W3C home > Mailing lists > Public > public-esw-thes@w3.org > December 2006

Re: Are skos:Concept and rdfs:Class disjoint?

From: Richard Cyganiak <richard@cyganiak.de>
Date: Fri, 15 Dec 2006 17:10:43 +0100
Message-Id: <48D4E74A-3C4F-47B8-9EB3-1195BAD65D78@cyganiak.de>
Cc: Dan Brickley <danbri@danbri.org>, public-esw-thes@w3.org
To: Bernard Vatant <bernard.vatant@mondeca.com>


On 15 Dec 2006, at 00:38, Bernard Vatant wrote:
> To complete Dan's answer, something I'm munching over those days is  
> that what's important with a RDF resource is more what one wants to  
> use it for (its functionality), than what it stands for (its  
> denotation). A skos:Concept is forged to index and search documents  
> and other information resources (maybe even philosophers - see  
> other thread about skos:subject), it's a librarian's tool. A  
> rdfs:Class is forged to sort and find out stuff outthere by  
> properties, it's a naturalist's tool, so to speak. So basically  
> they are different tools for different purposes, and as such,  
> should be kept distinct.

One more instance of the old "it's different things" hand-wave ... ;-)

> Now of course we have this permathread about how to link my-dog- 
> concept to your-dog-class, if they are indeed kept distinct. I  
> could again push the blank node connection here, but well ... In  
> the same spirit than above, I prefer to ask : what do you want this  
> for? Beyond the conceptual exercise, what is the use case? I ask  
> because honestly, despite an interest for this question which is  
> close to maniac obsession, I have not yet found a real-world, clear  
> business use case where such a requirement is in the critical path.

Here is my real-world business use case:

We have a large corpus of texts and are extracting "real" RDF  
resources from them -- e.g. if a text talks about Fido the dog, then  
we want an RDF resource representing Fido. This is the context.

Now the texts have been categorized into a very neat taxonomy. The  
Fido text might be classified under "Dogs". This categorizing has  
been a major investment. I have to figure out how we can use it to  
improve the generated RDF.

So I have a resource dog:Fido. I know it's related to a resource  
taxo:Dogs which is a skos:Concept. Thus my questions from this and  
the other thread:

1) Can I say that dog:Fido skos:subject taxo:Dogs?
2) Can I say that dog:Fido rdf:type taxo:Dogs?
3) Or do I have to go through intermediate resources to connect the two?

The third option is a non-starter because the project lead will not  
accept a solution that relies on some kind of intermediate resource  
hierarchy parallel to the other resources. (The way our  
infrastructure works, each new kind of resource is quite expensive.)

So, my choice is 1) or 2) or ignore the taxonomy for our RDF output.

Which option would you choose?

Richard


>
> Bernard
>
> Dan Brickley a écrit :
>>  Richard Cyganiak wrote:
>>>
>>> Another quick question:
>>>
>>> Can a skos:Concept be an rdfs:Class at the same time? For  
>>> example, if I have in my taxonomy a skos:Concept ex:Dog, then  
>>> would it be OK to also make this an rdfs:Class and say something  
>>> like this?
>>>
>>>     :ginger a ex:Dog .
>>>
>>> I suppose the answer is no because an RDFS class is something  
>>> else than a taxonomic concept. I'm interested in an explanation  
>>> that is a bit less hand-wavy than "it's something else."
>> How about: "I guess we could, but it would upset the OWL-DL  
>> constituency because it intermingles the ontological and instance  
>> data layers"?
>>
>> Some but far from all SKOS Concepts are, more or less, categories,  
>> ie. classes. Whether we indicate this in the Semantic Web by  
>> simple identity (ie. have the self-same thing simply be a Class  
>> and a Concept) , ... or whether we indicate this by named  
>> relationship (util:hasClass), is I think something still up for  
>> discussion. It is related to the question of how we indicate which  
>> SKOS Concepts "stand for" specific individuals, eg. a person, a  
>> place, or  event. I would be dissapointed if we answered those two  
>> questions separately, since it is the same core question: how does  
>> the (indirected, lowercase-r reified) SKOS worldview relate to the  
>> vanilla RDF/OWL worldview. The former is in terms of concepts, eg.  
>> the-concept-of-dogs, the-concept-of-fido; the latter is in terms  
>> of named classes, relationships and members of those classes: the  
>> class Dog, and the individual "fido" who is a thing in the class  
>> "Dog". In the RDF view, ... we get to ascribe arbitrary properties  
>> to Fido. In the SKOS view, we need to be careful when talking  
>> about individuals, since a SKOS concept for fido has different  
>> properties (creation date, for eg) than the thing it is the  
>> concept of. This is clearer in the case of individuals than in the  
>> case of classes.
>>
>> cheers,
>>
>> Dan
>
> -- 
>
> *Bernard Vatant
> *Knowledge Engineering
> ----------------------------------------------------
> *Mondeca**
> *3, cité Nollez 75018 Paris France
> Web:    www.mondeca.com <http://www.mondeca.com>
> ----------------------------------------------------
> Tel:       +33 (0) 871 488 459
> Mail:     bernard.vatant@mondeca.com  
> <mailto:bernard.vatant@mondeca.com>
> Blog:    Leçons de Choses <http://mondeca.wordpress.com/>
>
>
Received on Friday, 15 December 2006 16:10:55 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:38:55 GMT