xml-dev - Re: Syntax and semantics

[Eric Bohlman <ebohlman@netcom.com>]

On Thu, 18 May 2000, John Robert Gardner wrote:
> I shudder to reference a rose and its aroma by any other name . . . but
> the point in naming is not semantic content but distinguishing.  To name,
> even with a nonsense word like buwidgerist, is to make a crucial
> ontological distinction upon which all semantic content is necessarily
> predicated: it makes the thing named _not_ something else. Whatever else
> it is beyond that--semantically or otherwise--is dependent upon the
> ensuing exchange.
> 
> It is a misnomer, then, to procede under the premise that a name alone
> serves no semantic purpose/meaning value.  On the contrary, it is the
> fundamental premise upon which any expository utterance, digital or
> otherwise, is dependent.  In and of themselves, the binary "1" or "0" at
> the electronic--even subatomic--level (and chaos theory attests to the
> import of naming as saying _something_, cf. Heisenberg Principle), say
> only what the state of that bit is not.  

This reminded me of an observation Michael Crichton made in his
(non-fiction) _Five Patients_ thirty years ago: that while it may seem
strange to hear a physician refer to a particular pathology as
"idiopathic" or "cryptogenic" and then speak of it as if it were a
definite entity, this usage is entirely legitimate, because the terms
don't just imply that we don't know the reason for the pathology; they
also imply that we've explicitly ruled out certain reasons.  If we say
that a patient has "idiopathic cirrhosis of the liver," we've specifically
stated that the patient's liver problems are *not* due to alcohol
toxicity, among other things.  If we say somebody died of "electrical
heart failure," we've said that he did *not* die of blocked coronary
arteries, malformed valves, autoimmune destruction of the heart muscle,
etc.  In these cases, the name of the entity doesn't permit discovery of
what the entity *is*, only what it *isn't*.  But that's still useful.

All human uses of language are based on using symbols to reference shared
context, with the mapping between the symbols and the elements of the
context having been negotiated (constructed) in advance, though not
necessarily by the two parties communicating (they don't have to have
participated in the negotiations themselves, just to be aware, however
dimly, of the result).  I am not aware of *any* form of human-to-human
communication in which the message (string of symbols) itself encodes not
only the references, but also the entire context and the mapping.  That is
to say, I am not aware of any form of human communication that is
meaningful without respect to _a priori_ shared context.  Language is
constructed, not received.  Some of the visions of the Semantic Web I've
seen strike me as little more than a desire for robots to "communicate" on
behalf of their masters without this necessity for shared context and when
I see "autonomous and anonymous" I find myself muttering "and autistic"
under my breath.


***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************