2.5 - Manage vocabularies

Moderator note: Comments are from the previous discussion platform.

Submitted by Vladimir ALEXIEV on Sat, 27/04/2024 - 14:12
Hi Charline!

Will this also include binding vocabularies to data, eg to describe semantically columns of a dataset, or even to enable their semantic representation on the fly? Or would that be part of another requirement?

Eg see these publications

Semantic Interoperability for Data Spaces. Alexiev, V.In SEMIC: Data Spaces in an Interoperable Europe (SEMIC 2022), December 2022. Paper link bibtex
Data Wants To Be Truly Sovereign: Designing Data Spaces with Linked Data Principles In Mind. Petkova, T.; and Alexiev, V.Ontotext blog post, November 2022. Paper link bibtex abstract
Data Spaces vs Knowledge Graphs: How to Get To Semantic Data Spaces?. Alexiev, V.In Data Spaces & Semantic Interoperability Workshop, Vienna, Austria, July 2022. Paper slides video blog link bibtex abstract

Submitted by Mikael af Hällström on Wed, 24/07/2024 - 10:54
Hi,

Vladimir seems to be the only one asking the same questions regarding "semantic interoperability" and data spaces. The previous Gaia-X answer to my question about shouldn't there be a requirement to actually model the data products themselves as semantically interoperable structured data (preferrably first modeled with SKOS/RDF/OWL/SHACL and then serlialized using JSON-LD) was that "no, no, that is a domain specific matter" - the self-descriptions in the Gaia-X context are merely DCAT-AP type of stuff aka labels with metadata about the data product, not the actual detailed description of the data itself with the use of commonly defined semantics.

This "vocabulary" stuff is still a bit obscure to me - but if I've understood the recent DSSC Blueprint 1.0 > Data Modeling correctly, the data space communities (IDSA included) are actually moving towards the idea of sharing data as JSON-LD based on W3C Semantic Web data descriptions.

This would actually lead to the possibility of introducing data spaces that could decide that their data products MUST be defined and shared in the following way:

pick the underlying global or regional ontologies that you want to build your data understanding upon; for business data this could be something like the UN/CEFACT BSP Vocabulary or GS1 Web Ontology which seem to be standard W3C Linked Data
model your data space specific SKOS ontology and map it's concepts to a RDF/OWL data space core vocabulary
link the chosen underlying standards or vocabularies to your core vocabulary
define data products as SHACL shapes based on your own core vocabulary
create JSON-LD data products of your existing data sets, mapping your own data model to the core vocabulary of the data space
and finally - share the data product as a W3C Verifiable Credential (JSON-LD content wrapped in something like SD-JWT or similar format) using the upcoming EU digital wallets or similar
Not a real expert on these issues as Vladimir :-) but still experienced enough to feel that there is something more to "semantic interoperability in data spaces" than just "use DCAT-AP to describe your data product - which itself can then be basically anything like a PDF, JPEG or XML or whatever - that's not of interest for the data space - and you can share it by any technology you want, doesn't matter what, we don't care, it's the consumers problem, they'll know what to do"