RE: Antwort: RE: Semantic web article in Nature Biotechnology

- Phillip, 

> How does this work with properties. You suggest, for example, 
> that "complex ontologies should be avoided"--a complex 
> ontology being something that references other ontologies. 

I don't understand, what difference does a property make?  If, for instance,
we have a property "foo" and you intended to related a:D and b:R, which are
two classes, for the rdfs:domain and rdfs:range of the foo. To avoid complex
ontology, you create a new ontology "c", where you say, 

c:foo rdfs:domain c:D.
c:foo rdfs:range c:R.

And in a prfoile, p1, you say

c:D owl:equivalentClass a:D
c:R owl:equivalentClass b:R.

And then you use your p1 for application.

It may sound strange at first, but when you create c:foo, you make an
ontological commitment to certain reality.  People may like your "foo", but
not necessarily your "conceptualization" that your c:D is the same as a:D.
To make it this way, you give others a chance to use your c:foo without
subscribe your view to others.  If they like your "binding", they just use
your profile.  This extra work maximize ontology reuse, and the key of SW's
success is ontology reuse.

> If we take a biological example, OBO is currently coming up 
> with a core Relationship Ontology. Now they use a different 
> formalism from semantic web "proper", but, to me, having a 
> core set of vocabularies describing is a positive thing. 

Yes, core is good. I don't have objection to that.  But if your tie all
other ontologies to it, it kills science because what if the core turns out
to be faulty, then all other ontologies die with it as well.

> It also appears to me that if you consider the formal 
> semantics of OWL if any two ontologies use the same 
> relationships (from a third ontology) then you may well be 
> able to infer relationships between concepts in the two 
> ontologies that even if they do not directly reference each other. 

Yes, or not.  RDF is open world semantics, you are free to add your own
constraints on the terms developed elsewhere.  Using Profile actually help
you to manage this personal view.

> I think Robert also has a deeper point about modularisation. 
> We need to be able to split ontologies up. But we don't have 
> the tools and language support that most programming 
> languauges offer. Should an ontology not be able to declare a 
> "public" interface, as with many programming languages?

Yes, I can't agree more on content partition.  But it is such a subjective
matter and cann't be applied without a context.  The O3 is introduced to
help guiding the ontology deployment not development. They are two different
issues, don't you agree?

Xiaoshu

Received on Thursday, 13 October 2005 13:14:17 UTC