Re: (Round 2) Proposed Extensions to OWL

[Roger L. Costello

> Question: are all models of the length of the Yangtze River related by a
> conversion factor?
>

Maybe the lengths are - modulo the issue of different definitions of what
"length" means (how fine a granularity do you use in following every little
twist of the coastline?).

But there are plenty of transformations that are not one to one.  Think of
switching between an angle (like 45 degrees) to a complex representation
(like e^i*pi/4) (which is still one to one in a way) to (1 + i)sqrt(2)/2,
which has two pieces, real and imaginary.  Or non-isotropic properties
expressed as tensors.  Or areas where the two sides are expressed in
different units ("1 mile long by 1 inch wide",  or acre-feet of water).

That is why I prefer to use an abstract Tranformation thing that could be
specialized for one to one length comparisons, for example.  Better not to
have to remodel when we hit other cases, within reason of course.

> Question: what is the role of OWL with respect to these models?  Should
> an OWL document be responsible for stating the conversion factor among
> models?
>
> Jon Hanna made an interesting suggestion last week:
>
> "... (Separating the conversion factors from OWL) seems analogous to the
> way several successful internet technologies work, e.g. HTML has a way
> of saying 'image goes here' but relies on other technologies to actually
> encode, transmit and render the image. So maybe OWL could benefit from
> being able to say 'conversion goes here'".
>

Just so.

> Perhaps the role of an OWL document in describing the relationships
> between the length models may be something as simple as this:
>
> <owl:Class rdf:ID="LengthInKilometers">
>     <owlx:convertibleModels>
>         <owl:Class>
>             <owl:unionOf rdf:parseType="Collection">
>                 <owl:Class rdf:about="#LengthInMiles">
>                 <owl:Class rdf:about="#LengthInInches">
>                 <owl:Class rdf:about="#LengthInCentimeters">
>             </owl:unionOf>
>         </owl:Class>
>     </owlx:convertibleModels>
> </owl:Class>
>
> That is, an OWL document simply identifies which length models are
> convertible. An application must "out-of-band" locate the conversion
> routines.
>
> What do you think about this?

I think that OWL could declare that two transformations have equivalent
results, as I tried to sketch out in my post from earlier today.  I say
"equivalent" to allow for tolerance (and perhaps measurement errors too -
they really ought to be provided for in some way, because they are often
critically important).

Cheers,

Tom P

Received on Wednesday, 2 July 2003 21:46:51 UTC