Re: MISC: Internet Media Type registration: proposed TAG finding

>On Wed, 2002-05-29 at 11:35, patrick hayes wrote:
>[...]
>>  >It wouldn't be unprecedented, by the way:
>>  >
>>  >   http://www.w3.org/2000/10/swap/log#Truth
>>
>>  Wow, scales fall from my eyes!
>
>Why is this so crazy? It doesn't seem interestingly
>different from wtr in KIF.

The difference is fundamentally that the meaning of wtr in KIF is 
specified by a KIF model theory, not by English comments in a 
particular KIF document. So the KIF is (1) mathematically precise (2) 
part of the actual language specification (3) part of a semantics 
(and also, by the way (4) provably equivalent to a complete deductive 
system, although this isnt described in the KIF specs.)  None of that 
is true of the swap file.

>[...]
>>  Seriously, that document (1) does not define logical truth in any way
>>  whatsoever (2) says this: - <rdfs:Class
>>  rdf:about="http://www.w3.org/2000/10/swap/log#Truth">
>>     <rdfs:comment>Something which is true: belive it as you would
>>  belive this. Understood natively by cwm in that it will execute rules
>>  in a formula declared a Truth within a formula it is already taking
>>  rules from.</rdfs:comment>
>>
>>  which seems to indicate that log#Truth in fact is simply supposed to
>>  mean 'asserted', which is perfectly meaningful, but is not the same
>>  as 'logically true'.
>
>Er... close to that; it's a de-quoting mechanism.

OK, that's fine. However, I note that this isn't specified in the RDF 
spec anywhere.

>  > And in fact, the document does not define *any*
>>  meanings in RDF, or constrain the RDF interpretations, in any way
>>  whatsoever. It is just English with RDF decorations added. (The CWM
>>  code might be said to be a kind of implicit machine-readable
>>  constraint on interpretations of this vocabulary - along the lines of
>>  'this means whatever it takes to make CWM produce valid conclusions'
>>  - but it goes well beyond what an RDF engine would be able to make
>>  use of.)
>
>[...]
>>  >>  Who discovers this, and how?
>>  >
>>  >As explained above, I (i.e. anybody using
>>  >the framework) use the deployed URI infrastructure
>>  >to dereference http://www.daml.org/2001/03/daml+oil#UniqueProperty
>>  >and I see:
>>  >
>>  ><rdfs:Class rdf:ID="UniqueProperty">
>>  >   <rdfs:label>UniqueProperty</rdfs:label>
>>  >   <rdfs:comment>
>>  >     compare with maxCardinality=1; e.g. integer successor:
>>  >     if P is a UniqueProperty, then if P(x, y) and P(x, z) then y=z.
>>  >     cf OIL FunctionalProperty.
>>  >   </rdfs:comment>
>>  >   <rdfs:subClassOf
>>  >rdf:resource="http://www.w3.org/1999/02/22-rdf-syntax-ns#Property"/>
>>  ></rdfs:Class>
>>  >
>>  >The comment there is reasonably clear as a constraint
>>  >on interpretations, no?
>>
>>  NO!!! It is not in any way a constraint on interpretations, any more
>>  than a comment in a program is code.
>
>How is that not acceptable as a constraint on interpretations, but
>stuff like this is?
>
>
>[[
>for ?D an XML Schema datatype, IO(?O) is the singleton set containing
>the element of IC(?D) that has lexical representation ?L, provided that
>there is one, otherwise IO(?O) = { }
>]]
>  -- http://www.w3.org/TR/2001/NOTE-daml+oil-model-20011218#3

The difference is that this stuff is in a description of the model 
theory, and is part of the language specification intended to be read 
by humans, not a comment attached to a piece of the formal language 
intended to be read by software. Nobody expects this to be readable 
by machines. But that earlier rdfs:comment isn't in a specification 
document intended for human readers, it's in a piece of formal RDF.

I agree with you that this particular rdfs:comment is so exact that 
it COULD be part of a semantic specification. Unfortunately, however, 
as a matter of fact it ISNT part of a semantic specification. It's 
labelled "comment",  and its not part of a published specification 
document, and it occurs inside some RDF, which is supposed to be 
machine-readable.  Comments, pretty much by their nature, are not 
parts of a formal semantic specification and are not supposed to be 
machine-readable.

>  > A comment is a COMMENT, and that
>>  is all. YOU can read that and understand it, Dan, because you are a
>>  HUMAN BEING WHO UNDERSTANDS ENGLISH. The whole point of the semantic
>>  web is to allow SOFTWARE AGENTS to do a little understanding.
>
>Yes, so, I read the comment and write some code.

But we are talking about something that is supposed to be readable BY 
THE CODE, not by YOU. Thats why we are working on all these 
interminable specification documents, right? So that people can write 
code which will process the RDF unambiguously, and all their various 
pieces of code will do the right things with one another. RDF isn't 
meant for people to first read, then having read it, write code to 
interpret it. Its meant to be read by code that was already written 
by people who have never seen the particular RDF graph, but have read 
the RDF spec. Would you expect HTML to work if I had to re-write the 
code of my browser in order to read each web page?

>  > When
>>  you can write a Perl script that can figure out the content of the
>>  English comments, then maybe you can claim that the meaning of the
>>  comments is part of the meaning of the formalism. It still wouldn't
>>  be part of RDF, but you could call it RDFE .
>
>It seems to me to be a perfectly reasonable part of
>the Resource Description Framework. I guess I'm willing
>to call it an extension to RDF, if that makes you happy.
>But it seems pretty arbitrary, to me.

It seems absolutely fundamental to me. Why are we even bothering to 
write a spec for RDF if one can include anything in it by writing 
arbitrary scripts?

>[...]
>>  >Let's put it this way: does dublin core fit into the framework?
>>  >Or is RDF+dc an extension? How about a document that
>>  >uses RDFS, DAML+OIL, and dublin core together? Is that
>>  >another sort of extension?
>>
>>  Dan, you cannot possibly be this obtuse. Surely you know the
>>  distinction between a language and a set of axioms in that language.
>>  DC is a set of assertions in RDF, right?
>
>No; it's a set of terms defined by a community of practice.

OK, then it isn't representable in RDF. You can't have it both ways. 
If RDF is warm and fuzzy and determined by human societal use, then 
it isn't processable by machines (until we have AI licked). If its 
processable by machines then all the formal meaning in it is 
determined by the machine-processable aspects of the spec: in the 
case of RDF, by RDF entailment of some kind. Of course humans might 
declare in some human-readable way (eg in a comment) that some RDF is 
*intended* by them to mean something, and maybe then they can be held 
to account for having published that and tried by a human jury and 
fined by a human judge and whatever. All true, and all totally 
irrelevant to the machine-processable aspects of RDF meaning. I could 
get prosecuted for writing libelous graffiti, but that doesn't make 
graffiti into RDF.

>There are semantics to the dublin core terms that aren't
>written in the RDF specs.

Of course. But can they be represented in an RDF graph,  so that 
given the RDF graph and the RDF specs together you can figure out the 
semantics? If not, then claiming that there are such semantics, and 
simultaneously that the dublin core is 'in' RDF, seems to me to be 
close to dishonest. If RDF is a formalism, then RDF can only express 
RDF-expressible content.  And if RDF isn't a formalism, then why are 
we bothering to write all these formal specs that sure make it LOOK 
like a formalism? Are we trying to fool somebody, or what?

(Sorry if I sound like I feel strongly about this, but I do feel 
strongly. Ive been worrying and writing about what it means for a 
formal language to express meanings all my working life, and believe 
me, some pretty silly stuff has already been written; but at least in 
AI we now have a fairly clear overall picture of what it means to 
represent knowledge, including what it DOESNT mean. I joined the RDF 
effort in good faith to try to help make the SW happen. Now I feel 
like Ive been conned into signing a religious manifesto.)

>  > That means that one can
>>  interpret those DC assertions using the RDF semantics and they mean
>  > something that approximates to what the DC writers had in mind,
>>  roughly. If you do that to DAML, you often *don't * get an
>>  approximation of what the writer of the DAML had in mind.
>
>I must be quite obtuse; I don't see any fundamental
>difference between partial understanding of DC semantics
>and partial understanding of DAML semantics.

The difference is that you are here making a pun on 'semantics' . 
That word is used informally to just mean 'intended meaning' and 
formally to mean something like a model theory. Choose one sense and 
stick to it. In the second sense, DC doesn't even have a semantics. 
In the first sense, "DAML semantics" could be anything.

Basic AI/KR insight: if you formalize a thought or an idea, you can't 
just assume that what you had in mind is what is actually captured in 
the formalization. You have to examine the formalization carefully, 
poke it and shake it, to see what it really has in it.  Just writing 
"Loves(Pat, Jackie) " does not, by itself, capture that idea that I 
love somebody called Jackie. No amount of talking *in English* is 
going to change this. I can say it means that on my home page, write 
letters to the NY Times about it, put up billboards, stand on street 
corners and yell that that is what it means; none of that *makes* it 
mean that.  The only way to make a formal sentence mean something is 
to say enough *in the formalism* to constrain the interpretations *of 
the formalism* sufficiently, and/or to somehow ground the formalism 
to reality, e.g. by perceptual machinery.

Now, maybe English(/French/Kanji...) meanings could be used as a kind 
of social grounding, so that linking the formalism to NL sentences in 
a way that uses other people's experience as a kind of remote 
grounding is itself a social way of attaching meanings. So I can use 
the name 'Melbourne' with confidence even though I have never seen 
the place, never will, and have no idea what it looks, feels, sounds 
or smells like, because Im pretty sure that other people somewhere 
else have done enough seeing, feeling, listening and smelling to 
ensure that the name pins down a unique referent. That is what your 
vision of the 'greater RDF' sounds rather like. And that is indeed a 
great ambition; I mentioned it (briefly) in a paper I wrote in 1985. 
But we aren't going to achieve that by writing little essays inside 
rdfs:comments; we would need some genuine NL understanding software 
to be incorporated into the semantic web. Then (some of) NL will have 
been integrated into the formalism itself, and that *certainly* will 
not be RDF. Again, don't get me wrong: that's a fine long-term 
ambition to have for the SW. But please don't call it 'RDF', because 
that gives the impression that you are claiming that RDF 1.0, this 
useful but rather wimpy little language, can *already* incorporate NL 
meanings, human intentions, and arbitrary semantics; but in fact, it 
can't, no matter how much we put in the rdfs:comment strings.

Pat
-- 
---------------------------------------------------------------------
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 Tuesday, 11 June 2002 19:32:31 UTC