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

Pat Hayes a écrit :
> 
> On Nov 6, 2009, at 1:31 PM, Simon Spero wrote:
> 
>> On Fri, Nov 6, 2009 at 11:58 AM, Pat Hayes <phayes@ihmc.us 
>> <mailto:phayes@ihmc.us>> wrote:
>>
>>
>>     On Nov 5, 2009, at 4:05 PM, Simon Spero wrote:
>>
>>     FWIW, I have no trouble with imaginary entities. Still, there is a
>>     clear distinction between the concept of a unicorn and a
>>     particular unicorn, eg the one depicted here: http://bit.ly/3Hgz0P
>>
>>  
>> [...]
>>
>>>     Once one starts thinking extensionally this whole discussion
>>>     becomes much easier ("Word and Subject?").
>>>
>>>     For example:
>>>
>>>     Everything that is-about something is a document.
>>>     Everything that something is-about is a concept.
>>
>>     My problem is that this second assertion is blatantly false. I
>>     have shelves full of books that are not about concepts at all.
>>     Biographies are about people, not (usually) concepts of people. So
>>     at this point, SKOS simply vanishes into never-never land. I have
>>     no idea what it is talking about (quite literally). 
>>
>>
>> [To clarify, I am talking about standard Knowlege Organization System 
>> (KOS) semantics, not SKOS directly].
> 
> OK, thanks. 
> 
>>
>> The second assertion is an axiom...
>>
>> The problem we're having here is that the word "Concept" has different 
>> meanings in different disciplines.  
> 
> True, though I think Im using it in its normal English sense:
> 'an abstract or general idea inferred or derived from specific instances '
> 'Something understood, and retained in the mind, from experience, 
> reasoning and/or imagination'
> 'a general notion around which ideas are developed'
> 
> Notice the "mental" emphasis of these common definitions.
> 
>> An alternative term used in the Knowledge Organization literature is 
>> "Subject".  That term can lead  to even worse confusion, especially in 
>> the context of RDF, but is used to good effect by Elaine Svenonius in 
>> the following quote:
>>
>> Subject language terms differ referentially from words used in 
>> ordinary language. The former do not refer to objects in the real 
>> world or concepts in a mentalistic world but to subjects. As a name of 
>> a subject, the term Butterflies refers not to actual butterflies but 
>> rather to the set of all indexed documents about butterflies. 
>> (Svenonius 2000, p. 130) 
> 
> OK, a set of documents is something I can understand. That would be a 
> class in an ontology of documents. 
> 
>>
>> This is an allusion to Leonard Cohen's "How to speak poetry":
>>
>> The word butterfly is not a real butterfly. There is the word and 
>> there is the butterfly. If you confuse these two items people have the 
>> right to laugh at you.
> 
> Kozybski: the map is not the territory. Right, exactly. 
> 
>>
>> The importance of Svenonius's  distinction can be seen by considering 
>> the relationship between two Subjects. Let's stick with our examples, 
>> and choose the strings "Unicorns" and "Pictures of unicorns".
>>
>> As ordinary language, these strings refer to  different "kinds" of 
>> things.  One refers to the set of horses with horns;  the other refers 
>> to the set of pictures of horses with horns. These sets are completely 
>> disjoint.
>>
>> Now consider what these strings refer when treated as subjects.  One 
>> string refers to the set of all documents about horses with horns;  
>> for example, the novel "The black unicorn".  The other strings refers 
>> to the set of all documents about pictures of horses with horns; for 
>> example, a wiki page containing a list of freely usable pictures of 
>> unicorns.
>>
>> These two sets are both sets of documents.  Not only are the two sets  
>> not disjoint; the second set is a subset of the first.
>>
>> A similar relationship can be seen between the strings "Horses" and 
>> "Diseases in horses".
> 
> Point taken, and I quite understand. But then the SKOS documentation is 
> woefully unclear on what it is (presumably?) intended to mean. By using 
> the word 'concept' and the phrase 'unit of thought' it seems to claim 
> much broader applicability than this tidy library-science image would 
> suggest. Which is why it is even being discussed in the same breath as 
> dbpedia, I presume.


Yes, the applicability of SKOS is indeed wider than the very nice class/subclass example. The problem is that not all practioners in the KOS field have read Svenonius and designed and used their thesauri in accordance.
When you ask them whether a concept can be seen as a class of documents, then most of them will say yes. But if you ask them whether the set of documents associated with a super-concept should blindly contain all documents associated to its sub-concepts, then the reactions are more mitigated. From "yes, certainly" to "not at all". 

So in the end, if we want to get that sort of legacy KOS data on the SW, we have to make compromises. The documentation has to accomodate a wide diversity of situations, not to specify an ideal view on what KOSs should be.
And believe me: I was in fact among the ones who embarked that second SKOS effort, thinking that we could put great and precise semantics (and a lot of OWL axioms, as a metter of fact) in it...

Antoine

Received on Sunday, 8 November 2009 17:52:20 UTC