search

opendata-swiss / dcat_ap_ch

Examples for geocat and DCAT data-catalogs are given here
5 stars 3 forks source link

rdfs:seeAlso is too similar to dct:relation and the defintion is contradicting itself #53

Open sabinem opened 3 years ago

sabinem commented 3 years ago

Property: rdfs:seeAlso Class: dcat:Dataset Conformance Problem: rdfs:seeAlso is a custom Swiss property to express relationships between datasets. DCAT already includes properties for this purpose, however. DCAT gives the choice between dct:relation and dcat:qualifiedRelation. In order to conform to DCAT, one of those properties MUST be used instead. Proposal: Abondon the property and use DCAT vocabulary dct:relation or dcat:Relationship instead.

metaodi commented 3 years ago

I support this proposal, but want to remind you that dct:relation is now being used for all sorts of "extra" information. I'm not sure if it's possible to add more information to the relationship (like "documentation", "further information", "related dataset"). If this is not possible, I would strongly suggest to keep this extra property.

p1d1d1 commented 2 years ago

I support this proposal, but want to remind you that dct:relation is now being used for all sorts of "extra" information. I'm not sure if it's possible to add more information to the relationship (like "documentation", "further information", "related dataset"). If this is not possible, I would strongly suggest to keep this extra property.

dcat:qualifiedRelation

andreasamsler commented 2 years ago

I support this proposal, but want to remind you that dct:relation is now being used for all sorts of "extra" information. I'm not sure if it's possible to add more information to the relationship (like "documentation", "further information", "related dataset"). If this is not possible, I would strongly suggest to keep this extra property.

I strongly support to keep this extra property because of the dependencies/situation @metaodi isreferring to.

p1d1d1 commented 2 years ago

IMHO dct:relation and dcat:qualifiedRelation are enough for this. I support the proposal to abandon rdfs:seeAlso

metaodi commented 2 years ago

@p1d1d1 but dcat:qualifiedRelation is currently not part of DCAT-AP Switzerland, correct? So you would propose to replace rdfs:seeAlso with dcat:qualifiedRelation?

p1d1d1 commented 2 years ago

@metaodi if we can avoid having a custom property, I'd say yes. How is currently used rdfs:seeAlso? Do you have some examples?

metaodi commented 2 years ago

@p1d1d1 see here Dataset: https://opendata.swiss/en/dataset/covid_19-massnahmen-im-kanton-zurich Turtle: https://ckan.opendata.swiss/en/dataset/covid_19-massnahmen-im-kanton-zurich.ttl

seeAlso is used to link to related datasets as intended by the current DCAT-AP Switzerland standard.

Juan-Juan-1 commented 2 years ago

@p1d1d1 "Description: Link to related datasets. Contains the identifier of the linked dataset." (see https://handbook.opendata.swiss/de/content/glossar/bibliothek/dcat-ap-ch.html?highlight=dcat%20ap%20ch#dcat-ap-ch-dataset). See examples above from @metaodi

p1d1d1 commented 2 years ago

For me that stuff is simply a dct:relation (or one of its sub-properties hasPart, isPartOf, references, isReferencedBy) or if you want to qualify the relation something like:

... dcat:qualifiedRelation[ a dcat:Relationship ; dct:relation http:.../673@fachstelle-ogd-kanton-zuerich ; dcat:hadRole http://www.iana.org/assignments/relation/related ] ;

sabinem commented 2 years ago

@metaodi I just want to add a link to a section of DCAT Version 3 with examples and a discussion on relationships between datasets and resource; https://www.w3.org/TR/vocab-dcat-3/#qualified-relationship.

I also want to com back to your example: (Dataset: https://opendata.swiss/en/dataset/covid_19-massnahmen-im-kanton-zurich, Turtle: https://ckan.opendata.swiss/en/dataset/covid_19-massnahmen-im-kanton-zurich.ttl)

See here how rdfs:seeAlso is implemented there: rdfs:seeAlso "671@gesundheitsdirektion-kanton-zuerich", "673@fachstelle-ogd-kanton-zuerich", "691@statistisches-amt-kanton-zuerich", "696@gesundheitsdirektion-kanton-zuerich", "699@gesundheitsdirektion-kanton-zuerich", "706@gesundheitsdirektion-kanton-zuerich", "710@gesundheitsdirektion-kanton-zuerich", "725@gesundheitsdirektion-kanton-zuerich", "728@gesundheitsdirektion-kanton-zuerich", "740@statistisches-amt-kanton-zuerich" ;

To me this is not understandable on the level of data and just makes sense in the context for opendata.swiss. Out side of it these would be just Strings. I think a datastandard should not need the context of a dataportal to be understood. To understand this statement above, you need to knowt hat from the above strings it is possible to derive the perma links of the datasets (a knowledge that comes from the dataportal)

If that Swiss property is kept, I think it at least its definition should be revised https://www.dcat-ap.ch/releases/2.0/dcat-ap-ch.html#dataset-see-also, so that the range is dcat:Dataset.

metaodi commented 2 years ago

@sabinem 100% agree, those strings should be URIs to begin with, and having more information about the relationship as @p1d1d1 suggested makes sense.

My point is, that we should keep the functionality of linking to related datasets. If another attribute is better suited for that than rdfs:seeAlso, then let's use another one.

sabinem commented 2 years ago

@metaodi I would suggest to use the property dct:isReferencedBy that is an optional property of dcat:Dataset instead of rdfs:seeAlso:

<https://ckan.opendata.swiss/perma/707@statistisches-amt-kanton-zuerich> a dcat:Dataset ;
  rdfs:seeAlso "671@gesundheitsdirektion-kanton-zuerich" , 
                       "673@fachstelle-ogd-kanton-zuerich" .

would then become:

<https://ckan.opendata.swiss/perma/707@statistisches-amt-kanton-zuerich> a dcat:Dataset ;
  dct:isReferencedBy <https://ckan.opendata.swiss/perma/671@gesundheitsdirektion-kanton-zuerich"> ,
                                  <https://ckan.opendata.swiss/perma/671@gesundheitsdirektion-kanton-zuerich"> .

To me an approach as above would make much more sense: especially using the perma links instead of the identifiers. One could argue against dct:isRefrencedBy and replace it by a subproperty of dct:relation instead. But I don't really why we should not go with this already existing property of DCAT-AP. What do you think about this?

metaodi commented 2 years ago

I agree, I just have two things to add: