Wednesday, January 25, 2006
Generative Methodology Glass Bead Games
Philip said:
The OWL semantics for having multiple things at the end of range (or
domain), is that they are implicitly intersected. So if you have disjoint
classes at either end of a property, then you are requiring the property to
link no individuals. If you do this with domain you're likely to get a entirely
inconsistent ontology, since domain forces reclassification, while range
doesn't.
If you want to say "either DNA or Protein", then you need to
stick a union class expression in instead.
I *think* that protégé encodes this semantics straightforwardly; that
is two classes in the domain box will be serialized as being intersected, but I'm not sure.
Comment
One needs to always be very clear about what is being said. You are speaking from the OWL technical point of view and not about (say) category theory as might be understood regarding naturally occurring categories (what ever that might mean) as perceived by a biologist. Right?
It is the differences and similarities between biology and OWL definitions and "axioms" that have to be made clear.
When you say
"The semantics for having multiple things
at the end of range (or domain), is that they are implicitly intersected."
this is a "what" statement, whereas I am really interested in "why" they are implicitly intersected (as part of the development of Protege) and "how" this implicit intersection occurs in OWL.
Is there a "how" in terms of natural category formation (metabolic pathway anticipatory mechanisms) and how well is this "how" expressible in OWL?
I suspect that a "category" (and I use this in a non-technical common sense fashion) is implied when a property, say "property1", is defined in OWL. The category that "is" this property is then modified as the domain and range (of the property) are defined or changed.
In fact if a concept has a property2, and that concept is also in the domain, or range, of property1; then in some way the "semantics" of the property1 is changed when the definition of property2 is defined (initially) or is changed in some way.
So
if
< concept, property1, {range1} > is defined and later on
< concept, property2, {range2} >
is defined; then semantic changes propagate to the concept (even if no transitive relationships are defined).
(These appear to be the hard issues related to what is called "formal semantics").
The issue of interpretant adds an additional set of issues, which are sometimes called implicit semantics... under the assumption that this interpretation is fixed but simply not yet made explicit.