Re: SEM: comprehensive entailments without dark triples

Dan Connollywrote:

> On Mon, 2002-04-22 at 16:19, Dan Connolly wrote:
> > On Mon, 2002-04-22 at 15:56, Jonathan Borden wrote:
> > [...]
> > > "Patients with a dominantly inherited disease  have a father with a
> > > dominantly inherited disease, and/or a mother with a dominantly
inherited
> > > disease"
> > >
> > > <Class rdf:ID="DominantInheritance">
> > >     <unionOf>
> > >         <Restriction>
> > >                 <onProperty rdf:resource="#mother"/>
> > >                 <toClass rdf:resource="#DominantInheritance" />
> > >         </Restriction>
> > >         <Restriction>
> > >                 <onProperty rdf:resource="#father"/>
> > >                 <toClass rdf:resource="#DominantInheritance" />
> > >         </Restriction>
> > >     </unionOf>
> > > </Class>]
> > >
> > > ...
> >
> > OK; that looks like a stumper: it's a defined class,
> > and it's circular. I doubt Jeremy's rules
> > provide for the relevant conclusions.
>
>
> I take that back.
>
> I tried to make a test case out of this, but it doesn't
> involve any conclusions about the existence of circular stuff;
> the existence of the circular stuff is in the premise
> of the argument.

Perhaps the premises would be something like:

<Person rdf:ID="John">
    <ex:hasDiagnosis rdf:resource="#DominantInheritedDisease"/>
</Person>
<Person rdf:ID="David">
    <ex:father rdf:resource="#John">
</Person>

I would very much like the classifier to conclude that "David" has a
"DominantInheritedDisease" -- actually I assume that there will be numerous
subclasses ("dominant inheritance" is a first year of medical school sort of
concept). The classifier should conclude that when a father has any of these
subclasses as a diagnosis, the children will have the diagnosis.

Note: that I've simplified this concept, e.g. the penentrence of an
inherited disease is not often 100%. The intention is to say: "An individual
has something if either of his or her parents have this thing"

>
> So I don't see a problem.

I don't know if there is a problem. I was simply providing a real world
example of a 'circular' class.

It was asserted (sic) that circular classes caused problems (or 'the
problem'). I am simply pointing out that there is an actual real world need
to deal with circular classes/concepts. Does this example capture the
problem? If not, we should restate the problem.

Jonathan

Received on Wednesday, 24 April 2002 14:45:46 UTC