Re: question about DL/Lite semantics + agenda request

This includes my review of the changes to the annotation semantics.
I conclude that some WG telecon time should be spent on this and the "layering 
bug?" thread.

Jeremy:
> Does
> 
> Class( <foo> annotation( <a> <bar> ) )
> AnnotationProperty( <a> )
> Class( <bar> )
> 
> entail
> 
> AnnotationProperty( <a> )
> Class( <foo> annotation( <a> individual() ) )

Peter:
>I propose to fix this by modifying the direct semantics to admit ``junk''
>elements of the domain, i.e., elements of the domain that are not
>individuals.  This makes the direct semantics more ugly, but more like the
>RDF semantics.

>One side effect would be that Jeremy's entailment above would not hold,
>because <bar> would not have to be an individual.

Minutes June 5
http://lists.w3.org/Archives/Public/www-webont-wg/2003Jun/0050.
>ACTION: Jeremy Carroll - review proposed change to direct semantics that
>        handles annotations better.

S&AS editors draft section 3
> EC(owl:Thing) = O ⊆ R, disjoint from LV 
appears as the definition of O, but the first use of O is higher in the table. 
I found this somewhat confusing.
However, as far as I can tell, the changes are doing what you wanted ...

but ...

while

> Class( <foo> annotation( <a> <bar> ) )
> AnnotationProperty( <a> )
> Class( <bar> )
> 
does not now
> entail
> 
> AnnotationProperty( <a> )
> Class( <foo> annotation( <a> individual() ) )

we have that

T(
> Class( <foo> annotation( <a> <bar> ) )
> AnnotationProperty( <a> )
> Class( <bar> )
)
OWL Full entails
T(
> AnnotationProperty( <a> )
> Class( <foo> annotation( <a> individual() ) )
)

because the individual() item is transformed into a bnode (with type 
owl:Thing?) and the type triple is trivially true, so that the OWL Full 
contentful part is:

<foo> <a> <bar> .
entails
<foo> <a> _:b .

which is true in RDF.

This seems to be another instance of the "layering bug?" to which you say:

> if an entailment holds in OWL DL then it holds in OWL Full,
> not vice versa.

IMO it is worth highlighting this to the WG, and ensuring that the other 
documents capture this.

for example, TEST needs to change to make it clear that an OWL Full 
consistency checker and an OWL DL consistency checker may get different 
results on a DL document.

That's not good.

Jeremy

Received on Friday, 20 June 2003 03:41:36 UTC