Re: Proposal for new type : Vocabulary

Hi Bernard,

I think we agree on many points. It's possible to have several types, arranged into an hierarchy if needed. And schema:additionalType could be used for typing the "vocabularies".
We also need an isDefinedOf that is more general than the current RDFS one. And yes, I can see vocabularies as collections: we talk about the "art" of vocabulary crafting and re-use, after all ;-)

Also: the line is blurred because some resources, like the elements of LoC relator are of both rdfs:Property and skos:Concept types. *But* one will find and re-use LoC relators either as an Ontology or a skos:ConceptScheme, not as one or the other, i.e. a generic "vocabulary" they wouldn't know the purpose of.
Such artifacts are (knowledge) engineering artifacts, the form is essential to the use. What's the case where people would be searching for such things, without knowledge and interest for the specific goal they serve??
The knowledge that these are "vocabularies" doesn't say anything on what one can do with them: this comes only with the lower-level info, the one of OWL and/or SKOS.

Note that if you see a point in a blurred level, I can make it even more blurry: as we had discussed in the Library Linked Data incubator [1], a (SKOS) vocabulary is also not far from reference datasets that don't show vocabulary flavour and yet can be used in many same scenarios as the SKOS concepts. Especially datasets like DBpedia, or Geonames, as you point out in your email. and in a way many ontologies do include knowledge base-style axioms...

Antoine

[1] http://www.w3.org/2005/Incubator/lld/XGR-lld-vocabdataset-20111025/#Introduction:_Scope_and_Definitions



On 12/9/13 12:45 PM, Bernard Vatant wrote:
> Hello Antoine and Chaals
>
> 2013/12/7 Antoine Isaac <aisaac@few.vu.nl <mailto:aisaac@few.vu.nl>>
>
>     Hi Chaals, all (I guess there was a problem in my email's receipients)
>
>     I think I'm in favour of having owl:Ontology, and if there's a need to get an umbrella for glossaries/thesauri/other classifications, to re-use skos:ConceptScheme or something like that.
>
>
> As you write yourself, owl:Ontology and skos:ConceptScheme are different animals, both specified in such a way that they lead to a variety of different interpretations, as we have learnt in the LOV project, but in any case specified for specific types of vocabularies.
>
>     Diane says there are use cases, but http://metadataregistry.org/ still makes a big difference between "Vocabularies"  (there, akin to SKOS ConceptSchemes) and "Element Sets" (there, akin to OWL Ontologies).
>
>
> Yes indeed. But having a general type in schema.org <http://schema.org> allows specific resources to have their specific subtypes, named in their local parlance. If I understand well Chaals' point, I agree. From a search engine viewpoint (which is all schema.org <http://schema.org> is about), users will certainly look for terms without specifying if they are looking for a skos:Concept, a rdfs:Class, a wordnet:Synset etc or just a word in a flat glossary not specified in any RDF dialect. I would see owl:Ontology etc fit in additionalType
>
>     In fact Bernard has not answered my question on starting to have everything in LOV.
>
>
> This is a rather different issue. We have restricted so far the scope of LOV to RDFS and OWL ontologies in order to stay focus, cope with limited resources etc. For example the VOAF constructions and queries computing links between vocabularies are taking into account so far RDFS and OWL constructions, not SKOS mapping relationships for instance.The next version of LOV on which we are working will certainly be more open in scope, but in any case the schema.org/Vocabulary <http://schema.org/Vocabulary> proposal goes beyond such considerations. It could be used for non-RDF vocabularies, which will stay in any case outside the scope of LOV.
>
>     To be more precise: I do not see the the use case for having *only* the generalization of Ontology and ConceptScheme. I see the point in making inventories of both types of resources; but because the two categories are usually made to meet quite different requirements, I'm skeptical about the value of blurring the line.
>
>
> The line is already blurred anyway, as examples I mentioned are showing. LoC relators are both skos:Concept and rdf:Property. Th Geonames ontology includes the feature codes as SKOS concepts etc.
>
>     About rdfs:isDefinedBy vs an hypothetical "uses" property (between Vocabulary and individual classes/properties or concepts). I agree with Chaals: the differences may not survive in this case. More precisely: I think it is the "usage" link between vocs and their elements that will get more traction. The "defined" link is more a provenance/process aspect; it's important for specific scenarios, which may not be really sought by the general schema.org <http://schema.org> users. Therefore it would seem dangerous to me to use rdfs:isDefinedBy, because its original meaning would be 'highjacked' by the more prominent requirement.
>
>
>   Indeed. So you agree with a generic "isDefinedBy". Or  reuse http://schema.org/isPartOf which has a very restricted use so far. Unless Vocabulary can fit under http://schema.org/CollectionPage. A Vocabulary is after all somehow a collection of terms as an ImageGallery is a collection of images, but this seems a bit far-fetched ...
>
>     Antoine
>
>     On 12/6/13 9:46 PM, Charles McCathie Nevile wrote:
>
>         On Fri, 06 Dec 2013 11:19:06 +0100, Antoine Isaac <aisaac@few.vu.nl <mailto:aisaac@few.vu.nl>> wrote:
>
>             Hi Bernard,
>
>             +1 for including Vocabulary. It would fill in the gap of not having included an equivalent to skos:ConceptScheme in the MiniSKOS proposal.
>             I'm however a bit skeptical about having just one class for ontologies (sets of classes and properties a la schema.org <http://schema.org>) and other SKOS-level vocabularies (classifications, thesauri) in one big bag. I'll be the first one to agree with you that the two types of vocabularies are not exclusive. But the functions are quite different.
>
>
>         I'm not so sure - and I certainly don't know of a practical way to describe the difference to non-professionals.
>
>
>             In fact trying a litmus test: if the "Vocabulary" class was to include SKOS Concept schemes, would the Linked Open vocabularies start gathering SKOS concept schemes as well?
>
>
>         That would seem like a reasonable thing to me. Although they might be selective about them for other reasons.
>
>
>                 An extra would be to have a "definedBy" property to link instances of the oncoming Topic class to an instance of Vocabulary.
>
>             Is this skos:inScheme in the SKOS world, or really rdfs:isDefinedBy?
>             The nuance (creating vs. including/re-using) can be in fact really important. (and I personally feel that both aspects are worth representing).
>
>
>         The problem is that in practice, where a term is originally defined stops being relevant once people start including it in other places with differences, and getting traction.
>
>         I have nothing in principle against preserving the differences, but I am sceptical that they will survive contact with reality, and not sure they are worth struggling for.
>
>         cheers
>
>         Chaals
>
>
>
>
>
> --
> *Bernard Vatant
> *
> Vocabularies & Data Engineering
> Tel : + 33 (0)9 71 48 84 59
> Skype : bernard.vatant
> http://google.com/+BernardVatant
> --------------------------------------------------------
> *Mondeca*****
> 3 cité Nollez 75018 Paris, France
> www.mondeca.com <http://www.mondeca.com/>
> Follow us on Twitter : @mondecanews <http://twitter.com/#%21/mondecanews>
> ----------------------------------------------------------

Received on Monday, 9 December 2013 13:47:58 UTC