Re: How does RDF/OWL formalism relate to meanings?

John Black wrote:
>>From: Thomas B. Passin

>>Well, I'm tired of your trolling tactics.
> 
> 
> "Trolling: Deliberately posting false information in order to 
> elicit responses from people who really want to help." -
> http://www.datatronics.net/glossary_of_web_terms.htm#T
> 
> I sincerely appologize to you, Thomas, and to the other readers 
> of this list for my trolling tactics.
> 
> Specifically, when you truly helpfully responded to my 
> semi-rhetorical, semi-searching questions as though I was a 
> complete newbie writing to this list to ask "How do I use the 
> semantic web?", I should have stopped and explained that I had 
> been participating in this public list for some time because I 
> have opinions and ideas about the resolution of TAG issue 
> rdfURIMeaning-39. 
> 

OK, now we have this behind us.  And besides, you really have been 
asking about some real and hard matters, and Pat Hayes has posted a very 
cogent post or two a bit ago that capture them well.

> However, I really did think you knew that my examples about 
> my company and its employees and my strong employee were just 
> that, examples to use as thought experiments.  

I would like to suggest that the way this thread has been spinning out 
has not helped any of us to formulate, let alone answer, the issues very 
well.  That is why I asked for something more concrete, to have 
something to focus on.  Pat, of course, is well capable of getting past 
this weakness and has done been doing so, with more knowledge and 
eloquence that I can.

  Pat Hayes -

"This is exactly the point where John's question has some force: HOW 
does one assign a resource to a URI?  What do you DO to get the 
assignment done? ... But how can you attach a unique referent to a name 
by making assertions (in OWL, say)? All you can do it to relate URI 
referents to other URI referents. You have to have some referring names 
to get the process started."

In the face of the difficulty, or even seeming impossiblility of 
attaching a precise and rich denotation to a (word/term/URI/...), then, 
how in fact can we proceed?

Not that I have *the* answer - no doubt there is not a single one 
anyway.  In the face of lack of understanding we have to start simply, 
or at least I do.  Here are a few things I have observed that I think 
are potentially useful -

1) Sometimes, manipulation of the symbols is enough.  They do not have 
to be grounded in a human understanding.  For example, consider a 
database where the allowed range of a variable is specified by a logical 
formula.  The database code can perfectly well check for valid entries 
by applying this formula without having the slightest notion of what any 
of the symbols mean.  A lot of math is like that, and, really, 
practically all of logic.

For these cases, then, we hardly have to attach any "meaning" to the 
symbols - translate to "symbol" URI if you like.  It is enough to be 
able to use them in accordance with some formula, code, or constraint. 
Probably every Prolog program you have ever written falls into this 
category.

2) Sometimes it will be enough to inform a machine (or a person) that 
the new term is equivalent to one the system already knows about.  How 
did it find out about that other one?  Never mind, just say it is given. 
  I bet you don't know how you learned the meaning of the word "turtle", 
but it doesn't matter.

3) Sometimes it will be enough to say how and where a term can be used. 
  I'm especially thinking of range and domain here, but of course they 
are not the only possibilities.

4) Sometimes a term (or read "resource" or "concept" if you like) can be 
known well enough for some purpose by its relations with other terms.

OK, I know you know all this but I am trying to cast it into a useful 
progression in the hopes that we may start to get somewhere.

So far, one can see how RDF/RDFS/OWL/... and URIs can work for items 1) 
- 4).  Wouldn't you agree?  And the URIs could even be useful because at 
least you can know that the same symbol is being talked about, and you 
might even be able to dereference the URI and get some useful 
information about it (never mind whether it would be human or 
machine-usable for now).

One of the points here is that how thoroughly grounded a term must be 
depends on its projected use and users.  That's where I wanted John to 
get more concrete.

Beyond items 1) - 4), things would appear to get more difficult in terms 
of relating the meat of a concept/resource to an arbitrary term.  If you 
read up on how children learn words, it is incredibly complex, and also 
there is no one way they do so.  Furthermore, there is generally, 
probably always,  a social component to that learning.

I would say that one of the tasks at hand is to learn how to make as 
much mileage as possible out of items 1)-3), and to learn to to make 4) 
more feasible and more powerful.  Beyond that, things get pretty murky, 
at least me.  Do we have to have Cyc-like "common-sense" reasoning 
abilities to make any headway?  Can we make progress along the lines of 
Doug Hofstader's analogical Lisp programs that solve series-completion 
problems?  Or what else?

I don't know.  But I think we can do a lot using just items 1) -4). 
That will not let either a human audience or a computer "understand" the 
depths of your notion of "strength", but I agree with Pat that they 
usually won't have to.

Let me just close by returning to the notion that to make this 
discussion productive, we need to work out how to engage in it in a way 
that could lead somewhere.   That in itself seems hard to do, doesn't 
it? It might require setting aside for the time being some of the 
weightier philosophical issues, I don't know.  For example, if we ask 
too hard about what a "resource" is, we are likely to get into an 
interminable discussion about Platonism, but I don't think that will 
help us.  That hasn't been resolved for millenia, and if we have to 
resolve it to make progress then we are dead in the water.

Well, I better stop here because I am getting way out of my depth.

Cheers,

Tom P

Received on Friday, 9 April 2004 20:36:42 UTC