Re: Building terminological consensus, part 1: Foundations

Pat Hayes wrote:
>
[snip]

>
> OK, back to the beginning.  Following the informal precedent of this
> discussion so far, I'll use prefixes to mark technical terms with
> their owners, so, e.g. pl:denote for 'denote' per Philosophy of
> Language, webarch:identify for 'identify' per WebArch,
> la:depict for 'depict' per Goodman's _Languages of Art_ [3] and
> vsr:expression for 'expression' per Cantwell Smith's "Varieties of
> Self-Reference" [4].
>
> > Oh, that last one is rather a tricky source for such a bland word.
Again, the real question is should we use that word "expression" or
should we fall upon the intension/extension difference. I believe the
word "expression" is trying to basically say something slightly
different - that here's a material substrate (i.e. bits and bytes of a
web-page, or some RDF statements)  that e*express* someone's intension
(i.e. a picture of the Pluto or a OWL Class for Pluto)  - or in Smith's
words, "impression"). Whether or not the extension/denotation of that
impression actually ends up mapping to Pluto is of course something
completely different. Anyways, expression here seems to equal
webarch:representation!
>
>
> pl:referring_expressions, in particular pl:names, pl:denote
> pl:referents.  pl:proper_names approximate pl:rigid designators, which
> have exactly one pl:referent.  Some pl:referring_expressions, known as
> pl:indexicals, may have different referents depending on the context
> of use, for example 'I', 'over there', 'next Thursday'.  One response
> to this is to distinguish pl:meaning from pl:interpretation,
> describing pl:meaning as a mapping from pl:context to
> pl:interpretation, where it is pl:interpretations (of names
> respectively sentences) which are pl:referents or have truth values.
> Thus when two disputants each say "I'm right, you're wrong", their
> utterances have the same pl:meaning, but distinct pl:interpretations,
> of which at most one is true.
>
> How the relation between pl:proper_names and their pl:referents is
> established, transmitted and maintained is the subject of considerable
> debate, but one currently popular position associated with Kripke [5]
> distinguishes categorically between pl:proper_names and other
> pl:referring_expressions, appealing to a notion of original pl:baptism
> for the former alone.
>
> So far so good -- I think we can establish some uncontroversial
> parallels:
>
>   pl:proper_name == webarch:URI
>   pl:denote == webarch:identify
>   pl:referent == webarch:resource
>
> > I don't find ANY of these even plausible, let alone uncontroversial.
> A URI is like a proper name?!? In what possible sense? How can one
> make any sense at all of the notion of an ontology all of whose names
> are proper names?? And (sorry to be so persistent, but) if
> webarch:identify means pl:denote, then why is there anything at all in
> the webarch document about http? Almost all of that document is about
> topics which have nothing to do with pl:denotation; so what justifies
> this equation? The last equation is categorically wrong, since
> 'referent' isn't really a category: it refers to something at one end
> of a relationship, which can in fact be anything at all.
p1:proper_name == webarch: URI if you buy the line "URIs denote
*exactly* one thing", which drags in a bunch of controversial
assumptions, ala the world is neatly divided into individual things, and
those things remain static over time. The question is - it seems that
assumption is needed to make the SemWeb work, so how can it be
*engineered* in?

I think Pat's objection is that he's noticing that proper names refer
traditionally to individuals, not classes, and you need classes for an
ontology. I think Henry's thinking *things* in a more general sense that
includes both classes and individuals, as well as perhaps more fuzzy
"things". That's stretching the traditional definition of "proper names"
to more of a loose analogy than a strict defintion.

The p1:denote = p1:identify is perplexing - but it seems like on the Web
identify means something a bit different than denotation in the logical
sense - as Pat's earlier observation about URIs giving not only
reference/denotation, but *access* - and this enables the
"Self-Describing" Web. The real reason why it's identify seems to be
that, unless you buy a strict ontological separation between the thing
the URI denotes and the representations served by the URI, then to some
extent the URI gives you actual access to the referent, something *not*
implied by p1:denote.
>
> Both sides of this set of equations gloss over the
> pl:meaning/pl:interpretation distinction, although in slightly
> different ways.  "John Smith" is actually a pl:indexical, and needs a
> context to uniquely determine a pl:referent, whereas webarch:resources
> can themselves _be_ pl:indexical -- WebArch would say that _the_
> webarch:resource webarch:identified by http://www.guardian.co.uk/ is
> something like "the current front page of the Manchester Guardian".
>
> > As I understand the REST architectural model, the actual resource is
> the *function* from times to the MG front page at that time. Then you
> don't have to say its indexical, which probably makes the whole theory
> simpler.
So, the resource is a function from time to Webarch:representations?
Correct? I think Roy says that, but I think Tim and the SemWeb community
are often saying something a bit different, as in URIs that have no
representations can still identify resources.
>
> So wherea the pl:meaning of a pl:referring_expression is a function,
> and thus something quite different in kind from most pl:referents,
> webarch:resource does double duty, corresponding to both pl:meaning
> _and_ pl:interpretation in a more careful set of equations:
>
>   pl:proper_name == webarch:URI
>   pl:denote == webarch:identify
>   pl:meaning == webarch:resource
>   pl:interpretation == pl:referent == webarch:resource
[snip]

To recap my take on Henry's take, maybe one way to put it would be to
ditch the proper name analogy (while noting the concept of baptism is
important, although the implementation details are fuzzy), and just say:

  pl:name == webarch:URI
  (p1:names may webarch:identify one or more things, including classes
of things)
 
  webarch:identify = (p1:access or p1:denote)

  p1:thing == webarch:resource

  pl:refer == webarch:identify
  (This is broken since p1:refer can both p1:denote p1:things that has
no causal access to the Web, or have direct causal p1:access to a p1:thing.)
 
  pl:interpretation/denotation/referent == webarch:resource == p1:thing

So
(p1:name/webarch:URI) (webarch:identify/(p1:access or p1:denote))
(p1:thing/interpretation/denotation/referent/webarch:resource).

To draw out the parallel:

p1:name         p1:denote              
p1:thing/interpretation/denotation/referent
webarch:URI  webarch:identify   webarch:resource.

Then the key is that you can also:

webarch:URI p1:access p1:thing/interpretation/denotation/referent

webarch:URI p1:access webarch:resource

This last statement only makes sense if webarch:resource is composed as
a mapping of webarch:representations.

   Lastly, what Henry was pointing out I think is that the
denote/meaning/interpret function results in
denotations/referents/interpretations can depend on context, and so may
not be one individual. If the only context is time, then we can include
that by making the definition of resource time-varying, and the problem
goes away. Is there any more context?

  . Again, the two problems here are:

    1) Resources do not have to be individuals, but can be
webarch:representations or classes.
    2) URIs give one an access relationship in some cases to the
thing/referent/denotation/interpretation itself in some cases. In other
cases, it does not give you direct access, but indirectly gives you
"more information" about the dentotation/referent/interpretation that
lets you pin it down. And that, is *new* on the Web I think, and one of
the more important features of the Web that separates it from ordinary
language.
>  In both cases such authority may be
> exercised felicitously or spuriously.
> > Right now it can't be exercised at all. There aren't any suitable
> ceremonies to exercise it in.
>
> > Pat
>
>
> ht
>
> [1] http://lists.w3.org/Archives/Public/www-tag/2007Jun/0056.html
> [2] http://www.w3.org/TR/webarch/.
> [3] Goodman, N., _Languages of Art: An approach to a theory of
>     symbols_, second edition, Hackett Publishing, 1976.
> [4] Cantwell Smith, B., "Varieties of Self-Reference," in Joseph
>     Halpern, ed., _Theoretical Aspects of Reasoning about Knowledge_,
>     Morgan Kaufmann, 1986, pp. 19-43.
> [5] Kripke, S., _Naming and Necessity_, Cambridge, MA: Harvard University
>     Press, 1980.
> --
>  Henry S. Thompson, HCRC Language Technology Group, University of
> Edinburgh
>                      Half-time member of W3C Team
>     2 Buccleuch Place, Edinburgh EH8 9LW, SCOTLAND -- (44) 131 650-4440
>             Fax: (44) 131 650-4587, e-mail: ht@inf.ed.ac.uk
>                    URL: http://www.ltg.ed.ac.uk/~ht/
> [mail really from me _always_ has this .sig -- mail without it is
> forged spam]
>
> -- 
>         -harry
>
> Harry Halpin,  University of Edinburgh
> http://www.ibiblio.org/hhalpin 6B522426
>

Received on Tuesday, 10 July 2007 04:36:59 UTC