W3C home > Mailing lists > Public > public-esw-thes@w3.org > November 2009

Re: [Dbpedia-discussion] Using DBpedia resources as skos:Concepts?

From: Richard Cyganiak <richard@cyganiak.de>
Date: Thu, 12 Nov 2009 21:55:43 +0000
Cc: Antoine Isaac <aisaac@few.vu.nl>, Pat Hayes <phayes@ihmc.us>, SKOS <public-esw-thes@w3.org>
Message-Id: <350B9696-5B52-4664-896C-4977AADEF1B6@cyganiak.de>
To: Bernard Vatant <bernard.vatant@mondeca.com>
Bernard,

Thanks for summing up the thread so nicely! And apologies for anyone  
who might be bored by this thread by now ...

So you have me convinced that my skos:Concept labelled "Michelle  
Obama" is a "library business object" which is a different entity from  
Michelle Obama the living person. It follows that an owl:sameAs  
statement between the skos:Concept and the foaf:Person is  
inappropriate, because it conflates creation dates of different  
entities.

On 12 Nov 2009, at 15:13, Bernard Vatant wrote:
> OTOH, looking with librarian glasses on, DBpedia entries are also  
> somehow "concepts", since they could also bear Dublin Core stuff  
> like the first version of DBpedia in which they have been published,  
> number of related DBpedia concepts ... or any librarian workflow  
> record of the same kind. In this case, Richard could assert a  
> skos:exactMatch relationship between his entry and the DBpedia one.

I'm intrigued. So could I say the following:

    <http://mydataset/433256>
        a skos:Concept;
        skos:closeMatch <http://dbpedia.org/resource/Michelle_Obama>.

Note that the DBpedia resource *is* Michelle Obama the living person,  
typed as a foaf:Person, and bearing properties such as :spouse  
and :child. But I will just ignore that, just to see what rat's nest  
that will get me into.

Because of the range of skos:closeMatch, the triples above imply the  
following:

    <http://dbpedia.org/resource/Michelle_Obama>
        a foaf:Person;
        a skos:Concept .

So this would mean that there exists a concept that shares all  
properties (including creation date) with the person. I'm assuming  
that DBpedia does not assert any SKOS properties about the resource.

Can a single URI refer to a person and to an equivalent concept at the  
same time?

Anyone care to shoot holes into that idea?

Richard


>
> But if the DBpedia entry is supposed to represent the thing herself,  
> as the predicates it bears, and the excellent (as usual) arguments  
> of Pat, seem to prove, we're back to the initial question : what  
> predicate should poor Richard (and myself) use if we want  
> nevertheless to express the fact that his specific record/entry/ 
> heading/concept is a proxy in his system for the person Michelle  
> Obama? We definitely miss, as I have kept stressing a lot of times  
> in the past, a specific vocabulary to indicate this level of  
> indirection, as also suggested in Danbri's answer.
> And a side question is, if the DBpedia URI represents the thing  
> itself and not a DBpedia proxy of it, where do I put the above  
> workflow information? The date of publication of the DBpedia entry  
> vs the birth date of Michelle Obama ...
>
> Bernard
>
>
> 2009/11/8 Antoine Isaac <aisaac@few.vu.nl>
> Pat Hayes a écrit :
>
>
> On Nov 5, 2009, at 2:51 PM, Antoine Isaac wrote:
>
> Hi Kingsley,
>
> when specifically does one use "skos:exactMatch" etc? Based on my  
> response John (few minutes ago), I am assuming that the partitioning  
> of so called Named Entities and Subject Matter Concepts was the line  
> of delineation sought in SKOS which is about Subject Matter/Heading  
> style Concept Schemes?
>
>
> Well, as Leonard just put it, there can perfectly concepts for  
> persons (as part of authority files in Libraries, for example).
> You could have a skos:Concept Mrs_Obama with a Library A as  
> dc:creator, and another Concept Michelle_Obama created by Library B.  
> Using owl:sameAs between those is certainly not ideal, as you end up  
> with one resource being created by two different agents, and  
> probably at different times, and so on. exactMatch fits that case.
>
>
> If I follow this and your previous post, then sameAs will almost  
> never be true in the SKOS world, correct?
>
>
> Yes.
>
>
> The only case I can think of would be where a library puts a new  
> indexing system in place (for its existing records), and retains the  
> old one for legacy reasons, with sameAs links between the old and  
> new indices.
>
>
> Not even then... If C1 is the legacy concept, and C2 the new one,  
> they would have different management info, perhaps different notes/ 
> definitions attached to them, semantic relations (broader/narrower/ 
> related) to different concepts...
> So I would advise against using sameAs in such a concept.
> sameAs would be rather used in case where different identification  
> schemes have been created for one concept, as in [1]
>
>
> That is, co-reference between items in thesauri (both referring to  
> FLOTUS) is irrelevant to identity of the *concepts*.
>
> Do I have this more or less right?
>
>
> Yes. Of course one can argue that it is a necessary condition, but  
> not a sufficient one.
> By the way, Pat, I hope that the example [1] can shed a bit more  
> light in your quest for what a SKOS concept is. Especially you can  
> look at how the dcterms:created and dcterms:modified are used.
> I guess that the general idea of Murano glass did not come into  
> existence at 1986-02-11T00:00:00-04:00.
> SKOS concepts are very practical entities, almost  
> "documents" (albeit very specific, controlled documents) about more  
> general ideas. And of course these things can exist for persons, eg.  
> [2].
>
> Antoine
>
> [1] http://id.loc.gov/authorities/sh85055118.rdf
> [2] http://stitch.cs.vu.nl/vocabularies/rameau/ark:/12148/cb11944615b
>
>
>
>
>
> -- 
> Bernard Vatant
> Senior Consultant
> Vocabulary & Data Engineering
> Tel:       +33 (0) 971 488 459
> Mail:     bernard.vatant@mondeca.com
> ----------------------------------------------------
> Mondeca
> 3, cité Nollez 75018 Paris France
> Web:    http://www.mondeca.com
> Blog:    http://mondeca.wordpress.com
> ----------------------------------------------------
Received on Thursday, 12 November 2009 21:56:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:39:05 GMT