Re: Ambiguity (was: Comments on "SPARQL 1.1 Uniform HTTP Protocol for Managing RDF Graphs")

On Sun, 2011-03-20 at 22:10 +0000, Graham Klyne wrote:
> It seems to me that the goal here is not so much unambiguity of
> reference, but reliablility of prediction (or inference).

I think that's a reasonable way of looking at it.

> 
> My reading of model theory and interpretations, etc., is that by
> proving that some conclusion holds under all possible satisfying
> interpretations of an expression, then it must also hold true for some
> particular satisfying interpretation that one may have in mind all
> along.

Agreed.

> 
> So, while we as (non-logical) observers may usefully (and incorrectly)
> believe that some term refers unamibiguously to some thing, the logic
> may not support this, but ultimately it doesn't matter because the
> answers that the logic may yield are still correct.

Exactly . . . for *that* application.  So an application that only needs
to prove the color of an object can prove that in every satisfying
interpretation, the color of <http://example/vxb24#a37f> is "black". 

On the other hand, a different application using the same URI in a
different RDF graph, and attempting to prove that the object is a TV or
a shoe may not be able to do so if all the information it has about the
object is:

  <http://example/vxb24#a37f> :color "black" .

I.e., for the second application, that same URI is ambiguous.

> 
> (If we were able a priori to prescribe a particular domain of
> discourse, is it still not possible for logic to constrain terms
> unambiguously within that domain?)

Perhaps, but I suspect that the problem of ambiguity of reference would
merely have been shifted to the description of the domain of discourse,
leaving that as the "then a miracle occurs" step:
http://tinyurl.com/5dqf48 

> 
> Thus, I think it's a useful fiction to think of the terms as
> unambiguous (at least in the scope of some discourse), even when this
> cannot be proven or enforced.

I agree, in the sense that being unambiguous is a good *goal*.  But I
think it is also important to recognize that this goal is not
*universally* achievable -- it is only achievable within a certain class
of applications.  There will always be other applications that require
finer (or different) distinctions.  Thus, one should not think of
ambiguity as necessarily being due to sloppiness.  This is what I've
been calling myth #3:
http://dbooth.org/2010/ambiguity/paper.html#myth3

In other words, when you mint a URI and publish a URI declaration that
defines its identity, by deciding what assertions to include in your
definition you are making a choice about the class of applications that
your URI will support.  Your URI will be unambiguous within that class
of applications, but ambiguous to other applications that require finer
distinctions.


-- 
David Booth, Ph.D.
http://dbooth.org/

Opinions expressed herein are those of the author and do not necessarily
reflect those of his employer.

Received on Monday, 21 March 2011 15:51:00 UTC