Re: (Round 2) Proposed Extensions to OWL

[Richard H. McCullough]

> This conceptual model is OK, but unnecessarily complicated,
> and produces RDF descriptions which are just too long.
>
> Since these physical quantities are just literal numbers,
> they can be expressed as attributes of attributes, e.g.
>     <ex:River ex:Yangtze>
>         <ex:length units:kilometer=6300/>
>     </ex:River>
>
 Well, that is just the point.  Physical quantities are __not__ just numbers
with units attributes.  They - or rather their measurements, which is really
what we are talking about here - may have sets of mesaurements, various
units some of which may be named and some of which may be unnamed (what is
the unit which is dimensionally kPa/m-s^1.5 ?), they have accuracy and
precision, they have provenance, the measurements may change over time
(after the Russians changed their standard thermometer for measure outdoors
temperatures in the 1920s (I think it was), all subsequent temperatures
turned out to be inconsistent by a few degrees with data taken using the old
style thermometers - you have to  know when a Russian weather reading was
taken to be able to make proper use of it).

Furthermore, conceptually there can even be different kinds of things called
lengths, etc.  One example is the area of a rectangle vs the surface area of
a rectangular pice of rough cloth the with the same width and length.  Thus
there are physical quantities, measurements of physical quantities, and the
values of those measurements.  The values may not even be simple numbers.
They may be vectors or matrices, for example.

Of course, it all depends on what you want to do.  Roger Costello asked how
to represent  an equality test to state that two lengths written in
different units were equivalent.  You cannot assess equality unless you know
the units conversion formula and the precision to be used in the test, and,
if the two numbers represente different field measurements rather than just
a conversion between two units, the precision and accuracy of the
measurements as well.

The model you said is too complicated is intended to allow one to make
statements about

1) The kind of physical quantity
2) A measurement of a physical quantity
3) The measurement value(s) of a particular measurement.

Thus you need a stripe for each of the three, plus whatever you need to
model the measurement values.

But no, it is not the simplest way you could record a single number that is
somehow associated with the subject.

Cheers,

Tom P

Received on Monday, 7 July 2003 09:35:28 UTC