[seek-kr-sms] property restrictions in growl

[Ferdinando Villa]

I have seen OWL tutorials advise to never fill in domain and range, but
specify everything through restrictions... I don't think it's too good
to mandate the full definition before putting in restrictions unless
we're really forced to.

On a more general note, I think there are two very different purposes to
develop ontologies, and the OWL way - open world, axiomatic, optimized
for reasoning - is very different from the RDFS way, which is much more
like defining the conceptual skeleton of an object-oriented class
system. I think OWL suffers for wanting to support RDFS idioms like
property and range, which extend its applications but in my view muddle
the overall model. And OWL-DL does not live up to RDFS anyway for
constraint-based modeling, having things like concrete/abstract roles,
properties linking to classes, etc.

In my apps, after much thinking I have decided to support both RDFS and
OWL and use each for what it does best - constraint-based schema
specification for the first, axiom specification and reasoning for the
second. The problem is that they're not really that interchangeable. I
don't know what we should do in SEEK also because it's not clear to me
how much schema definition will be done based on ontologies.  But if our
uses of OWL is going to be predominantly to build concept maps that we
can reason on, I suggest that we downplay the hardwired constructs like
property range and domain, and find ways to make the axiomatic view as
intuitive as possible.

ferdinando

On Mon, 2004-08-02 at 10:47, Serguei Krivov wrote:
> Hi All,
> 
> We have been working hard on first alpha release of growl. Yet we have
> got a couple of important issues which probably should be resolved
> beforehand. A long discussion between me and Rich have not brought us
> to a clear consensus, so the feedback from the group is wanted at
> least for the most controversial issues. Here is one:
> 
>  
> 
> Issue #1  Should we allow user to define property restrictions without
> defining  explicitly the repective property? Or should we demand that
> user alwase define a property with domain and range and only then get
> permision to define (any) restriction on this property.
> 
>  
> 
> We are not sure what is legal in owl, specifically in owl-dl, but we
> have different intuitions about it. On one hand when reading a
> property  restriction with no explicit property definition owlapi
> complains  that it is not an owl-dl construct. Also just to exclude
> the possibility of typos while defining restriction it would be
> advantageus to have combo boxes that strictly confine  the names for
> restriction to link them to the properties that have been already
> defined.
> 
>  
> 
> On the other hand, we know it that definition of domains and ranges is
> not realy requred for tableaux algorithms to reason. Moreower –
> semantic of  domains and ranges in owl-dl is defined   via property
> restrictions which are  the elementary constructs in DL. Therefore the
> complains in owlapi about non-dl construct  is weared. It may be
> related to the fact that  from owl syntax it does not follow if a
> restriction pertained to data property or object property. For example
> here is the code:
> 
>  
> 
> <owl:Class rdf:about="http://a.com/ontology#Object1">
> 
>   <rdfs:subClassOf>
> 
>   <owl:Restriction>
> 
>     <owl:onProperty rdf:resource="http://a.com/ontology#Relation1" />
> 
>         <owl:minCardinality
> rdf:datatype="http://www.w3.org/2001/XMLSchema#int"
> 
>        >2</owl:minCardinality>
> 
>   </owl:Restriction>
> 
>   </rdfs:subClassOf>
> 
> </owl:Class>
> 
>  
> 
> Is Relation1 an object property or a data property? From owl syntax it
> does not follow and may be this is a reason for the complain. This
> uncertainty as we see it here may be a problem of owl design , but it
> is not problem of growl, since growl has different node types for
> object property restrictions and data property restrictions. And if
> property restrictions with no explicit property definitions are
> allowed in owl-dl we have to support their editing in growl.
> 
>  
> 
> So what we have to do with so called  “non typed” property restriction
> –allow them in growl or banish them from growl? Any of your thoughts
> on this subject would be welcome.
> 
>  
> 
> Thanks,
> 
> Serguei
-- 
Ferdinando Villa, Ph.D., Associate Research Professor, Ecoinformatics
Gund Institute for Ecological Economics and Dept of Botany, Univ. of Vermont
http://ecoinformatics.uvm.edu