If you already publish linked data, you don’t need a separate file for your dataset description: the NDE Dataset Register can read it straight from your triplestore with a SPARQL CONSTRUCT query.

De Art & Architecture Thesaurus (AAT) van het Getty Research Institute is regelmatig slecht bereikbaar. Daardoor levert zoeken in de AAT via het Termennetwerk geen of vertraagde resultaten op, ook in collectiebeheersystemen die het Termennetwerk gebruiken. Andere terminologiebronnen werken gewoon; de AAT-termen zelf zijn nog wel direct bij Getty te raadplegen.

The Network of Terms is increasingly becoming part of the infrastructure used to search, identify and reuse terminology across the Dutch cultural heritage network. Grafana gives us a useful view of how the service is being used.

For the period from 1 January to 10 August 2026, the supplied Grafana export records 11,238,305 user queries and 15,256,333 requests to terminology sources.

Users want access to datasets, not dataset descriptions. The dataset descriptions that you can search, filter and view in the Dataset Register are a means of getting to the dataset’s source data.

A while ago we added worldwide GeoNames to the Network of Terms. That post announced the what; this one is the how, because getting GeoNames in there was less obvious than it sounds.

GeoNames has no usable RDF. There is official ‘RDF’, but it dates from 2020, it is broken RDF/XML and incomplete anyway. There’s also per-resource RDF, but that’s hard to work with for 13M+ resources. So we fell back to the TSV dump: messy, but recent and complete. This is the story of converting that TSV to RDF using only the SPARQL we already know. And of why the pipeline around the conversion is what turns a one-off experiment into an actual service.

Today NDE releases SCHEMA-AP-NDE 1.0: the first stable version of the Schema.org Application Profile every dataset in the Dutch cultural heritage network is expected to publish. It comes with a backward compatibility promise: as long as we stay on 1.x, data that conforms keeps conforming, and software built against it keeps working.

That is what a 1.0 really is: not just a number, but a signal of stability. It is worth asking what that promise means for a specification, and why a shared profile needs a version number at all.

Version 2.0 of the Requirements for Datasets is on the horizon. It tightens rules that are warnings today into full requirements, aligns the Dataset Register with DCAT-AP-NL 3.0, and raises the baseline quality of dataset descriptions across the Dutch cultural heritage network. Publishers and collection management system vendors have until Monday 3 May 2027 to prepare.

Schema.org is het meest gebruikte vocabulaire voor gestructureerde data op het web, ondersteund door Google, Bing en talloze dataplatforms. Het definieert typen als CreativeWork, Person en Organization, waarmee machines inhoud gestandaardiseerd kunnen verwerken. Precies daarom is het NDE-applicatieprofiel gebouwd op Schema.org. Correct gebruik maakt cultureel-erfgoeddata vindbaar en herbruikbaar tussen instellingen.

Maar als je met Schema.org in RDF hebt gewerkt, ben je dit tegengekomen: er is http://schema.org en er is https://schema.org. Dat ene lettertje verschil kan voor echte problemen zorgen: SPARQL-query’s die niets teruggeven, SHACL-validatie die goede data afwijst of slechte data negeert, of datasets die niet aan elkaar te koppelen zijn – vooral als je data uit meerdere bronnen combineert, zoals binnen NDE gebeurt.

Het Termennetwerk biedt nu de mogelijkheid om binnen een terminologiebron te filteren op onderwerp.

We hebben een publieke statuspagina gelanceerd die real-time inzicht biedt in de beschikbaarheid van NDE-diensten.