Re: Proposed extensions to OWL?

[Roger L. Costello]
> I believe that it will be useful to delve into the meaning of
> "equivalence".
>
> Would someone please explain what this means:
>
>    <owl:DatatypeProperty rdf:ID="length">
>       <owl:equivalentProperty rdf:resource="#distance"/>
>    </owl:DatatypeProperty>
>

equivalentProperty means that one property is a sort of alias for the other,
so that they are interchangable.  Of course, this is done in a typical OWL
fashion, by looking at the collections of individuals involved in the
relationships.  If A is related by property P1 to {B, C, D}, and property P2
is known to be an equivalentProperty of A, then you are entitled to infer
that A is also related through P2 to the exact same set of individuals {B,
C, D}.

In addition, P2 is a subProperty of P1 and vice-versa.

Well, that at any rate is what I get from the OWL documents, but I am not
too clear on how it relates to properties that are literal values.  May we
translate "the same set of individuals" to "the same set of values"?  It
seems reasonable but I did not see I written down.  Maybe it is in the
formal semantics and I just could not recognize it.

I think that it would be easy to understand if we had properties like this:

isOneInchAwayFrom
is2Point54CmAwayFrom

Here if we declare the two properties to be equivalentProperties, it is
obvious that
{Point1 is2Point54CmAwayFrom Point2} is implied if we know that {Point1
isOneInchAwayFrom Point2}.

What can we say about isOneInchAwayFrom so that it will be known to have a
value of one inch?  Obviously we could write some RDF statements, and since
a property is a resource (that is still so, isn't it?) we can say that is
value is "1" or "1 inch" or whatever.  We still need the conceptual model
and the vocabulary to assign units to the measure, so the problem looks like
the same one where we had a general property like Length.

If we wanted to, we could derive a datatype by restriction that had only the
single value we want, and make it the range of our property.  That would
seem to be formally correct but rather clumsy if we have a lot of property
values to assign.

Returning to Length, it seems me that we do not want to use a literal for
the value of Length at all.  Suppose we create a resource OneInchLength,
representing the concept of a length on one inch.  If we like, we could say
it is an instance of LengthInInches.  Notice that this is a noun, not a
verb, and we are talking about a subject or object, not a predicate. Now we
can say in RDF

     {Stick1, hasLength, OneInchLength}

If we also define a resource TwoPoint54Cm, we can assert that it is the
SameIndividualAs OneInchLength.  Now we have our unit equivalence under
control.  It is not a matter of equivalent properties but of equivalent
value resources.

Again, this would seem to be conceptually correct but rather clumsy if we
have a lot of objects like this.

I am going to stop here.

Cheers,

Tom P

Received on Tuesday, 17 June 2003 22:19:45 UTC