Re: log:forSome/#rdfms-identity-anon-resources

>pat hayes wrote:
> >
> > >Graham Klyne wrote:
> > >[...]
> > > > (1) Unknown binding...
> > >[...]
> > > > Following this line, any pair of names can bind to the same 
>object in the
> > > > domain of interpretation that matches what we know about 
>them, so from your
> > > > examples:
> > > >
> > > > [[[
> > > > <http://skolem.example#432oj34oij2o3ijo23j>
> > > > <http://purl.org/dc/elements/1.1/title> "Fractals Everywhere" .
> > > >   :
> > > > ]]]
> > > >
> > > > and
> > > >
> > > > [[[
> > > > <http://booksRus.example/inv2001-06-25#item342323>
> > > > <http://purl.org/dc/elements/1.1/title> "Fractals Everywhere" .
> > > >   :
> > > > ]]]
> > > >
> > > > Can be matched by an interpretation in which
> > > > <http://skolem.example#432oj34oij2o3ijo23j> and
> > > > <http://booksRus.example/inv2001-06-25#item342323> indicate 
>the same value.
> > >
> > >But that's the case only for *some* interpretations;
> > >using the formula in existentially-quantified
> > >form, before skolemiziation, *all* interpretations
> > >that satisfy the bookseller's database satisfy
> > >the query; i.e. going from one to the other
> > >is a valid inference.
> > >
> > > > The problem here seems to be one of computation rather than logic:  if
> > > > *any* pair of constants must be tested to see if they match the same
> > > > conditions, then the search space becomes impossibly large 
>for a practical
> > > > problem.
> > >
> > >No, it's a problem of logic: going from
> > >       (title item34 "f a")
> > >to
> > >       (title sk1 "f a")
> > >is not a valid inference; but going from
> > >
> > >       (title item34 "f a")
> > >to
> > >       (exists (?x0) (title ?x0 "f a"))
> > >is.
> >
> > True, but Graham has a point.
>
>Please let's not go into a tangent about tractable reasoning.

OK, just allowing the point. It is of considerable practical importance.

>My point is that this very common usage of RDF, describing
>something you want, is not possible if RDF parsers skolemize
>away all the existentials; i.e. if the abstract syntax
>of RDF doens't have existentially quanitified terms in it.

I disagree. I think that this common usage, if it really means what 
you are saying it means, isn't in fact an existential at all. Im not 
yet quite sure what it is.

>When you say "True", it seems that you are agreeing
>with the my position on this issue,
>namely, from my message of Mon, 25 Jun 2001 02:44:47 -0500 :
>
>| So an RDF document with anonymous nodes is *not*
>| interchangeable with a skolemized version of that document.

No, I was agreeing with what you said above (that

>going from
> > >       (title item34 "f a")
> > >to
> > >       (title sk1 "f a")
> > >is not a valid inference; but going from
> > >
> > >       (title item34 "f a")
> > >to
> > >       (exists (?x0) (title ?x0 "f a"))
> >is.

>It also seems that you agree with me on this issue
>when, in your message
>of Thu, 28 Jun 2001 10:56:44 -0500, you write
>| >In the current RDF specification (that we are chartered to clarify),
>| >is this a purely syntactic issue, or does the "model" also allow for
>| >anonymous resources?
>| If "model" means the set-of-triples abstract graph syntax, the answer
>| is surely yes, in the m&s.
>
>At least a few people in the WG are of the opinion
>(when last I heard from them) that the parsers can
>just make up constant names/URIs
>for all the anonymous nodes in an RDF document without
>substantially reducing the utility of RDF; i.e. that
>we don't need the _:xyz construct in n-triples at all.

If all the reasoning using RDF is strictly valid, I think that is 
true. It may be slightly more 'dangerous' to use skolem names, in 
that it may encourage an invalid (but not unreasonable) usage, but if 
that is the case then we ought to get that out into the open and 
maybe change the semantics to make it valid.

> > If the logic has equality then it is
> > logically easy to express what you want, ie (= item34 ski),
>
>No, not in this use case:
>	* I, book buyer, can't state that, because I
>	do not know that the book that I want is
>	known, to the bookseller, as item34, when
>	I make my query.
>
>	* the book seller can't validly infer that
>	if the query comes to him/her in skolemized form.

I didnt say it would be easy to infer, only that it would be 
expressible. I agree that the bookseller couldn't validly infer that 
without some extra knowledge about uniqueness. The constant being a 
skolem constant or not isnt really important, other than if it really 
is a skolem constant that you generated then the bookseller couldnt 
possibly know anything about it other than what you tell him. Skolem 
constants are indistinguishable from nonskolem constants as far as 
the logic is concerned.

> > but there
> > is a computational cost to allowing equality in the logic which is
> > exactly what Graham says: the reasoner has be constantly checking to
> > make sure that one name isnt equal to every other name, and it gets
> > expensive. Classical resolution equality reasoners generate subgoals
> > of the form (not (= a b)) like kudzu, which is what led to the desire
> > for general principles for rapidly deciding that a wasn't = to b, eg
> > the unique-names assumption.
>
>I don't care how much computational horsepower the bookseller
>brings to bear on this problem; their inference engine
>had better not deduce from
>
>       (title item34 "f a")
>and
>       (title sk1 "f a")
>that
>	(= item34 ski)
>
>because it's not a valid inference.

Right, but hes in the same position if you tell him an existential. 
He can't infer (foo item34) from (exists (?x) (foo ?x)) either.

>
>I'm saying that RDF must be able to express
>
>	(exists (?x) (title ?x "f a"))
>so that booksellers can deduce from
>	(title item34 "f a")
>that they have a book that meets the description of the buyer.

BUt if that description is just a bare existential, then they still 
can't infer that the one they have is the one you want. There might 
be several of them satisfying the existential.

If you want your order to mean "I want ANY book matching this 
description" then its not an existential: its a universal. I think 
this is what Jos DeRoos is doing with anonymous nodes as well, by the 
way. I have no beef with this, as long as we are clear what we want 
the anaymous nodes to mean.

Pat


>
>--
>Dan Connolly, W3C http://www.w3.org/People/Connolly/

---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Thursday, 28 June 2001 19:22:30 UTC