Dxwg: Erratum: in DCAT 1.0 RDF dcat:keyword is a sub-property of dct:subject

[dr-shorthair]

In the RDF representation of DCAT v1, dcat:keyword is a sub-property of dct:subject - see https://www.w3.org/ns/dcat.ttl .
In the recommendation document this is not mentioned - see tabulation at https://www.w3.org/TR/2014/REC-vocab-dcat-20140116/#class-dataset
so this row in the description was omitted in the draft moving forward,

I think this is an unintentional omission, but perhaps worth checking with @makxdekkers and @aisaac if you have any insight? Else it should be added here https://w3c.github.io/dxwg/dcat/#Property:dataset_keyword

[andrea-perego]

I think it is probably more correct to say that dcat:keyword, whose range is rdfs:Literal, is a subproperty of dc:subject. By contrast, dct:subject (quoting) "is intended to be used with non-literal values" - as dcat:theme.

[makxdekkers]

Given the definition at DCMI, I don't think dcat:keyword can be a subproperty of dct:subject. The note with the definition means that dct:subject is an owl:ObjectProperty, while dcat:keyword is an owl:DatatypeProperty. I think a sub-property should be the same property class as its parent. I agree with @andrea-perego that it could be a sub-property of dc:subject.

[fellahst]

I think dct:keyword should remain as is. dct:subject typically refers to a skos:Concept and enable semantic search and discovery. Making dct:keyword subproperty of dc:subject will just add confusion. Almost no OWL ontology uses DC today. They are mostly using DCTerms ontology.

[andrea-perego]

I think dct:keyword should remain as is. dct:subject typically refers to a skos:Concept and enable semantic search and discovery. Making dct:keyword subproperty of dc:subject will just add confusion. Almost no OWL ontology uses DC today. They are mostly using DCTerms ontology.

@fellahst , I agree that the range of dcat:keyword (rdfs:Literal) should not be changed, but the statement making it a subproperty of dct:subject is formally incorrect, and it should be marked as an erratum / deleted.

[dr-shorthair]

Definition of dct:subject here: http://dublincore.org/documents/dcmi-terms/#terms-subject
Indeed, it says "This term is intended to be used with non-literal values", which would be inconsistent with dcat:keyword, which has an explicit range of rdfs:Literal.

So it appears there is another erratum in the RDF representation of DCAT 1.0 - the rec document appears to be correct, so the RDF file has an incorrect sub-property axiom.

@draggett @kcoyle @pwin another erratum?

[fellahst]

@andrea-perego I think you misunderstood my suggestion. I never said that dcat:keyword should be a subclass of dct:subject. It would be incompatible indeed.

[andrea-perego]

@fellahst said:

@andrea-perego I think you misunderstood my suggestion. I never said that dcat:keyword should be a subclass of dct:subject. It would be incompatible indeed.

Sorry for the misunderstanding.

[davebrowning]

Since we're nearing the approval process for DCAT2, should we not at least log this on the 2014 DCAT errata page at https://www.w3.org/2001/sw/wiki/DCAT_Errata ? [It looks like a wiki page editable by any W3 WG member though I haven't tried to save it yes]. SInce we will be providing a new dcat schema file it would be good to have some audit trail on the old file?

(I'm not sure of the process for actually correcting the file, since this isn't an erratum in the Recommendation which would need WG and AC approval. The schema makes it clear that the document has authority over the schema and not vice versa, but it would be nice to have a public erratum record.)

[andrea-perego]

@davebrowning , unless you think we should still deal with this DCAT 2014 erratum, I propose we close this issue.

[andrea-perego]

Noting no objections, I'm closing this issue.