Wikidata SPARQL Query Service Data

It is the window onto Wikidata's knowledge graph - the seat where questions become tables. Datadory catalogs two publishing datasets from Wikidata: the twin record holds the shelf itself, 122,983,238 entities of editions, works, authors and publishers connected by typed statements, while this one covers the query service that reads that graph. The service speaks SPARQL 1.1 over the Wikibase RDF data model: you declare the shape - which editions, which properties, which markets - and it returns result sets assembled to match.

The seat is unusually well appointed. A visual Query Builder and an interactive Query Helper assemble triple patterns without hand-written syntax, a curated examples collection supplies ready-made starts, and result views cover table, map, graph, timeline and image grid - a publisher-location pattern renders as pins before any code exists. Every one of those capabilities was confirmed live during the August 2026 research pass, alongside a probe joining ISBN-13 (P212), title (P1476) and publication date (P577) that came back with language-tagged titles and typed date literals.

Datadory takes the machinery out of the browser and hands you the result sets as flat, typed rows. Context sits on the publishing data hub; get a sample of this dataset cut to your publishers, languages or windows.

Two captured solutions beat a specification paragraph:

# one solution row - delivered grain is one row per query binding
book                : Q100708395
title    (P1476)    : Silver Economies, Monetisation and Society in Scandinavia,
                      AD 800-1100                    [language tag: en]
isbn13   (P212)     : 978-87-7934-585-0
pubdate  (P577)     : 2011-01-01T00:00:00Z             # xsd:dateTime

# a second row shows the same shape carrying another language market
book                : Q100708731
title    (P1476)    : Ovo gigante da Mothra : Consumo do Pacifico Sul no Japao
                      dos anos 1960                   [language tag: pt]
isbn13   (P212)     : 978-85-98112-63-3

# the response envelope travels beside every result set
head.vars           : book | title | isbn13 | pubdate
results.bindings    : one solution object per row, each value typed uri or literal

# the graph underneath the rows
entities.total      : 122983238        # counter read August 2026

Read the anatomy rather than the antiquity of the first title. First, the grain is honest: one row per binding, so a publisher-year count and an edition listing differ only in the question, never in the shape. Second, identity arrives as entity references - Q-numbers that resolve to publisher and class nodes - rather than loose strings, which is what keeps a thousand-row pull joinable instead of fuzzy. Third, the typing is load-bearing: dates land as xsd:dateTime, titles carry their language tag, and a Portuguese-market edition reads natively beside an English one. Your scope ships with the sample in exactly this layout.

Seven documented fields in two layers, definitions verified against live captures in August 2026 - a bar met by 85.7% of the 1,744 datasets Datadory catalogs.

The envelope layer makes every response self-describing: head.vars names the columns, results.bindings carries one solution object per row with each value typed as URI reference or literal. Bind against that contract once and the shape survives however the query changes.

The solution layer carries the publishing payload: ?book addresses the edition entity by Q-number, ?title (P1476) arrives monolingual and language-tagged, ?isbn13 (P212) keys the join onto trade catalogs, ?publisher (P123) points at the publishing-house item, and ?pubdate (P577) preserves its recorded precision. Publisher-label resolution and rollup columns fold in on request rather than cluttering the core.

Geography - global and multilingual, mirroring the graph itself. Language-tagged titles mean regional reads are native rather than derived, with statement density varying by language community; screen fill rates for any scope before leaning on completeness.

Temporal - current-state coverage of a continuously edited graph, not a dated snapshot. Each row carries its own publication date (2011-01-01T00:00:00Z on the captured row), so a historical cohort study and this quarter's new-edition watch share one dictionary. Scheduled captures add the history dimension the current state lacks.

Granularity - triple-level statements aggregated into user-defined result sets per query: one row per binding, shaped by the question. Result size is bounded by per-query limits rather than by the dataset, so large analytical joins stage in pieces - handled at scoping, not discovered mid-project.

Scored honestly: 9 of 10 on Datadory's rubric against a 7.81 mean across all 1,744 cataloged datasets, sitting just under the sixteen-record publishing slice's 8.5 average - held up by the verified response contract, docked because property-level fill rates vary and need measuring per scope rather than being promised blanket.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; query construction, staging under per-query limits and publisher-name resolution stay our problem. Result sets arrive flattened to the field dictionary above - one typed row per binding, language tags intact - so consecutive deliveries diff cleanly into new editions and corrected statements instead of piling up as undifferentiated snapshots.

Hourly suits teams watching statements land on titles they track. Daily suits catalog-enrichment and audit jobs that run between warehouse loads. Weekly suits network work, where publisher graphs move slowly enough that snapshots lose nothing. A sample cut to your named publishers, languages, subjects or date windows comes first either way.

• Publisher output measurement - editions per imprint per year, grouped straight from bindings on publisher and publication date; one of 367 cataloged datasets serving market sizing.
• Rival catalog watching - competitors' edition arrivals and identifier hygiene diffed week over week; one of 195 tagged for competitor tracking.
• Missing-identifier audits - the editions lacking ISBN-13, publisher or date statements, listed outright; data-quality arguments become work queues with a length.
• Multilingual market reads - language-tagged titles separate markets natively, no translation layer in between.
• Citation-grade metadata pulls - typed, sourced edition values that survive review; the catalog holds 765 datasets clearing that bar across 133 industries, and this one resolves editions across languages inside it (citation-grade research).

Data scientists and ML engineers land closest fit: prototype a join interactively, then industrialize the extract - typed bindings drop into pipelines without a bespoke parser; see data scientists x publishing. Developers and builders embed lookup, deduplication and audit features behind one stable envelope via developers & builders view. Market researchers and consultants build publisher-by-year output curves and multilingual footprint maps through market researchers view. Competitive intelligence teams run rivals' edition counts as a standing feed via competitive intel view. Journalists, academics and students verify bibliographic detail across languages with typed values attached to the editions they describe - journalists & academics view.

This record owns the question-answering layer of the Wikidata pair. The twin Wikidata - Publishing Knowledge Graph + SPARQL is the shelf: entity-level rows for enrichment, validation and network loads. The dedicated comparison walks the split - if the deliverable is a dataset, take the shelf; if it is an answer, take the window; most teams take both in one feed.

Elsewhere on publishing's shelf: Crossref REST API is the DOI registration layer with 185,678,115 works; OpenAlex API is the breadth play at roughly 322 million works including 5.9 million books; the International ISBN Agency registers cover numbering-group plumbing without per-title statements. None of those return arbitrarily shaped result sets over a multilingual graph. When the question is who published what, in which language, when, this is the record.

Because the raw artifact answers one question at a time in a browser tab, and every interesting business question wants thousands of bindings across publishers, years and languages in one frame. Datadory industrializes the extract: queries constructed and staged under per-query limits, publisher names resolved beside every identifier, deliveries keyed so consecutive pulls diff into clean timelines, and a sample in hand before any commitment. Browse the rest of the slice on the publishing data hub, see where the pair ranks on best publishing datasets, or get a sample of this dataset cut to your scope.