Institution-name research

Use this procedure when you correct a library name or add an access_point_title or library_alternate_names entry. The tracked research/name-evidence.csv register holds one preferred-name decision for every directory record. Legacy labels start as pending; that status does not mean the name has been checked. The first review queue includes the 25 record IDs already selected for the linked-data pilot. Review them in small batches; their name decisions remain separate from authority and IIIF decisions.

Identify what the record describes

Read the record's website, city, nation, aggregators, and any existing authority evidence. Decide which institution or relevant unit holds the manuscripts and which portal or collection provides the access point. Two records can name the same institution while describing different access points. Do not merge or renumber them because their names, authority identifiers, or websites match.

Use the institution's current official name as it publishes it, preserving its script and diacritics. Use a public institutional page or issuing registry to confirm the exact unit. A search result, last_checked date, or shared URL does not verify the institution. Keep a doubtful existing label pending and explain the uncertainty in the register.

Add access_point_title only when the source names a distinct portal or collection. Add a current official variant or translation under library_alternate_names; add its BCP 47 language when known. A former name, incorrect spelling, or portal title is not an alternate institution name. These source fields support reader discovery. They do not create an RDF institution resource or an identity relationship.

Record and review evidence

The register columns are fixed:

record_id,field,value,language,status,source_url,checked_on,note

Use field library, access_point_title, or library_alternate_names. Keep exactly one library row per record. Use pending for an unverified current name or unpublished candidate. Use verified only after opening a public source that names the correct institution, unit, or access point. Verified rows need the exact published value, source URL, and real check date. For alternate names, the register's language must match the data entry.

Before changing library, check whether the record's ISIL, Wikidata QID, and any approved holding-institution relationship still describe that holder or unit. Follow identifier research and the existing assertion approval process if one of those values must change. A name-only correction does not add a row to link-assertions.csv. Preserve last_checked unless you actually reassessed the collection URL.

For a changed library, append its new name-derived slug under the same ID in library-aliases.json. Keep every earlier slug. Set last_edited to the real date of a material correction; keep added and the stable ID unchanged. Document the public name evidence and any unresolved scope question in the pull request for maintainer review.

Command safety: Safe

python scripts/validate_data.py
python scripts/validate_name_evidence.py

The first check validates catalogue shape and alias coverage. The second checks register completeness and agreement with published fields. In pull requests, verify_name_history.py also requires a verified row for every newly published name or title. These checks cannot establish that a source's institutional claim is true; a maintainer must inspect the cited page.

For the full build and publication checks, follow How we validate data and Linked data and persistent identifiers.