The class Evidence should not be a subclass of dcat:Dataset

[andreea-pasare]

No description provided.

[bertvannuffelen]

Dear @andreea-pasare,

can you provide more context for your comment?

The motivation came from the consideration that an Evidence is a abstract notion for a collection of data. This data can be shared in different formats.

Since dcat:Dataset + dcat:distribution in an abstract way address that consideration, the WG accepted to make that explicit in the model using a subclass relationship.

The impact for implementations is that when additional properties are required by the implementation, a semantical check with DCAT must happen. It was our assessment that this would not pose major issues, but rather would facilitate the design of the implementation.

[costezki]

There are models. Some of them are considered metadata models because they describe other data in a particular way (e.g. DCAT for cataloguing purposes). So, generally, there are models for metadata and they are not the same as the models for the content of datasets (assuming that the datasets are in RDF, which may not necessarily be the case).

By bringing DCAT into the picture in CCCEV, we tend to push it into the category of metadata model. YET, I am not sure we want to view the CCCEV core vocabulary as a metadata model.

If we start from the premise that mixing content data and metadata is not always a good idea, then we need to carefully monitor what sort of metadata do we allow to be mixed with the "content" data.
DCAT is a cataloguing type of metadata model.

To my knowledge, instances of DCAT descriptions are NOT mixed with the content of the datasets they describe. (And, I would be very interested to learn of the contrary practice).

In eProcurement Ontology, we need to Re-use CCCEV vocabulary. If we accept that cccev:Evidence is a subclass of dcat:Dataset, we accept that a metadata model is instantiated and mixed with the ePO content model. Thus violating the premise above.

[bertvannuffelen]

@costezki and @andreea-pasare

I agree it might confusing, but here dcat:Dataset is used at a very finegrained level. And not to be mixed with the usage of dcat:Dataset in e.g. DCAT-AP.
Here we are at the level of an individual record in a register.

This usage of dcat:Dataset is maybe not the most common one, but it is not excluded from the semantics.
So we reuse here the class as-is from the DCAT vocabulary. And not the class in the context of the DCAT-AP profile.

Maybe we have to make this more explicit in the usage note.

[jimjyang]

Related to CPSV-AP, we have provided examples, at the descriptive level (metadata level, so to say), of being able to model Evidence as Dataset, and to model Output as Dataset.

In one way or another, in order to implement "once only" and to increase data sharing and reuse, we need to be able to "relate" Evidence and Output to Dataset.