Appearance
To get started with Wikidata for heritage, create a verified account, learn the item-property-value data model on three or four test items, then reconcile your existing catalogue against Wikidata before any bulk upload. Begin small: model one object by hand, query it back with SPARQL to confirm it behaves, and only then scale. The single most common mistake is uploading hundreds of items before understanding how statements, qualifiers and references fit together.
What is Wikidata and why does it suit heritage data?
Wikidata is a free, multilingual, structured knowledge base where every fact is a triple: an item (like Q42), a property (like P31, "instance of") and a value. For heritage work this matters because your catalogue stops being a flat spreadsheet and becomes a graph that links to VIAF, GeoNames, the Getty vocabularies and other institutions automatically.
All content is released under CC0, so anything you contribute enters the public domain. That is a feature for shared authority control but a constraint you must respect: do not upload rights-restricted descriptions or personal data about living people without care.
How do I set up an account the right way?
- Register at
wikidata.orgwith a real email so you can recover the account. - Add a short user page describing your institution and project; reviewers trust transparent contributors.
- Make a handful of manual edits to reach autoconfirmed status (4 days plus 50 edits). This unlocks the batch tools.
- Request a bot flag only if you will exceed a few thousand automated edits.
Keep institutional and personal editing separate, and never share a login across staff — edit history is your audit trail.
What is the minimum viable model for one heritage item?
Model a single object by hand first. A 17th-century parish register might look like this in QuickStatements syntax:
text
CREATE
LAST Len "Parish register of St Mary, Beverley (1601-1650)"
LAST P31 Q87167 /* instance of: manuscript */
LAST P195 Q1234567 /* collection */
LAST P571 +1601-00-00T00:00:00Z/9 /* inception, year precision */
LAST P195 Q1234567 S854 "https://archive.example/ref"Notice the S854 reference (reference URL) attached to the statement — every claim that could be challenged needs a source. Date precision /9 means "year only", which is honest about what you actually know.
How do I avoid creating duplicates?
Search before you create. A practical check sequence:
| Step | Tool | What it catches |
|---|---|---|
| 1 | Wikidata search box | Obvious exact-label matches |
| 2 | Mix'n'match | Items in catalogues already aligned |
| 3 | Institutional ID property | Records linked by your own accession number |
| 4 | OpenRefine reconciliation | Fuzzy matches across a whole dataset |
For anything beyond a dozen records, reconcile in OpenRefine rather than searching by hand — it scores candidates and flags ambiguous ones for review.
How do I check my work before scaling up?
Query it back. Run a small SPARQL query against the live endpoint to confirm your item appears as intended:
sparql
SELECT ?item ?itemLabel ?inception WHERE {
?item wdt:P195 wd:Q1234567 . # in your collection
OPTIONAL { ?item wdt:P571 ?inception . }
SERVICE wikibase:label { bd:serviceParam wikibase:language "en". }
}If the row is missing, your statement did not save or you targeted the wrong property. Fixing one item is trivial; fixing 800 is a weekend lost.
What pitfalls trip up newcomers most?
- Treating descriptions as prose — Wikidata descriptions are short disambiguators, not catalogue notes.
- Forgetting references, which makes statements look like unsourced opinion.
- Over-precise dates that imply certainty you do not have.
- Modelling collection structure with free text instead of
P361(part of) /P195(collection) links.
Key Takeaways
- Learn the item-property-value triple model before touching batch tools.
- Reach autoconfirmed status, then graduate to QuickStatements and OpenRefine.
- Model and verify one item by hand before scaling to hundreds.
- Reconcile against existing items to prevent duplicates.
- Every challengeable statement needs a reference (
S-prefixed source). - Respect CC0: no rights-restricted text, careful handling of living-person data.
- Use date precision honestly instead of padding unknown days.
Frequently Asked Questions
Do I need permission to add my heritage collection to Wikidata?
No special permission is required, but each item must meet Wikidata's notability policy and the data must be compatible with the CC0 public-domain dedication. Sensitive personal data and copyrighted descriptive text cannot be uploaded.
Is Wikidata free to use for a cultural institution?
Yes. Reading, editing, querying via SPARQL and bulk export are all free with no account tier. The only cost is staff time for modelling and reconciliation.
How many edits can a new account make before hitting limits?
Unconfirmed accounts are rate-limited to roughly 8 edits per minute and cannot use some tools. After 4 days and 50 edits an account becomes autoconfirmed, which unlocks QuickStatements and higher throughput.
Should I create one item per object or per record?
Create one item per real-world entity. A single archival record describing five distinct objects should usually become five items linked back to one collection item.
What is the fastest way to find an existing item before creating a duplicate?
Search the label plus a disambiguating word, then check the Mix'n'match catalogue and any institutional authority ID property. Reconciliation in OpenRefine catches duplicates at scale.