W3C home > Mailing lists > Public > www-rdf-logic@w3.org > June 2001

Re: rdf as a base for other languages

From: pat hayes <phayes@ai.uwf.edu>
Date: Fri, 1 Jun 2001 15:58:26 -0500
Message-Id: <v04210111b73dadd0a476@[205.160.76.212]>
To: Sandro Hawke <sandro@w3.org>
Cc: www-rdf-logic@w3.org
>In message <06c801c0eab9$dcada230$0a2e249b@nemc.org> you wrote:
> >Sandro Hawke wrote:
> >> The only potential for confusion I see is that some people might want
> >> to jump from having a triple described (with ground facts) to assuming
> >> the described triple is true, but that seems clearly wrong.
> >
> >Calling something a "fact" implies that it is "true". You might try to
> >assert a falsehood but that would be false or inconsistent. To be consistent
> >you must assert a fact, i.e. the fact _must be_ true. What am I clearly
> >missing?
>
>I don't understand where we're not understanding each other.
>
>For example, here are 7 facts:
>  1.  I can imagine a condition, X.

What is a 'condition', exactly?

>  2.  X can be expressed accurately as English sentence Y.
>  3.  Y has four words.
>  4.  The first word of Y is "the".
>  5.  The second word of Y is "sun".
>  6.  The third word of Y is "is".
>  7.  The fourth word of Y is "shining".
>
>Now I'm not exactly sure which of X and Y might be best called a
>"condition", "state", "statement", "declartion", "sentence",
>"situation", etc, etc. but using only true ground facts I have
>essentially communicated X and Y to you without asserting them.

I have no idea what X is supposed to be, but you havnt communicated Y 
to me. You have DESCRIBED Y, but you havnt communicated it. Now, of 
course, you and I are educated human beings who have enough wit to be 
able to go back and forth between words and meanings without even 
being consciously aware of it, but that talent is taking up about 35% 
of the processing power of our two associative cortexes, and nobody 
on God's earth knows how our brains do that. It might as well be a 
miracle.

This illustrates nicely why we shouldnt use intuitions from ordinary 
human discourse as a guide to good practice. You used a magical 
phrase in there :"X can be expressed accurately as sentence Y". That 
is a (miraculous) de-reification statement: it talks about a sentence 
and about its meaning, and asserts they are connected.

>If you take everything I say as true and I add an 8th fact: "X is
>true" or "X is a fact" then you are licensed to infer that the sun is
>shining.
>
>If instead I add an 8th fact: "X is not true" then I have made no
>contradiction; I have actually made a reasonable and useful statement,
>licensing you to infer that the sun is, in fact, not shining.
>
>Do you have any problem with this approach, beyond style?

It isnt an 'approach'. It works for human beings, but it does not 
work for machines. It would require AI to have been done already, and 
AI has hardly got started.

>I'm sorry I haven't looked at your strawman yet, but I think it has
>a somewhat different syntax for facts like:
>
>   1.  Statement 1 is an asserted statement that I can imagine a
>       condition, X.
>   2.  Statement 2 is an asserted statement that X can be expressed
>       accurately as English sentence fact 3.
>   3.  Statement 3 is an UNASSERTED statement that the sun is shining.
>
>The differences between these approaches seem trivial.  Any formal
>language using the first approach can be machine translated to & from
>any formal language using the second approach, given a vocabulary for
>(1) describing the parts of a sentence and (2) asserting sentences.

It stops being trivial when the sentence being asserted requires, in 
order to be asserted, an expressive power beyond that used to 
describe it. What do you think it means to assert, say, KIF 
sentences? To assert a sentence it is not enough to just display a 
sequence of symbols: one needs an engine of some kind that can 
understand, in some useful sense, the LANGUAGE that the sentence is 
written IN. You and I can understand English, but that is because we 
are damnably smart and have specialized language-processing circuitry 
in our heads. There are fairly hairy pieces of code that can, in a 
sense, understand KIF (say), in the sense that they can draw valid 
conclusions , construct/check proofs, and so on. But one does not get 
that ability just by having KIF sentences assembled and injected into 
one: it has to be provided. In fact, even to *parse* the sentence 
being constructed here (your "Y") requires knowing quite a lot about 
the language in question.

So for example if you are going to dereify sentences using KIF's 
'wtr', as Dan Connolley suggested, then you need enough of KIF to be 
already implemented to be able to draw sufficient conclusions 
involving 'wtr' and your KIF sentences to get done whatever job needs 
doing. In other words, to use KIF, you need a KIF engine. If you know 
how to compile a KIF reasoner mechanically from RDF, I would suggest 
that you get it written and patented as soon as possible.

Pat Hayes

---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes
Received on Friday, 1 June 2001 16:58:35 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:40 GMT