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

Re: Using DBpedia resources as skos:Concepts?

From: Antoine Isaac <aisaac@few.vu.nl>
Date: Wed, 04 Nov 2009 08:37:47 +0100
Message-ID: <4AF12F4B.9080001@few.vu.nl>
To: Richard Cyganiak <richard@cyganiak.de>
CC: dbpedia-discussion@lists.sourceforge.net, SKOS <public-esw-thes@w3.org>, Pat Hayes <phayes@ihmc.us>, Dan Brickley <danbri@danbri.org>
Hi Richard,

I've recently been caught in discussions on the fact that it was not very good to have "universals" (in your case, concepts) mixed with "particulars" (in your case, persons). But to tell the truth, in the SKOS Recommendation there is no axioms stating that skos:Concept is distinct from foaf:Person. So there is no contradiction. I'll let you judge, maybe Dan and Pat can say more of it. 

In fact there is an ontologically cleaner solution, namely to create some intermediate skos:Concept and use a specific property to link that concept to the physical entity it is intrinsically associated to. There was a suggestion (skos:it), again Dan may say something about it. But we never had the time to address that issue properly, so that was not included in the standard. One thing is sure, it would make all the paths to access the relevant data elements a bit longer.

About the skos:exactMatch vs. owl:sameAs issue. Your argument against skos:exactMatch is not really right. If you are using owl:sameAs, you are asserting that dbpedia:Michelle_Obama is a skos:Concept too, by virtue of owl:sameAs's semantics. At least if using skos:exactMatch you are not transfering other statements, e.g., if you have one that says that you created http://mydataset/433256 (using dc:creator), then according to owl:sameAs semantics you would have created dbpedia:Michelle_Obama as well.
There's a nice example of such not-so-good practice at the NYT dataset released last week (which is otherwise an excellent work!). There they have something like:
  <rdf:Description rdf:about="http://data.nytimes.com/64870337666324078863">
    <dc:creator>The New York Times Company</dc:creator>
    <dcterms:rightsHolder rdf:datatype="http://www.w3.org/2001/XMLSchema#string">The New York Times Company</dcterms:rightsHolder>
    <owl:sameAs rdf:resource="http://dbpedia.org/resource/John_Frankenheimer"/>

By stating those triples, they in fact say (for those who will apply the owl:sameAs semantics on the dataset) that they created the Freebase concept, and hold the rights for it. when using owl:sameAs, you would have to be careful about that. At least with the exactMatch solution you are a bit safer.

(I'm sorry this is maybe not the solution to your issues, but I have no time now for more than giving you caveats...)



> Hi,
> I want to get some broader feedback and opinion on this question.
> There is a use case where an external dataset is modelled in SKOS, and 
> we want to map it to DBpedia. Something like:
> <http://mydataset/433256>
>     a skos:Concept;
>     owl:sameAs <http://dbpedia.org/resource/Michelle_Obama>;
>     .
> I've used owl:sameAs here. Now the problem is that my:433256 is a 
> skos:Concept, while dbp:Michelle_Obama is a foaf:Person. I wonder wether 
> that's a problem. I can't see any immediate contradiction arising from 
> that, but I'm uncertain.
> Another option would be to use skos:closeMatch or skos:exactMatch, but 
> these are intended for use between skos:Concepts, while I'm trying to 
> connect a skos:Concept to a foaf:Person.
> The main question, I guess, is wether people in the DBpedia project 
> would consider the assertion
>     <http://dbpedia.org/resource/Michelle_Obama> a skos:Concept .
> to be acceptable or erroneous, or wether SKOS folks tell me that 
> skos:Concept is obviously disjoint from foaf:Person.
> I know that this is a complex issue, so I'm not really looking for a 
> "right" or "wrong" answer. I'm more interested in getting all the pros 
> and cons and pitfalls and caveats on the table, so please, if you have 
> any opinion on the issue, I want to hear it.
> I'm cc'ing Pat Hayes, because he said he's interested in that kind of 
> question.
> All the best and thanks for your time,
> Richard
Received on Wednesday, 4 November 2009 07:38:29 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:45:59 UTC