URIs, ambiguity and RDF

At 07:28 PM 8/5/02 +0100, Bill de hÓra wrote:
> > Jonathan Borden
> >
> > Miles Sabin wrote:
> >
> > >   URIs are context dependent identifiers.
> >
> > No. URIs are most certainly not context dependent identifiers.
>
>Then there exist no non-reserved vocabularies in RDF.
>
>More specifically, the interpretative function IS in the RDF Model
>Theory is redundant in all cases, if URIs are neither context dependent
>nor ambiguous referents. There's no need for a Tarskian semantics to
>determine the truth or falsity of a statement with respect to a given
>denotation, when there is only ever one denotation to be had. That's a
>remarkable property.

Thinking about:

* a URI always denotes a single concept

* the meaning of a URI is ambiguous

Are these contradictory?  I don't think so.  At least, not as far as RDF is 
concerned.

I agree with Bill insofar that different people will use a URIs with 
different intent.  People also contradict themselves.

An RDF graph describes a set of possible worlds, each characterized by an 
interpretation that assigns denotations to the URIs used (the IS in RDF 
model theory [1]).  By asserting the graph, we say that only 
interpretations for which the graph is true correspond to the real world 
(or some world that circumstance indicates we are describing).  When an RDF 
graph contains contradictions, we may end up with graph which cannot be 
satisfied by any interpretation -- there is no possible world that it 
describes.  So, in general, taking the web as a whole, contradictions and 
all, it probably doesn't describe any world.

To make meaning of all this information, using RDF, we need to be 
selective.  How we select is not for RDF, or web architecture, to say.  But 
such selection will yield descriptions that are satisfied only by different 
interpretations of the URIs used.  I think there is much useful work to be 
done here, but that such work need not and should not muddy the lower 
layers of [semantic] web architecture.

And finally, I note that in his work on TAP [2], Guha is advocating a style 
of "reference by description", a key to which I understand is using a 
relatively small, well-defined vocabulary of URIs (thousands rather than 
millions of terms) to describe a far larger number of concepts using 
RDF.  I guess that the relatively restricted number of vocabulary URIs used 
will make it easier to establish sufficient levels of agreement about what 
concepts they actually denote.

#g
--

[1] http://www.w3.org/TR/rdf-mt/

[2] http://tap.stanford.edu/
     http://tap.stanford.edu/protocol/
     http://tap.stanford.edu/protocol/rbd.html


-------------------
Graham Klyne
<GK@NineByNine.org>

Received on Tuesday, 6 August 2002 13:31:26 UTC