Re: An intuition pump

At 8:25 AM -0400 9/24/03, Peter F. Patel-Schneider wrote:

>I don't think that the web is qualitatively different from even written
>natural language in this respect.  Readers of natural languages (even
>software) can agree on the words in an written sentence - these are, to me,
>quite similar to the boxes you mention above.
>

no. no, no - this is very different -- in English the string "person" 
may have many definitions.  On the web the string 
"http://www.../foo#person" may be said to refer to many concepts, but 
the label is unambiguous - think of it as if in English I was to write
  "No(1), No(1), No(1), this(17) is(42) very(6) different(7)"
where each of the subscripts refers to a specific definition on a 
specific page of a specific dictionary.   This would not make all 
problems go away magically (or we wouldn't need this group), but it 
is not the same as the first sentence I typed because you have a 
place to reference to somehow learn more about the intended semantics.


>I do agree that there could be a qualitative difference between the
>Semantic Web and natural languages, as the Semantic Web *could* end up with
>many more ``words'' than even English.  (It could also end up with much
>*fewer* ``words''.)  This could be taken as a chance to have a single-box,
>single-meaning situation.  However, as Pat has pointed out, single meanings
>can easily split into multiples when looked at more closely.
>
yes, but it would require a new URI if we then wanted to name one of 
these new concepts.

>>  >When are two systems in the Semantic Web referring to exactly the same box?
>>  >I see very, very few chances for this to happen.
>>
>>  whereas I see this all the time - our OWL page uses lots of links to
>>  other people's boxes, and then displays what is there when
>>  appropriate -- so we have a link to the W3C RSS feed without knowing
>>  what messages will appear on our home page when that link is
>>  exercised (via a RDF Query)
>
>If by box you mean URI reference, then sure, just as people use other
>people's words.  Common meaning is, I think, another situation.

well, maybe - most people coming to http://www.cs.umd.edu/~hendler 
seem to expect to find my web page there... common "meaning" is hard 
to discuss, common referent seems a lot more straight forward, and 
something to ground the meaning discussion on much more precisely 
(See Harnad's many papers about symbol grounding and language)

>
>>  >>  So, in a sense, I interpret Tim's approach as "The owner of the URI
>>  >>  gets to define what HE/SHE/IT means by that URI"   anyone else is
>>  >>  welcome to say things about the owner, the predicate, the URI, etc -
>>  >>  but they cannot change the "meaning" of that specific URI unless they
>>  >>  do it in their space - in which case it is their claim about the
>>  >>  meaning, not the owner's (and their URI is where they state what they
>>  >>  claim the meaning is).
>>  >
>>  >I really don't understand what you could mean by ``meaning'' here.  I don't
>>  >see that it can be ``denotation'' or ``defining information'' or anything
>>  >else that I can relate to.
>>
>>  well, my point on this one is simple too - I don't understand those
>>  either, so let's not use those words -- I still am looking for a test
>>  case where I don't need to dereference peter:meaning to see what
>>  breaks.
>
>Well, how about ``Different systems produce different results when given
>the same initial set of information''?

I would love an operational definition like this - now can we find an 
example? (I've been trying to create one, but keep getting stymied 
because I'm not yet sure the implications of some of the different 
quasi-proposals floating around)

>
>>  >>  Note also that a URI with nothing there (i.e. no dereferencerable
>>  >>  document or a non-document URI) works in the above anyway - the owner
>>  >>  makes no claim as to the meaning of the URI, but other systems
>>  >>  grounding at that URI can at least agree they are refering to the
>>  >>  same "box" despite it's null content
>>  >
>>  >Well, I would like to be able to have the Semantic Web concern itself with
>>  >something besides boxes.  Perhaps all that you mean here is a common
>>  >vocabulary.
>>
>>  I mean building common vocabulary more easily because "symbol
>>  grounding" is an essentially solved problem - the main difference
>>  between KR on the Web and KR in the real world IMHO
>
>[...]
>
>Huh?  I fail to see how the Semantic Web provides any traction here at all.
>I even think that the Semantic Web makes the ``symbol grounding'' problem
>harder because it beings some of the thorny issues in KR closer to the
>fore.  My view is that this effort should try very hard to come up with
>partial solutions that don't need to open up these cans of worms (while
>still not excluding incorporation of the worms into future extensions).

I agree with the goals, although as you know we don't always agree on 
how to get there :->

>
>Peter F. Patel-Schneider

-- 
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	  *** 240-277-3388 (Cell)
http://www.cs.umd.edu/users/hendler      *** NOTE CHANGED CELL NUMBER ***

Received on Wednesday, 24 September 2003 10:02:44 UTC