class 'discovery' was Re: revised version of semantics document

Jos De_Roo wrote:
>
> [...]
>
> > > the one that I can go with is
> > >
> > >   :John a :Student .
> > >   :John a :Employee .
> > >   :C owl:intersectionOf ( :Student :Employee ) .
> > > OWL-entails
> > >   :John a :C .
> > >
> > > as all models of the premis are also models of the conclusion
> > > and no new existentials are introduced in the entailment rules
> > > (the lists in the conclusions are identical clones)
> >
> > Do you have a complete way of transforming from the natural entailment
> to
> > this entailment?
>
> well Peter, I don't know what you mean with "natural entailment"
> I guess it's the one that you wanted to hold i.e.
>    :John a :Student .
>    :John a :Employee .
>  OWL-entails
>    :John owl:intersectionOf ( :Student :Employee ) .
>
> but I still don't understand the question (and it's 3:12 AM here)
...
>
> and I think it is quite natural to explicitly give
>   :C owl:intersectionOf ( :Student :Employee ) .
>
> as a premis, no?
>

An advantage of entailing a class that isn't a premise (i.e. isn't already
named) is that the classifier might 'find' classes that naturally exist but
haven't been explicitly labelled. This might be useful, for example, in the
case where a cluster of symptoms, findings etc. indicate a brand new
disease. The advantage of not requiring the disease to be already
known/named is that given a knowledge base, an OWL classifier might
otherwise be able to discover new facts. This could be a big benefit, I
imagine.

Jonathan

Received on Wednesday, 21 August 2002 21:21:32 UTC