Re: rule question

On 19.07.2004 16:57:42, paolo veronelli wrote:
>
>1)
>
>I think I can use the following:
>
>facts:
>  x1,p,y1
>  x2,p,y2
>  x1,rdf:type,c1
>  x2,rdf:type,c2
>
>inference
>
>c1,rdf:subClassOf,c2
>	or
>c2,rdf:subClassOf,c1
>
>is it right?

let's say
   peter age "5".
   myHouse age "500".

   peter rdf:type Person
   myHouse rdf:type Building
   and (age rdfs:domain rdfs:Resource)
   or (age rdfs:domain owl:unionOf[Person, Building])*

(*err, no, this is _not_ another rdf serialization ;)

as you can see, neither "Person" nor "Building" is 
neccessarily a subclass of the other one.

>is it possible to express this in RDFS?
>
>2)
>is it possible to infere x,rdf:domain,y from a set of triples that don't 
>contain x,rdf:domain,* ?
if you assume a closed world you could maybe find "best fitting" 
domains for a given set of facts if you have certain additional knowledge
about the schemas involved. (I'm not sure but I think the earlier 
versions of the RDF/iCal spec were partly generated this way.) 
In my example above, you couldn't even infer a common upper (named)
type for x1 and x2 from the fact that they are using the same
property. (well, other than rdfs:Resource or other general classes).


hth,
benjamin

--
Benjamin Nowack

Kruppstr. 100
45145 Essen, Germany

http://www.appmosphere.com/

>
>how is it composed this set?
>
>					Thanx for any hint
>
>
>
>-- 
>.....lotta dura  per la verdura
>
>
>

Received on Monday, 19 July 2004 12:54:47 UTC