Re: [foaf-dev] FOAF, geonames, and more

Hi,

Thanks for both of your answers. Yet, and sorry to bother, things are
not totally clear for me.  Well, actually, they became clearer while I
wrote this mail, but I'd like to be sure of that:

* If I understood well what Benjamin means:
foaf:based_near rdfs:domain geo84:SpatialThing
xxx foaf:based_near yyy
=>
xxx a geo84:SpatialThing

(which is rdfs2 entailment rule [1] ?)
Yet, it can involve coherence problems, that can be checked only with
an OWL reasoner and appropriate rules ? (i.e. to identify that xxx is
instance of 2 disjoint classes). Am I right until now ?

Reading Bernard's answer: I have:
> (1) "If a resource R is (of class X) in the domain of P, then you can
> attach a P-value to R".
Ok for this one.

> wheras the "sufficient" aspect
> (2) "If a resource R has a P-value, then R belongs to (some class in)
> the domain of P"
Which is what Benjamin said before ?

But:
> (3) "If a resource R is (of class X) not in the domain of property P,
> then you can't attach a P-value to R"
Means that if
xxx is not a geo84:SpatialThing => cannot have xxx based_near yyy

Yet, while writing, I'm wondering if there, "not" means "xxx is
explicitely stated as not a geo84:SpatialThing (i.e. instance of a
disjoint class)" rather that "we don't have any information if xxx is
a geo84:SpatialThing" - because of the open-world assumption.
In that case, I think that's clear for me :)

So, as soon as I define RDFS / OWL properties, I have:
p rdfs:domain xxx
aaa a xxx
=>
aaa p yyy
(necessary)

p rdfs:domain xxx
aaa p yyy
=>
aaa a xxx
(sufficient)

Then, I must deal with possible incoherence using OWL constraints /
axioms, is that right ?

Thanks a lot for your help :)

Alex.

[1] http://www.w3.org/TR/rdf-mt/#RDFSRules

> (1) and (3) are used to control edition interfaces, whereas (2) will
> need a reasoner/classifier. The former are more frequent and familiar
> than the latter, hence (2) is sometime forgotten or misunderstood.
>
> Bernard
> > On 28.01.2007 17:46:28, Alexandre Passant wrote:
> >
> >> Right, but what about using based_near for something that is not a
> >> spatial thing ?
> >> The only way to do is to assert this is a Spatial Thing, isn't it ?
> >>
> > I can't think of a non-spatial thing that's based near something
> > (unless you use "near" in a temporal sense).
> >
> >
> >> But since foaf:Organisation is not a subclass of geo84:SpatialThing
> >> (but foaf:Person is), I have to create an union class with foaf:org +
> >> geo84:ST to use based_near with this org ?
> >>
> > No, "Descriptive, not prescriptive" means that you don't have to
> > pre-define rules before you can use certain RDF terms. The orgs
> > you'd like to use based_near with are spatial things.
> >
> > From
> >
> >    :org foaf:based_near :x .
> >
> > you can infer that
> >
> >    :org a geo:SpatialThing .
> >    :x a geo:SpatialThing .
> >
> > The description
> >
> >    foaf:based_near rdfs:domain geo:SpatialThing .
> >
> > does not say that only resoures explicitly typed as spatial things are
> > "allowed" to use foaf:based_near. It's exactly the other way round:
> > Resources which use foaf:based_near *are* spatial things. Independent
> > of other types they may have. (In OWL you can construct axioms to
> > identify/prevent inconsistencies, but not in RDF Schema. The latter
> > can only increase the total number of triples, but never reduces them.)
> >
> > Bottom line: You simply don't use foaf:based_near with resources
> > [in the subset of foaf:Organization] that aren't spatial things.
> >
> >
> > cheers,
> > Benjamin
> >
> --
>
> *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 Monday, 29 January 2007 15:44:42 UTC