Re: log:forAll makes sense? [was: Can we agree on triples ?]

From: Dan Connolly <connolly@w3.org>
Subject: log:forAll makes sense? [was: Can we agree on triples ?]
Date: Mon, 02 Apr 2001 15:52:36 -0500

> "Peter F. Patel-Schneider" wrote:
> [...]
> >  For
> > example, if you represent quantification using "http://www.bar.com/logic#forall",
> > you could end up with triples of the form
> > 
> >         {http://www.bar.com/logic#forall,a,b}
> > 
> > which should not result in the assertion that there is some forall
> > relationship between the representation of the variable a and the
> > representation of the formula b,
> 
> I wonder about that...
> 
> > at least not one that will commingle with
> > assertions that come from triples of the form
> > 
> >         {loves,john,mary}
> 
> Er... why not? That's pretty much the design TimBL has coded up[swap].
> I've been trying to convince myself alternatively that is
> or is not a sound design. Maybe you can settle this for me/us...

The problem that I see is that there is no way for RDF agents (by which I
mean agents that understand RDF(S) only) to distinguish between

    {http://www.bar.com/logic#forall,a,b}
    {loves,john,mary}

They are both triples to an RDF agent, and must both be given
representational import.

> Looking at it in KIF, the latter is:
> 
> 	(loves john mary)
> 
> and the former is, say,
> 
> 	(log:forall '?a '(loves ?a mary))
> 
> where log:forall is defined so that this expands to
> 
> 	(wtr '(forall (?a) (loves ?a mary)))

Yes, but KIF has two big advantages here.  First, it is possible to have
axiomatizations of predicates in KIF, so you can, within KIF, tell a KIF
agent something like

      (log:forall '?a '(loves ?a mary))
      means
      (wtr '(forall (?a) (loves ?a mary)))

Second, KIF has a reasonably well thought out mechanism for expressing the
truth of a quoted sentence, and distinguishing between asserted sentences
and quoted ones, again from within KIF.

So a KIF agent can use the facilities within KIF to ``bootstrap'' itself
from a basic KIF agent to one that understands this different syntax.  Such
an option is no available within RDF.


Note that I am not arguing here that KIF is better than RDF for
representing information that can be represented in RDF.  I am also not
saying that I am completely happy with the KIF solution.


Peter F. Patel-Schneider

Received on Tuesday, 3 April 2001 06:50:25 UTC