Re: The Syntactic Web

On Tue, 10 Sep 2002, Jim Hendler wrote:

> Stevan- don't have time for a long response, and rarely respond to 
> such things - but since we've known each other a long time - let me 
> give you a quick response -- I personally believe that the single 
> most interesting thing about the web is that symbol grounding (in web 
> space) is doable, and that it is what the semantic web is based on in 
> almost every respect (although I think many in the SW community don't 
> realize it yet).  You might want to take some time to wander through 
> Tim Berners-Lee's writings about web design [1] - and particularly 
> his ideas about why URIs are so important and how they can be used to 
> anchor meaning (start with [2] and follow links).   I think you'll 
> see stuff relating to a lot of your interests -- the mapping to your 
> definitions of symbol grounding are not immediate and 1-1, but I 
> think you'll see a lot of interesting work there.
>
> I wish I had more time to discuss this with you in detail, but I'm 
> busy trying to build the SW and not worrying as much about the 
> philosophy and meaning as some people would like.  Pat Hayes is very 
> involved, he'd be a great foil for discussion of this stuff...
> 
> [1] http://www.w3.org/DesignIssues/
> [2] http://www.w3.org/DesignIssues/Meaning.html

Hi Jim,

Nice to hear from you again! 

First, as I admire your and Tim's work, I want to make it clear that this
is merely a "semantic" disagreement (about the semantics of "semantic"),
and not (necessarily) a philosophical disagreement -- and certainly not a
practical disagreement about the nature or value of the work.

I think the URL you sent me says it all: "Grounding the meaning of a
document in URI space" (or, as you put it, "symbol grounding (in web
space)": That's grounding symbols in symbols, not symbols in their
meanings!

So all I'm saying is that "The Syntactic Web" would have been literally
the right descriptor -- and still further testimony to how much
semantically interpretable mileage can be gotten out of pure syntax
alone! No shame in that!

Cheers, Stevan

PS The "Semantically Intrepretable Web" would be another straightforward
way of putting it, without needlessly turning the meaning of "grounding"
(or of "meaning") on its head...

> At 6:05 PM +0100 9/10/02, Stevan Harnad wrote:
> >Date: Tue, 10 Sep 2002 18:05:36 +0100 (BST)
> >From: Stevan Harnad <harnad@ecs.soton.ac.uk>
> >To: Michael Luck <mml@ecs.soton.ac.uk>
> >cc: Nick Jennings <nrj@ecs.soton.ac.uk>,
> >    Tim Berners-Lee WWW EJ <timbl@info.cern.ch>,
> >    James Hendler <hendler@cs.umd.edu>, ora.lassila@nokia.com
> >Subject: The Syntactic Web
> >X-ECS-MailScanner: Found to be clean
> >
> >Hi Michael!
> >
> >Your paper:
> >
> >>  Title:   Where are the Semantics on the Semantic Web?
> >
> >looks very interesting (and I wish I could be there!).
> >
> >I have some questions (based on the abstract -- if you have links to
> >full text, please send them to me!):
> >
> >>  In this paper, we address the question: How can we create a network of
> >>  semantically integrated communities on the World Wide Web? We first
> >>  clarify some confusion about what "semantics" means and introduce a
> >>  semantic continuum ranging from the kind of semantics that exist on
> >>  the Web today to a rich semantic infrastructure on the Semantic Web of
> >>  the future. We clarify what is meant by "semantic integration"
> >>  introducing and defining a "gold standard" whereby two agents that
> >>  have never met before can successfully exchange information.
> >
> >As far as I can tell, this is about coordinating the symbol use of many
> >(at least two) symbol-systems (software agents). Coordinated symbol use
> >is certainly potentially very useful -- but it seems to me it is not
> >semantics!
> >
> >Semantics involves 2 more components (actually 3, but the 3rd, conscious
> >meaning, we'd better ignore!):
> >
> >(1) The real objects (referents) to which symbols refer
> >
> >(2) The means by which the symbols are connected to their referents
> >
> >>  We acknowledge that this gold standard will only be reachable in limited
> >>  circumstances, and that a variety of approaches will be needed to
> >>  achieve successful agent interaction in practical situations on the
> >>  semantic Web. Towards this end, we introduce several architectures for
> >>  achieving semantic integration. Each are defined and compared on the
> >>  basis of how the following questions are answered. Who and when are
> >>  semantic mappings created between agent ontologies?
> >
> >As is (to some no doubt yawnfully) well-known in ECS, I've always had
> >problems understanding the notion of "ontologies": They seem to be used
> >in computation with a meaning almost diametrically the opposite of their
> >dictionary (or philosophical, for this is a term from philosophy) meaning:
> >
> >The dictionary meaning of ontology is: What there is, what exists, what
> >"reality" actually contains. Examples of things there are would
> >be people, chairs, storms, September 11 etc. (i.e., concrete objects,
> >states, events) plus greeness, roundness, differentiable functions, prime
> >numbers (i.e., parts and features of concrete objects, as well as abstract
> >properties). Examples of things there AREN'T would be unicorns (I don't
> >mean pictures of unicorns, I mean the creatures themselves), tachyons
> >(particles that move faster than light: ruled out by current-best physics
> >theory, when last heard from) and squared-circles (provably impossible).
> >
> >But the thing to notice is that "ontology" means those very
> >objects, states, etc. (that exist). (It can also mean a theory of
> >which things exist, Occam's razor, etc., but let's forget about that
> >meta-theoretic notion of ontology here.)
> >
> >In terms of the above, an ontology would contain (1), above, the actual
> >objects to which symbols (words) refer.
> >
> >And the symbol-grounding problem (the one I work on, on and off)
> >http://cogprints.ecs.soton.ac.uk/archive/00000615/index.html
> >is the problem of how to connect symbols to the real objects to which they
> >refer. For otherwise, mere symbol/symbol connections are "ungrounded"
> >-- their mean is not intrinsic to the symbol system, but needs to be
> >"projected" on them by the mind of an external interpreter.
> >
> >Now, for coordinating the exchanges of software agents it may not be
> >necessary to worry about the symbol-grounding problem at all. A great
> >deal of mileage might be derivable from symbol/symbol interactions
> >alone, without worrying about their meanings. Here "ontology" would
> >simply be a set of symbols that REFERS TO objects; not the objects
> >themselves.
> >
> >But then my only question is: Why would anyone want to call that
> >semantics, rather than the SYNTAX that it really is? Syntax is the
> >(algorithmic) manipulation of symbols purely on the basis of their
> >SHAPES (which are arbitrary in relation to their meanings, they are
> >merely a notational convention), not on the basis of their MEANINGS.
> >
> >"Semantics" is defined precisely in contrast to syntax: Syntax is
> >symbol-shape-manipulation, independent of meaning (apart from what an
> >external interpreter might project onto the symbols and manipulations).
> >Semantics is defined as what it is that the symbols actually mean (i.e.,
> >(1) and (2) above).
> >
> >>  Is the
> >>  architecture point to point between each agents, or mediated through
> >>  another ontology? What is the nature of agreements among the agents?
> >>  We conclude by making some predictions and reccomendations on how the
> >>  semantic Web will evolve in the coming years.
> >
> >I have no doubt that your algorithms will help coordinate the
> >interactions of software agents (i.e., of dynamic symbol systems, plus
> >data, i.e., static symbol systems). But I wonder why everyone does not
> >call the "Semantic Web" what it really is: the SYNTACTIC WEB! If the
> >syntactic manipulation of ungrounded symbols is sufficient to accomplish
> >all the things it accomplishes (in terms of coherent, semantically
> >interpretable -- by us -- symbols and symbolic agent-interactions),
> >then why not give the credit where the credit is due: to the syntax,
> >not the "semantics"?
> >
> >Best wishes,
> >
> >Stevan
> 
> -- 
> Professor James Hendler				  hendler@cs.umd.edu
> Director, Semantic Web and Agent Technologies	  301-405-2696
> Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
> Univ of Maryland, College Park, MD 20742	  240-731-3822 (Cell)
> http://www.cs.umd.edu/users/hendler
>

Received on Wednesday, 11 September 2002 11:44:24 UTC