W3C home > Mailing lists > Public > www-rdf-logic@w3.org > March 2006

Re: Mixed Object and Object Properties Hierarchies

From: Anne van Rossum <anne4theweb@yahoo.com>
Date: Mon, 27 Mar 2006 14:15:03 +0200
Message-Id: <6.1.2.0.0.20060327134838.025e7958@127.0.0.1>
To: www-rdf-logic@w3.org
Dear Pierre-Antoine,

I want to thank you personally for your reaction. Maybe I have to post my 
question somewhere else, because it's about the theory behind OWL (or 
DAML+OIL). Please correct me in that case.

Rephrasing: The question is: "Why does OWL - with many other ontologies - 
preserve such a strict separation between the object and relation 
hierarchies?" It appears that Sowa does the same in his conceptual graphs 
and also in Cyc the two are kept distinct.

Context: I'm building a framework for linguists (in a rather specific 
modality). These linguists should be able to use semantic relationships of 
their own, so this prevents me from using predefined ISA hierarchies etcetera.

Examples: You are right that a relation exists between two or more objects. 
However, a verb can also be seen as a relation between several objects (in 
for example Case Grammar). The number 4 can be seen as a concept or as a 
procedure of iteration. A plane as a concept or as a product of (relation 
between) two vectors. It is rather general to handle a relation like a 
normal concept, but that is what I want to do.
I do not know the objections that underly the decission of OWL or DAML+OIL 
to keep the "property" and "object" hierarchies distinct. Which literature 
about OWL or its predecessors handles this material?

Thanks a lot in advance!

Andy

At 15:41 24-3-2006, Pierre-Antoine Champin wrote:

>I know OWL better than DAML+OIL, but I trust what I will write is true
>in the latter.
>
>Any object property a corresponding concepts, called restriction in OWL,
>which can be expressed as
>
><owl:Restriction rdf:ID="Child">
>   <owl:onProperty     rdf:resource="#hasParent" />
>   <owl:someValuesFrom rdf:resource="#Animal" />
></owl:Restriction>
>
>In english: Child is the set of objects having at least one parent who
>is an Animal.
> >From the domain of hasParent, it follows that Child is a subclass of
>Animal (only Animals can have parents).
>
>Provided that you can define the inverse property of hasParent (which
>you can in OWL, I think you can also in DAML+OIL), you can as easily
>re-construct the concept of Parent as follow
>
><owl:Restriction rdf:ID="Parent">
>   <owl:onProperty>
>     <owl:ObjectProperty>
>       <owl:inverseOf rdf:resource="#hasParent" />
>     </owl:ObjectProperty>
>   </owl:onProperty>
>   <owl:someValuesFrom rdf:resource="#Animal" />
></owl:Restriction>
>
>It is hence possible to define two concepts for every object property.
>Note that the other way is not possible. You can not, in OWL, enforce
>every givesObject arc pointing to a Present to be also a givesPresent arc.
>
>So when this is possible, you should IMHO rather define a property than
>a class.
>It is due, in my understanding, to the fact that those concepts are mere
>"roles" : they do not intrinsically characterize an object, but rather
>the relation of that object to another one. You are not only "a parent",
>but "the parent OF somebody else". A flower is intrinsically not "a
>present", but can become "a present FROM somebody" when given (and a
>present TO somebody, for that matter, but this is not captured by the
>givesPresent property...)
>
>   pa
>
>
>
>Anne van Rossum wrote:
> > Good afternoon,
> >
> > Please redirect me if I'm posting in the wrong list. At
> > http://www.daml.org/2001/03/daml+oil-ex the following text is given in
> > the example file:
> >
> >     <daml:ObjectProperty rdf:ID="hasParent">
> >
> >           <rdfs:domain rdf:resource="#Animal"/>
> >
> >           <rdfs:range rdf:resource="#Animal"/>
> >
> >         </daml:ObjectProperty>
> >
> > *Observation*: The concept "Parent" is not reflected in the ontology,
> > while the property "hasParent" does have something to do with the
> > concept "Parent". The property "hasParent" is not derived from something
> > like "hasProperty" and "Parent". This is because object properties and
> > objects are maintained as disjoint sets.
> >
> > *Example*: Suppose an ontology for a visual language. In that language
> > "givesObject" can be represented by a certain picture. The property
> > "givesPresent" should subsequently be derived from "givesObject" and
> > "Present".
> >
> > *Recapitulating*: In general there are many situations in which
> > properties form a hierarchy together with objects. How to represent
> > combinations of object and property hierarchies in DAML+OIL?
> >
> > Dear regards,
> >
> > Andy
> >
Received on Monday, 27 March 2006 12:15:56 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:54 GMT