MOD3.0 Design issue #1: What class of DCAT does a Semantic Artefact specialize?

[jonquet]

mod:SemanticArtefact subclass of both classes

• a resource (« something that can be catalogued »
• a dataset (« collection of data ») … in the sense a collection of classes, concepts, instances, properties as captured to represent the knowledge

[jonquet]

Properties coming from DCAT

[agbeltran]

In terms of defining mod:SemanticArtefact as a subclass of dcat:Resource and dcat:Dataset, I would point out that:

• it certainly makes sense that a mod:SemanticArtefactis adcat:Dataset so that we can make use of the dcat:distribution property to point to the specific representations of the semantic artefact (dcat:Distribution)
• while we can inherit from both classes, I would keep it simple and only inherit from dcat:Dataset and not dcat:Resource, as that relationship will be inferred anyway

So, I suggest as a decision for this design issue that we add

mod:SemanticArtefact
  a rdfs:Class ;
  a owl:Class ;
  rdfs:subClassOf dcat:Dataset ;

[jonquet]

I mostly agree with the formal aspect (it would be infer).
Originally I mostly used the diagrams to illustrate the idea. And it was important that one can see the fact that a SemanticArtefact (also called a Semantic Resource) is a Resource.

[antony-wilson]

I also like the idea of keeping it simple and just inheriting from dcat:Dataset

[agbeltran]

Just a note to link to the discussion on the design issue 2, which refers to the point of distributions above #35

[agbeltran]

To contribute to this discussion, the definition of dcat:Dataset is "A collection of data, published or curated by a single agent, and available for access or download in one or more representations." (see https://w3c.github.io/dxwg/dcat/#Class:Dataset)

[jonquet]

FAIRsFAIR discussed about 3 years ago the fact of considering a Semantic Artefact a Dataset and the decision was to do it in order to be able to rely on DCAT for the next iteration of MOD.
The discussion also happened later in FAIR-IMPACT but without another concrete solution.

[agbeltran]

Further to the discussion on the FAIR-IMPACT meeting this week (that I aim to summarise here when I get a chance), there is further discussion on this on the DCAT issue tracker: w3c/dxwg#1576

[mariapoveda]

I wouldn't agree on having all types of semantic artefacts as datasets while some are (like skos vocabularies) for the case of ontologies doesn't seem natural to be classified as datasets. I would remove the subclassOf between SemanticArtefact and Dataset

[agbeltran]

Summary of the discussion and approach:

We had multiple discussions on this design issue (FAIR-impact project meetings related to T4.2.3), the discussion above, as well as discussion in the DCAT issue w3c/dxwg#1576).

While some team members are happy with deriving mod:SemanticArtefact from dcat:Dataset, others prefer to distinguish between a dataset and a semantic artefact, so prefer to derive from dcat:Resource that is DCAT's extension point.

However, for semantic artefacts, we still want to take advantage of the dichotomy between abstract entity and distribution, so the suggestion is that we will still use the dcat:distributionproperty for semantic artefacts. In effect, this implies that dcat:SemanticArtefactcan be inferred as a dcat:Datasetin the current interpretation of DCAT. This seemed to be a good compromise with those who disagree to make the subclass explicit.

Thus, we will represent:

mod:SemanticArtefact
  a rdfs:Class ;
  a owl:Class ;
  rdfs:subClassOf dcat:Resource ;

and still use the distribution predicate to refer to multiple representations of the semantic artefact.

[agbeltran]

I updated the issue #44

I also note the point raised by @jonquet about the schema.org term for sdo:DefinedTermSet, see https://schema.org/DefinedTermSet, which again distinguishes betwee Dataset and a set of defined terms

#51

[agbeltran]

Closing this discussion as the conclusion is available as an issue: #44