Reconciliation Service Extractor (OpenRefine / W3C)

Point this actor at any W3C Reconciliation Service API endpoint (the OpenRefine reconciliation standard) and pull structured results — no per-service code required. It works with Wikidata, the German National Library (GND / lobid), the Getty Vocabularies, and every other conforming reconciliation endpoint.

Reconciliation is how you turn messy text — a list of person names, place names, organizations, subject terms — into linked, canonical identifiers from an authority (a Wikidata Q-id, a GND number, a Getty ULAN/AAT/TGN id). It's the backbone of data cleaning, entity linking, and knowledge-graph building in libraries, museums, archives, and open-data projects.

What it does

Four modes, auto-gated on what the target service actually advertises in its manifest:

Mode	What it returns
manifest	The service capability manifest — name, identifier space, default types, and which optional modes (suggest/extend/preview) the service supports. The cheap connectivity probe; always run first.
reconcile	Batch-match a list of query strings to candidate entities. One row per candidate, carrying the raw score, the match auto-match flag, the entity type[], and description.
suggest	Autocomplete entities, properties, or types from a text prefix (only if the service advertises a suggest endpoint).
extend	Data-extension — fetch property values for a set of matched ids (e.g. birth date + birth place for a person id). Only if the service advertises extend.

Example input

{
  "endpoint": "https://wikidata.reconci.link/en/api",
  "modes": ["manifest", "reconcile", "extend"],
  "queries": ["Albert Einstein", "Marie Curie", "Isaac Newton"],
  "typeFilter": "Q5",
  "limit": 5,
  "extendIds": ["Q937"],
  "extendProperties": ["P569", "P19"]
}

Endpoints you can point it at

• Wikidata — https://wikidata.reconci.link/en/api (general knowledge graph; supports all four modes)
• GND (German National Library) — https://lobid.org/gnd/reconcile (library authority data)
• Getty Vocabularies — https://services.getty.edu/vocab/reconcile/ (art & architecture: ULAN, AAT, TGN)
• ...and any other endpoint conforming to the Reconciliation Service API spec.

Output

Each result is one flat row tagged with its mode. Reconcile rows carry the query, the matched id/matchName, the raw score (scales differ per service — kept un-normalized), the match flag, and the entity type[]. Manifest rows carry the full capability description. Results render as a clean table in the Apify Console and export to JSON, CSV, Excel, and more.

Notes

• Scores are not normalized — different services use different scales (0–100, unbounded float), so the raw score is preserved for honest cross-service comparison within one service.
• A query with zero candidate matches is a real, honest empty result (emitted as a no-match row), not an error.
• Optional modes are skipped with a warning if the target service does not advertise them.

Pricing

Pay-per-event: you are charged per result row produced. Fetch a manifest, reconcile a list, or extend a set of ids — you only pay for the records you get back.

Budget control: each row is billed as it is written (US$0.0005 per record on the FREE tier, lower on paid tiers). Set Max total charge on the run to cap spend: the Actor stops once the next row is no longer paid for, sends only as many reconcile queries as the budget covers, and makes no request at all if the budget cannot pay for one row. Every query returns at least one row (a no-match row when nothing matches), and each candidate is a row, so 3 queries with limit 5 plus the manifest can cost up to 16 records (≈US$0.008).