Crossref DOI Metadata Search

Search Crossref scholarly works or resolve DOI metadata into clean structured records — titles, authors, publishers, dates, types, citation/reference counts, licenses, ISSNs, ISBNs, subjects, and canonical DOI resolver URLs. Two modes in one actor:

• Search mode — supply a free-text query and get back ranked works.
• DOI mode — supply a list of dois and get back one structured record per DOI.

Both modes use Crossref's public REST API (https://api.crossref.org/). No API key, no proxy, no captcha, no browser.

What it does

• Hits /works?query=... for free-text search and /works/{doi} for direct DOI resolution.
• Projects each work into a stable schema with doi, title, authors[], publisher, containerTitle, publishedDate, issuedDate, referenceCount, citationCount, isReferencedByCount, subjects[], licenses[], language, issn[], isbn[], url, resourceUrl, sourceQuery, fetchedAt.
• Strips <jats:*> and other XML tags from abstracts; resolves ORCID URLs to bare ORCIDs.
• Deduplicates search results by lowercased DOI.
• DOI mode failures push an {doi, sourceQuery, fetchedAt, error} warning row instead of crashing the whole run.
• Polite-pool User-Agent + ~3 req/sec throttle between DOI calls.

Input

Field	Required	Default	Description
query	one of	—	Free-text work search (search mode)
dois	one of	—	DOIs to resolve (DOI mode); array of strings
maxItems	no	50	Max results from a search (1–500)
start	no	0	Pagination offset (0–100000)
fromPublishedDate	no	—	YYYY[-MM[-DD]] lower bound (search mode)
untilPublishedDate	no	—	YYYY[-MM[-DD]] upper bound (search mode)
types	no	—	Crossref work types (e.g. journal-article, book-chapter)
sort	no	relevance	relevance | published | is-referenced-by-count

At least one of query or dois must be non-empty. Mode is picked by which is supplied (search wins if both are given).

Output

One dataset record per work. Search-mode example:

{
  "doi": "10.1093/oso/9780198828044.003.0003",
  "title": "Machine learning with sklearn",
  "authors": [
    { "given": "Thomas P.", "family": "Trappenberg", "orcid": null, "affiliation": [] }
  ],
  "publisher": "Oxford University Press",
  "containerTitle": "Fundamentals of Machine Learning",
  "type": "book-chapter",
  "publishedDate": "2019-11-28",
  "issuedDate": "2019-11-28",
  "referenceCount": 0,
  "citationCount": 19,
  "isReferencedByCount": 19,
  "language": "en",
  "issn": [],
  "isbn": ["9780198828044", "9780191883873"],
  "url": "https://doi.org/10.1093/oso/9780198828044.003.0003",
  "resourceUrl": "https://academic.oup.com/book/36775/chapter/321905277",
  "sourceQuery": "machine learning",
  "fetchedAt": "2026-07-16T16:30:00.000Z"
}

DOI-mode records look identical except sourceQuery is the DOI itself.

Limitations

• Pure public REST wrapper — does not scrape paywalled publisher pages.
• reference-count and references-count may both appear in Crossref records; this actor prefers references-count (canonical newer field) and falls back to reference-count only if missing.
• subject is sometimes [] for works without FACET data — emitted as empty array, not null.
• abstract HTML/JATS tags are stripped; tag content remains as plain text with entity decoding for &, <, >, ", ',  .

Pricing

Free tier: 100 runs/month on the public Apify free plan. Public Store tier not yet enabled for this actor.

Changelog

• 0.1.0 — initial release.