Re: [OEP] "Classes as values" first draft

Natasha

I thought I would have a more complete response ready by now be events have intervened.
Just a few comments below on your comments.

Natasha Noy wrote:

> Alan,
>
> I think you are right that I wasn't clear in what I was trying to say.
> And I hope we can hammer out exactly the right words.
>
> >  My first question when presented with any KR question is "What are
> > you trying to say".
> >
> >  The  use case is
> >
> >  "We want to be able to say that an image of a Lion is also an image
> > of a Mammal"
> >
> >  My italics and boldface of the indefinite articles 'a'.
> >
> >  We are NOT saying that this is a picture of the "species Lion" or the
> > "class Lion."
>
> I think the confusion here comes from the ambiguity of the word
> 'subject'. I'll try to think of a different example. I used subject
> here (and, again, I think that's the intended interpretation of
> dc:subject) as the topic of the image rather than to describe something
> that is actually shown in the image. Kind of like "Ontologies" can be a
> subject of someone's PhD Thesis.

but the intention as I understand it also includes

      "...dc:subject value(Shakespeare)"

for a picture of Shakespeare.

In which case what is meant is literally that it is a picture of Shakespeare.  If there is another
property for this case, please correct me.




> I'll think of using a different
> example for the next draft to avoid the confusion (any suggestions are
> welcome!). In fact, think about books about animals and their subjects.
> Here a book about lions is really a book whose subject is a class Lion
> rather than an individual Lion.

I accept that the case of a "Book about Lions" is a ambiguous.

There is quite a bit to say here which I will illustrate in a later email.
However, we don't want people to have to understand all that before they use the Semantic Web.

I think we need two clear guidelines

    One for those who want to stay with OWL-DL,
    The other for those who want to use OWL-Full

With luck, if we can stick to only one guideline each for OWL-Full and OWL-DL, then we can suggest standard transforms to get from one to the other with a minimum, well defined, distortion of meaning.

I'll argue for my suggestion  for the OWL-DL case in a subsequent email supported with ontologies and examples.

>
>
> I think that with this interpretation, using class as the value for the
> property can indeed be considered a correct representation (my approach
> 1) and matches this interpretation pretty closely (with the benefit of
> being the most intuitive to most people and with the serious drawback
> for many of being outside OWL DL).

My point is partly that what I suggest works in OWL DL smoothly and achieves
exactly what you want - that any picture of a Lion is also a picture of Mammal, an Animal,
etc. I'll attach the ontologies when they are fully annotated and screen shot.

>
> I think the option that you propose actually corresponds to a different
> interpretation of subject, where the value for the subject property
> would be the specific lion shown in the picture. In fact, your
> statement
>
> >     LionImage rdf:type restriction(someValuesFrom dc:subject
> > someValuesFrom Lion))  (1)
> >
>
> says exactly that. It will also make any kb that doesn't have that Lion
> individual filling in the property value inconsistent with this
> definition.

NO.  It doesn't say anything about any specific Lion.  It only says that
in the specific KB with the picture, that picture is an instance of a
"picture of a Lion".

The advantage of this version over your case in which you instantiate individuals is precisely
that it leaves the individual indeterminate.  Any KB can point to the reference ontology and
indicate that some resource is an of type "image of a lion", e.g.

    zzz:myResource rdf:type (yyy:Image and restriction(xxx:subject someValuesFrom yyy:Lion))

where xxx is the namespace for the reference property, e.g. Dublin Core, and yyy is the name space for the reference ontology, and zzz is the namespace for the referring KB. (Obviously we could name the target class something like "Image_of_a_lion", but I have written it out in full here.)

Anyone with access to the KB at namespace yyy will be able to infer
that zzz:myResource is also or type  image of a Mammal, a Carnivore,
a Ferocious_animal, or whatever other superclasses are in the ontology at xxx.

The point is - we don't care which lion; just that it is 'a lion'.   (or 'some lions').

Furthermore, if they have another picture of 'a lion' or 'lions' they can reference it to the same point unambiguously.

I hope this is clearer.

Regards

Alan


>

--
Alan L Rector
Professor of Medical Informatics
Department of Computer Science
University of Manchester
Manchester M13 9PL, UK
TEL: +44-161-275-6188/6149/7183
FAX: +44-161-275-6236/6204
Room: 2.88a, Kilburn Building
email: rector@cs.man.ac.uk
web: www.cs.man.ac.uk/mig
        www.opengalen.org
        www.clinical-escience.org

Received on Friday, 23 April 2004 08:49:27 UTC