual treatments, semantic web and RDF standards, graph databases and engineering practice, ontology engineering, and the practitioner and SEO side. Each entry says what the book teaches, who should read it, and where it is dated.

Two of the best sources cost nothing. The multi-author survey Knowledge Graphs by Hogan and colleagues is a full book published in the Synthesis Lectures series and is readable in HTML at kgbook.org. Linked Data: Evolving the Web into a Global Data Space by Heath and Bizer is free at linkeddatabook.com and remains the clearest short explanation of why identifiers and dereferenceable URIs matter, even though its tooling references are old.

The books section serves readers who want a structured path through a topic rather than an answer to a specific question: someone new to RDF who needs to understand why a triple is shaped the way it is, an engineer choosing between a triple store and a labeled property graph (LPG) database, or an SEO who understands schema.org markup but not the entity model underneath it.

See /reading/books/.

Papers

The papers page collects research papers and reports that shaped the field, annotated with what each one contributed and whether it has been superseded. It is organized by theme: definitions and foundations, the public knowledge graphs (DBpedia, YAGO, Freebase, Wikidata, and Google’s Knowledge Vault), graph data models, embeddings and relational machine learning, and the recent work joining knowledge graphs to large language models.

Papers serve a different purpose from books. They are where a technique is defined precisely enough to reproduce, and where the limits of a technique are stated honestly. For an engineer evaluating whether graph-based retrieval beats vector-only retrieval on a particular corpus, the source that answers the question is a paper with an evaluation section, not a vendor tutorial. Papers are also where the primary evidence lives for claims that circulate as folklore, which matters when a design decision rests on one of them. Related background is at /concepts/knowledge-graph-embeddings/ and /build/graphrag/.

The page includes non-academic primary sources where they are the original record of something, such as Google’s 2012 announcement of the Knowledge Graph and Microsoft Research’s GraphRAG report.

See /reading/papers/.

Specs and standards

The specs page indexes the primary standards documents: the W3C Recommendations for RDF, RDFa, SPARQL, OWL, SHACL and JSON-LD, the schema.org vocabulary, the Open Graph protocol, and the relevant WHATWG material on microdata. Each is linked from the matching page in the /languages/ section so that a reader who wants the explanation and a reader who wants the normative text can each go straight to what they need.

This page serves anyone with a question that has an exact answer. Which characters are legal in a Turtle-prefixed name, what owl:sameAs formally entails, whether SPARQL FILTER placement changes the result, and what JSON-LD 1.1 added over 1.0: these are spec questions, and secondhand summaries are frequently wrong. Developers implementing against a standard should work from the Recommendation. SEOs should note that Google’s structured data documentation is separate from the schema.org vocabulary and imposes its own requirements, so they should consult both. Start at /languages/resource-description-framework/ or /languages/sparql/ for context before reading the specs themselves.

See /reading/specs-and-standards/.

Courses

The courses page lists structured teaching material, free and paid: university courses with public materials, vendor academies that teach a specific stack, and short courses on graph-based retrieval. Courses suit readers who learn by doing exercises against a real endpoint rather than by reading, and they are the fastest route to competence in a query language.

The key distinction is between vendor training and general education. Vendor material is usually excellent on its own product and quiet about the trade-offs that would favor a competitor. University material is more balanced and often older. Using both, and knowing which is which, produces a better result than either alone.

See /reading/courses/.

Where to start

For a developer with no background: read the definition material here, then the free Hogan survey book, then the W3C RDF primer, then work through a query language course against a live endpoint such as the Wikidata Query Service.

For a data engineer evaluating an architecture: read the graph database and enterprise knowledge graph books, the graph data model survey, and the specs for whichever stack is under consideration. Skip the embeddings literature until a link-prediction problem actually arises.

For an SEO: read the entity and structured data material on this site first, since most of the academic literature answers questions SEOs do not have. Then read the linked data book for the identifier model, Google’s own documentation for what search engines actually consume, and the practitioner books for applied technique. Background is at /seo/entity-seo/ and /ecosystems/schema-org/.

Unfamiliar terms across all four pages are defined at /glossary/.

FAQ

What is the best book to start with on knowledge graphs?

For a general audience, the survey book Knowledge Graphs by Hogan and colleagues covers data models, schemas, identity, context, reasoning, embeddings, and quality in one place, and it is free at kgbook.org. Readers who want applied engineering rather than a survey are better served by a practitioner book on building graphs, and readers coming from SEO should start with the entity material instead.

Are there free books and papers on knowledge graphs?

Yes, and several of the best sources cost nothing. The Hogan survey is free at kgbook.org, the Heath and Bizer linked data book is free at linkeddatabook.com, most of the major research papers are on arXiv or in open conference proceedings, and every W3C standard is published openly. Each entry on the books and papers pages notes free availability.

Should SEOs read the academic knowledge graph literature?

Selectively. Most academic work addresses construction, reasoning, and machine learning problems that do not arise in search marketing. The parts that repay the time are the definition papers, the linked data material on identifiers and sameAs, and the papers describing how public graphs such as Wikidata and DBpedia are built, since those are the sources search engines draw on.

How current is this reading list?

Entries are checked for existence and correct attribution, and each one notes where the material has aged. The field moves fastest at the boundary with large language models, so the retrieval and generation papers turn over quickest, while the standards and the foundational definition work have been stable for years.

Sources and further reading

Hogan, A. et al. Knowledge Graphs. Free online edition: https://kgbook.org/
Heath, T. and Bizer, C. Linked Data: Evolving the Web into a Global Data Space: http://linkeddatabook.com/
W3C. Semantic Web Standards: https://www.w3.org/standards/semanticweb/
W3C. RDF 1.1 Primer: https://www.w3.org/TR/rdf11-primer/
schema.org: https://schema.org/
Google Search Central. Structured data general guidelines: https://developers.google.com/search/docs/appearance/structured-data/sd-policies