On 7/13/2012 2:21 PM, Chris Menzel wrote:
> But, of course, that doesn't mean that a semantical model can't be
> exactly correct as far as it goes. BTW, it would be /great/ if in the
> future you numbered your slides. ;-) (01)
Re slides: The Adobe reader shows the slide number. To go to slide N,
just type N into the little box in Adobe. (02)
In any case, slide 23 leaves open the option that the model could
be exactly isomorphic (or even identical) to the subject matter.
But that case is possible *only* for mathematics -- or for
formally defined games like chess. (03)
To represent any chunk of the physical world perfectly, we'd need
some notation that could exactly represent every state of every
atom. That would run into quantum mechanical effects, and an
impossibly large amount of data. Approximations are the norm.
As engineers say, "All models are wrong, but some are useful." (04)
> What I meant of course was a predicate whose intended semantics
> is a relation between names and the things they name. (05)
I assumed that was your intention. But it's common practice in NLP
systems to use surrogates (such as GENSYM in LISP) as URIs for the
named entities. Computational linguists would normally use some
predicate to link the internal identifiers to the external names. (06)
> ... through the magic of Gödel coding, any theory containing a bit
> of number theory can construct such a predicate. I am talking about
> languages like that, in which semantical predicates... (07)
I have a very high regard for Gödel's achievement and related methods
in the foundations of mathematics. (08)
But very little, if any, of that work is of any relevance to AI, NLP,
cognitive science, or applications of ontology to the areas that
Ontolog Forum addresses. (09)
> you can introduce what you call semantical predicates and pretend
> they have a certain intended meaning, but if those meanings are not
> encoded in axioms and are not reflected in the semantics... (010)
I'm not pretending anything. I define them. (011)
> And I recommend R. L. Martin's classic Recent Essays on Truth and
> the Liar Paradox and the recent book Axiomatic Theories of Truth
> by Volker Halbach for understanding the issues I'm referring to. (012)
I read enough about the Liar Paradox, and I find it moderately
interesting. Those rare cases that occur in NL documents written
by people for the purpose of communicating with other people never
require any features beyond the Tarski hierarchy. (013)
> But the Tarski hierarchy doesn't solve the problem, it simply avoids
> it by restricting the languages in the hierarchy so that paradoxes
> can't even be expressed. (014)
By treating 'true' as an indexical, you can process English sentences,
including the Liar Paradox, in a way that classifies paradoxical
statements as ill formed because they create cycles in the hierarchy.
I have never seen any document written by people who are trying to
communicate with other people that requires anything more. (015)
The only exceptions are the 0.000001% of documents written by
philosophers for other philosophers. I am not against processing
that kind of language. But there are so many serious problems of NLP,
that I consider those issues to be a frivolous waste of time. (016)
> That's like avoiding illness by enclosing yourself in a sterile bubble.
> It might keep you healthy but it puts severe limits on what you can do. (017)
Comments like that sound like the kind of disease that Wittgenstein
said only philosophers ever succumb to. (018)
John (019)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (020)
|