Re: a few questions about literals

On Thu, 2002-02-28 at 08:49:12 PM, Dan Connolly wrote:
> On Wed, 2002-02-27 at 17:05, Pat Hayes wrote:
> [...]
> > At present, a graph entails its existential generalization which is
> > gotten by 'erasing' urirefs into bnodes, eg
> >
> > ex:Jenny ex:age what:ever .  |=    ex:Jenny ex:age _:x .
> >
> > Do we want this to be true for literals as well? Eg should this be a
> > valid inference?
> >
> > ex:Judy ex:age "10" .   |=      ex:Judy ex:age _:x .
>
> yes.
>
> It is in our "swap" software. I can make test cases if
> anybody likes.

as is in our stuff
i.e.
  ( <http://www.agfa.com/w3c/n3/p13.nt> )
    log:entails <http://www.agfa.com/w3c/n3/p14.nt> .

> > It seems to me that we should do this, since there is no doubt that
> > literals do denote something - themselves, in fact - in any
> > interpretation.
> >
> > This means that the following inference would be valid, for example:
> >
> > ex:Jenny ex:age "10" .   |=    ex:Jenny ex:age _:y .
> >
> > which might seem a bit worrying if there was an rdfs:drange assertion
> > around, eg
> > ex:age rdfs:drange xsd:number
> > which imposes the 'lexical' datatype in the first case, but looks
> > like it might impose the 'value' in the second case; but in fact it
> > is OK, since that conclusion would only trigger the value datatyping
> > constraint if the bnode were also the subject of rdfs:dlex; and that
> > in turn would require the original graph to have had something like
> > this in it:
> >
> > ex:Jenny ex:age "10"  .
> > "10" rdfs:dlex "12" .
> >
> > which is so crazy that no-one should be surprised if it has crazy
> > entailments, right?
>
> right.

indeed, assuming that rdfs:drange only imposes datatype checking on
literals

> > Anyway, if y'all agree that we should accept this inference, then I
> > think the simplest way to re-do the MT is to simply say up-front that
> > *all* RDF interpretations must include *all* literals in their
> > universe.
>
> yup; that seems straightforward.

I fully agree

> > Then we can just say that for literals E, I(E) = E, and not
> > talk about things like LV and XL at all. Does anyone have any
> > philosophical objections to this? It would allow quite a few of the
> > lemmas to be stated with fewer qualifications, and the proofs to be
> > simplified.

no objection, just praise

--
Jos

Received on Thursday, 28 February 2002 16:57:07 UTC