Sunday experiments with French museum collections.

Sunday experiments with French museum collections.

This is "L'Oiseau Bleu," a Jean Paul Gaultier gown from Palais Galliera, musée de la Mode de la Ville de Paris. It surfaced this morning as the first search result for "oiseau bleu" from a pilot test of the Paris Musées API inside #LinkedCulture.

The question for both sources added today was the same: is this collection viable for semantic treatment?

Paris Musées: 1,751 records indexed as a test set. GraphQL API, clean thumbnails, good field coverage. Quota is 1,000 queries/day on the public tier, which limits scale but was enough to validate the integration. Verdict: viable. Worth pursuing the private API for full indexing.

Joconde: 10,000 records as a pilot sample from a 1M+ record corpus. Open License 2.0, no API key, 1.1 GB bulk CSV from the French Ministry of Culture. The constraint: the open data export is text-only. No image field. We attempted to bridge that via Wikidata ID mapping to pull image URLs, but coverage was thin. Only a handful resolved. Verdict: viable as a large-scale French-language text corpus. Image enrichment via Wikidata is the right path but needs more work.

Both pilots are now live in the same multilingual vector index as 220,000+ English-language records. Cross-lingual queries work without a translation layer. A search in English surfaces French records.

Paris Musées: https://linkedculture.org/multilingual?source=parismusees

Jconde: https://linkedculture.org/multilingual?source=joconde

That's what I came to confirm. It works. Enjoy! C'est ce que je voulais confirmer. Ça fonctionne. Bonne découverte !