W3C home > Mailing lists > Public > public-esw-thes@w3.org > May 2008

Re: skos:Concept and "real world things"

From: Peter Ansell <ansell.peter@gmail.com>
Date: Fri, 16 May 2008 12:01:10 +1000
Message-ID: <a1be7e0e0805151901w5948c1b2v49f4644aed5cb24e@mail.gmail.com>
To: "François-Paul Servant" <francois-paul.servant@renault.com>
Cc: public-esw-thes@w3.org, fps <fps@semanlink.net>

2008/5/13 François-Paul Servant <francois-paul.servant@renault.com>:
>
> Hi,
>
> In the "Linked Data" community, people give URIs to "real world things",
> such as animals, cities or persons.
>
> In SKOS, we define and give URIs to concepts, "ideas or meanings that are
> unit of thoughts..."
>
> It seems obvious that there is a deep (and possibly complex relation)
> between skos:Concept(s) and "real world (physical) things" (I would even say
> that, except maybe in very particular domains such as mathematics, there is
> no way to define a concept without relating it to "real world things").
>
> For instance, there is a relation between the skos:Concept "ex:platypus"
> declared in a given SKOS scheme and the funny australian animal which was
> given the URI "dbPedia:Platypus" by the dbPedia project:
>
> skosex:platypus p dbPedia:Platypus
>
> What kind of property  should be used for "p" in that statement?
>
> We can think of:
> 1) owl:sameAs
> 2) skos:exactMatch
> 3) moat:meaningURI
> 4) dc:subject
> 5) ...
>
> 1) doesn't seem OK (for reasons that include the reasons why we do not use
> owl:sameAs to state that 2 skos:Concepts are the same)

I don't see why you wouldn't use this if your ontology was utilising
names in a compatible way to Wikipedia where articles (hence URI's)
have always been about real world concepts, unless otherwise stated in
the text of the article. If you are just using the letters "p + l + a
+ t + y + p + u + s" as a knowledge organisation tag than it won't be
compatible as you are explicitly not saying anything about the concept
that could be proved or disproved with scientific methods, other than
you are presumed to have made it up in your head and are utilising it.
If you are attempting to operate at that level than sameAs wouldn't
have very good semantics.

> 2) implies that dbPedia:Platypus is a skos:Concept. DBPedia doesn't state
> that, and it is controversial to say that a real world thing is a conceptual
> resource. SKOS doesn't seem to want to allow such statements: in the wiki
> [1] we can read:
> "So, for a resource of type skos:Concept, any properties of that resource
> (such as creator, date of modification, source etc.) should be interpreted
> as properties of a concept, and not as properties of some 'real world thing'
> that that resource may be a conceptualisation of."

This is the biggest reason why I am avoiding using skos in my own
work. Even skos:related, which could be a potential term for
indicating a relationship, relies on one accepting that everything
will be compatible with skos:Concept.

> 3) MOAT ("Meaning Of A Tag" [2]) relates tags to their meaning. The meaning
> of a tag is an instance of a class "Meaning" (that probably could be
> considered as a skos:Concept). A dedicated property moat:meaningURI is used
> to link the "Meaning" to a "real world thing". We have something like this:
>
> tag:platypus moat:hasMeaning moat:meaning_platypus
> moat:meaning_platypus moat:meaningURI dbPedia:Platypus

This is in a sense an indirect assertion similar to that of the skos
Concept being used just as a knowledge organiser, again, if that is
what you desire than use it.

> 4) seems weak

It isn't utilised for realistic terms. It is more used to identifier
what a document is conversing with relation to.

> 5) any better idea ?
>
> The pattern used in [3] (MOAT) seems OK to me: we have on one side the
> "Concept", and on the other the "real world thing". This allows to write
> statements about the two entities, which are indeed distinct, and at the
> same time allows to clearly define a concept, when existing URI of a real
> world thing already exists.
> However, I would'nt feel bad with solution 2 (I am ready to consider real
> world things as concepts). But of course, my preference is 5 ;-)

If you aren't quite ready to accept the skos, clear assumptive
statement that reality doesn't exist within the skos:Concept world
then maybe okay. However, you would be unpredicatably utilising it to
imply that Wikipedians aren't really talking about real world things,
which wouldn't enforce the implication that dbpedia resource is a
class of skos:Concept on your users. Even if they didn't understand it
you would have to accept that solution 2 is not quite right and it may
interfere with their practical work if they state somewhere that
something is disjoint with skos:Concept (implying that it isn't a
non-real world concept and hence unuseful in skos terms)

> Anyway, shouldn't SKOS define a property allowing to state the relation
> between a skos:Concept and the real world thing that that skos:Concept is
> the conceptualisation of?

That would be outside of the bounds of their self-imposed system.
Interestingly, skos is explicitly implicated via MOAT dependencies if
you look hard enough so it violates skos:Concept already and hence you
wouldn't be the first person to violate their statement. ;-)

Peter
Received on Friday, 16 May 2008 02:12:45 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:38:59 GMT