Re: Proposed extensions to OWL?

[Roger L. Costello]
> Jos De_Roo wrote:
> >
> > we tend to use RDF interpretation properties for that
> > and math: properties and N3 conversion rules such as eg
> >
> > {?X eg:length-in ?Y} => {?X eg:length-cm (?Y 2.54).math:product}.
> > {?X eg:length-cm ?Y} => {?X eg:length-in (?Y 2.54).math:quotient}.
>
> [snip]
>
> Hi Jos,
>
> Does this have an XML syntax?  My interest is in how to express the
> mathematical relationships using an XML syntax.  /Roger

Applying units involves modeling them, and that involves more than one
aspect.  I think that it is pointless to declare equivalence between
instances of variables.  Something more general is needed. Also, a simple
scale factor is one thing, but there are often more complex relatoinships
involved.

A unit is like data type, but not a lexical one.  As long as we restrict
ourselves to scalar quantities that would be expressed in floating point
numbers of some flavor, it seems to me that we could derive new types from
the XML Schema data types.  Thus we could derive a type "Meter" from type
"Decimal", and so on.  The lexical representation would be unchanged.

Since these new types would be derived from XML Schema, presumably they
would be suited for use in RDF.

Next, I would model each conversion as a separate resource -

{&units;inch-to-cm-conversion
   {&units;baseUnit, &units;Inch}
   {&units;targetUnit, &units;Cm}
   {&units;scaleFactor, xs:decimal:2.54}
}

This is a simple, clean approach that says exactly what is intended, and
could be applied by a processor that understands the constructions.  No
instances of actual physical quantities have to be annotated with equivalent
conversion factors or equivalent values in several units.  The only
requirement would be the ability to derive datatypes like Inch, and this can
be done with XML Schema datatypes.

Obviously an OWL processor would not know what to do with these statements
and would not be able to reason with them to see if a particular conversion
were correct or not.  But that is not OWL's job, anyway, IMHO.  If Roger
wants to extend an OWL processor to understand scalar conversions, I suppose
he can do so.  But any attempt to logically test equality between different
numerical measures is going to be very difficult, because of the lack of
infinite precision, and different computing errors and precisions.  A
non-trivial undertaking, I think, and one that is not in the domain of
ontology.

Cheers,

Tom P

Received on Sunday, 15 June 2003 11:10:44 UTC