OpenAlex API - Scholarly Book & Publisher Graph

It is the breadth layer of scholarly publishing data, filed under Publishing and scored 9 out of 10 on Datadory's quality rubric against a catalog mean of 7.81 across 1,744 datasets. OpenAlex, compiled by a US 501(c)(3) nonprofit as the successor to Microsoft Academic Graph, indexes the world's research; the August 2026 research pass observed 322,044,938 works, including 5,906,748 books and 22,015,762 book chapters - 27,922,510 long-form scholarly units combined.

Two things make it analytic rather than merely searchable. First, citation impact travels on the row: cited_by_count, the field-weighted fwci index, year-normalized percentiles and annual counts_by_year, so a 1988 French field report and a 2013 clinical manual compete on measured terms. Second, the publisher entity is structural: hierarchy_level, parent_publisher and lineage turn a flat list of 10,708 publishers into a corporate map. Datadory normalizes all of it into typed, documented tables delivered on your cadence.

One row per indexed work, wrapped in index-wide counters:

id                : 
title             : Diagnostic and Statistical Manual of Mental Disorders
type              : book
publication_year  : 2013
cited_by_count    : 115
language          : en

Read what the pinned row proves. The id anchors every join and survives downstream renames; type: book places the row in the 25-value vocabulary before any classifier guesses; and cited_by_count: 115 measures a reference work's standing directly on the row.

Two more rows show the schema's honest range. The Australian century (1999) carries 18 citations; Rapport de mission au Rwanda 28. 1 au 06. 2. 1988, a French-language government field report, carries zero - and that zero is the point. Long-tail non-English monographs persist as first-class rows rather than being dropped, which is what makes cohort and coverage studies possible. Index-wide counters - 322,044,938 works, 5,906,748 books, 22,015,762 chapters, roughly 125.7 million authors, 255,627 sources, 10,708 publishers - travel beside the rows they summarize, so headline totals stay reconcilable with the data behind them. Example rows are illustrative of the record shape; your sample ships pinned to real records cut to your filters.

Sixteen field groups define the work and publisher records, definitions verified during the August 2026 research pass rather than inferred from column names. Grouped as Datadory ships them:

• Identity: id as the stable entity URI, doi where registered, title, and type across the 25-value work-type vocabulary.
• Dates: publication_year beside the finer publication_date.
• Impact: cited_by_count, the field-weighted fwci, cited_by_percentile_year and citation_normalized_percentile bands, and annual counts_by_year series.
• People: authorships with positions, ORCID-linked author ids, institutions, countries and raw affiliation strings.
• Hosting: the open_access status object and primary_location/locations with landing page, PDF, version and hosting source per location.
• Classification: topics, keywords and primary_topic for subject filtering and trend slicing.
• Container: biblio volume, issue and page range.
• Graph: referenced_works and referenced_works_count - outbound citation edges as columns.
• Publishers: works_count and cited_by_count aggregates plus hierarchy_level, parent_publisher and lineage corporate-structure fields.

Fields whose worked example ships as structure rather than literal values - the nested objects among them - confirm against live records when your sample is cut rather than being padded with guessed values here. The full dictionary follows in tabular form below.

Five extensions reshape the same graph rather than adding columns to the core. Author and institution resolution joins authorships onto the author and institution entities, so country and institution become filterable dimensions across roughly 125.7 million authors. Publisher lineage tables flatten the hierarchy into parent-child pairs, turning the 10,708-publisher list into an account map of who owns which imprint. Citation-edge extracts re-serve referenced_works as an edge list ready for network analysis. Topic and year rollups re-aggregate output and citation uptake over any window you name.

The most requested extension is the crosswalk: joining this corpus onto Crossref, arXiv and DOAJ by DOI, title and author - posted, registered, indexed and open-access-screened in one pipeline instead of four stitched exports. Each extension confirms against your named scope when the sample is cut.

Geography - global research output across all countries and publishers. Language is recorded per work, so the French field report in the sample above is queryable as itself, not noise around an Anglophone core.

Temporal - historical through current. Decades of monograph backfile sit beside recent arrivals in one schema, which is why a single feed serves both a thirty-year citation-cohort study and a new-title watchlist without changes on your side.

Granularity - entity-level throughout: works, authors, sources, publishers and institutions as separate record sets, joined by citation-network edges rather than fuzzy name matching. Rollups arrive per publisher and work type, so record-level science and market-share arithmetic share one dictionary. One honest caveat: book coverage depends on upstream harvests, so titles without registered identifiers may be underrepresented - screen your scope at sampling if completeness against a known list matters.

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

You pick the channel and the cadence; extraction plumbing, pagination walks and change detection stay our problem. Rows arrive normalized to the field dictionary above and keyed on id, so consecutive deliveries diff cleanly into new-work and citation-growth timelines instead of piling up as undifferentiated dumps.

Hourly suits teams watching new arrivals or citation movement where latency matters. Daily suits competitive dashboards tracking publisher output through the lineage tree. Weekly suits trend work, where citation curves move slowly enough that point-in-time reads lose nothing. Whichever you pick, definitions travel unchanged and schema shifts are flagged rather than discovered. A sample cut to your named publishers, fields or years comes first either way.

Six jobs the record settles outright:

1. Bibliometrics and impact scoring - fwci and percentile bands put monographs on the same measured footing as articles; see citation-grade research.
2. Publisher competitive benchmarking - works and citation totals per publisher, resolved through the lineage tree, expose group-level output moves; see competitor tracking.
3. Research-trend monitoring - topics, keywords and the book-versus-article mix read each discipline's direction early.
4. Book-market and list curation - citation uptake per title gives editors a demand signal across 5.9 million books.
5. ML training corpora - titled, classified, language-tagged long-form works with citation edges as columns; see ML model training.
6. Systematic review automation - typed, dated, topic-filtered screening across millions of candidates in seconds.

Data scientists and ML engineers rank highest fit: 322 million nodes of citation edges, affiliation chains and topic labels are labeled structure raw dumps rarely provide; see data scientists' publishing use cases. Journalists, academics and students follow - cite the exact monograph, resolve its impact standing and trace the citation trail without hand-built bibliographies. Competitive intelligence and product teams get the lineage fields that survive corporate restructuring; see competitive intel product teams' publishing workflows. Market researchers and consultants size research output by country, field and work type; see market researchers' publishing workflows. Developers and builders ship discovery and reference products on one uniform entity schema; see developers builders' publishing use cases.

Publishing's shelf splits by lifecycle stage, and this record owns the breadth-and-citations layer. Crossref holds the registration authority upstream - the registry aggregators build on - with deeper deposit-side commerce fields per work; this one answers when the question spans books, chapters and citation networks in a single frame. arXiv sits further upstream at the preprint instant. DOAJ curates the journal-business layer. On the trade side, Internet Archive answers where a readable copy exists, while this record answers what scholarship did with the book - the contrast is worked through properly in the Internet Archive vs OpenAlex comparison.

If the question is who published what, and who cites whom, across all of research - this is the record.

Because the raw artifact underneath is built for application lookups checking one entity at a time, and every interesting question wants millions of works, many publishers and many years in one frame. Datadory hands over the graph itself: the sixteen-group dictionary kept verified and current, publisher lineage resolved into its own table, citation edges served as columns ready for network analysis, and deliveries keyed on id so consecutive pulls diff into clean timelines - with sample rows in hand before any commitment. Browse the rest of the slice on the publishing data hub, see how the source fits the wider catalog on the OpenAlex source profile, or start from the best publishing datasets ranking.