W3C home > Mailing lists > Public > public-sw-meaning@w3.org > April 2004

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

From: Pat Hayes <phayes@ihmc.us>
Date: Mon, 12 Apr 2004 13:41:27 -0500
Message-Id: <p06001f4fbca07ee081e1@[]>
To: "John Black" <JohnBlack@deltek.com>
Cc: "Bijan Parsia" <bparsia@isr.umd.edu>, <public-sw-meaning@w3.org>

>  > From: Bijan Parsia [mailto:bparsia@isr.umd.edu]
>>  Sent: Sunday, April 11, 2004 11:23 PM
>>  On Apr 11, 2004, at 11:04 PM, John Black wrote:
>>  [snip]
>>  > In the RDF Semantics Recomendation it states:
>>  >
>>  > "1.2 URI references, Resources and Literals.
>>  > This document does not take any position on the way that
>>  URI references
>>  > may be composed from other expressions, e.g. from relative URIs or
>>  > QNames; the semantics simply assumes that such lexical
>>  issues have been
>>  > resolved in some way that is globally coherent, so that a single URI
>>  > reference can be taken to have the same meaning wherever it occurs."
>>  > - http://www.w3.org/TR/rdf-mt/#urisandlit
>>  >
>>  > What is the effect of the language, "...so that a single URI
>>  > reference can be taken to have the same meaning wherever it
>>  occurs."?
>>  > How important is this assumption to RDF semantics?
>>  Upon reflection, that isn't the best wording.
>>  Roughly: In the *graph* there are only absolute URIs. There
>>  also are no
>>  contexts, so every node labeled with the same uri is equivalent.
>>  *Between* graphs, however, URIs can behave quite differently
>>  (until you
>>  merge them).
>>  I'd say it's pretty important :)
>And merging graphs would seem to be a factor affecting my ability
>to say what I mean when using the descriptive naming provided
>by functional and inverse functional properties.
>Lets say I create a URI within a descriptive graph using,
>UsGov:SSNo :123456789 and foaf:mBox :mailto:JohnBlack@mysite.com
>as inverse functional names (properties) for my concept of me. 
>Then I can use this descriptive graph to signify me in an on-line
>resume, for example, and say that my title is 'Senior Software
>Architect'.  But so can anyone else use those inverse functional
>names in descriptive property graphs to make statements.

Sure. People can lie, using the same vocabulary as is used to speak 
truth. Sorry about that. Adding to the vocabulary doesn't help, 
however.  If we introduce an explicit "this is true" predicate, then 
liars can still lie using this larger vocabulary: all that we get 
that is new is the possibility of expressing the liar paradox by 
constructing self-referential lies.

>may feel the need to dispute me and put a graph on his web site
>saying that UsGov:SSNo :123456789 has the foaf:mBox
>:mailto:Peter@hissite.com and has the title 'President of the
>United States'.  When the sw-google global search/inference engine
>merges the two graphs, what happens to my meaning?

That question is meaningless. Your INTENDED meaning isn't changed by 
anything Peter says. If Joe merges yours and Peter's graphs then it 
would follow logically (and Joe's OWL reasoner could come to this 
conclusion) that you were Peter, with two titles. No doubt with a bit 
more information the reasoner could detect an inconsistency.  The 
RDF/OWL specs do not specify what to do if one finds an 
inconsistency. Some intelligence may be required at this point.

>  Who are my
>statements about?

This question is meaningless. A chunk of OWL is 'about' whatever 
things (are in the universe of interpretations which) satisfy it. As 
you may have figured out from my recent exchange with Dan Connolly, 
there is as yet no a generally-agreed-on way to formally connect this 
'rigidly' to the actual world which you and I inhabit, although there 
are plenty of techniques which seem to work well enough in practice 
for many purposes

>So how do I communicate to sw-google that when I post an on-line
>resume in RDF, and post a descriptive RDF graph to signify me, so I
>can make RDF statements about me, that it should interpret my graph,
>and not Peter's, to determine who I meant to be the subject of the RDF
>statements of my resume?

That is the key, seems to me. When getting information about YOU, I 
will tend to trust sources that originate with you more than I trust 
random sources originating from others. But its no good expecting 
RDF/OWL to provide this kind of correspondence between description 
and origin, since by their very nature, RDF and OWL are supposed to 
be the common languages that anyone can use to say anything (that is 
sayable in RDF or OWL, that is.)  So liars can use them just as 
easily, and with the same meanings, as trusted truth-speakers. A 
public vocabulary can't, by its very nature, provide a secure basis 
for trust based on origins, because anything A can say, B can say 
just as easily. You need some way to check the signatures, externally 
to the RDF/OWL, so that a mere falsehood becomes an actual forgery, 
in order to get a bit more security into the picture. But that goes 
outside RDF/OWL semantics. Some of us have some ideas about how to do 
this neatly, but they havnt made it to anything like W3C WG status 

>It seems to me that if a human interprets the RDF, they might parse
>the mailto:JohnBlack@mysite.com semantically and realize that the
>@mysite.com is the same as the mysite: prefix of the resume
>statements on my site and conclude that statements located on my site
>were more authoritative about who I was referring to than the statements
>made on Peter's site.  But RDF/OWL engines are not required to do this,
>because the RDF semantics ignore the URI semantics.

They are not required to do it in order to be RDF compliant, but that 
doesn't mean they are obliged not to do something like this. The RDF 
WG was chartered not to, and so deliberately did not, go into issues 
of trust and provenance. You are right to emphasize that they will be 
needed, and people are thinking about that right now.

>"There are several aspects of meaning in RDF which are ignored by
>this semantics; in particular, it treats URI references as simple names,
>ignoring aspects of meaning encoded in particular URI
>forms [RFC 2396]..." - http://www.w3.org/TR/rdf-mt/#intro
>These topics were discussed extensively last year.  But while I can see
>the value of the ability to dispute the factual validity of RDF
>statements, using RDF statements, I don't see the same value in the
>ability of someone to dispute who or what I mean to refer to by a name
>or a term or a descriptive graph.

It is impossible to distinguish these topics.

>To go back to a previous example, it doesn't seem that useful
>if I were a math teacher and wrote on the board, as the preliminary
>givens of an example, 'let a = 3' and 'let b = 4' and one of my
>students jumped up and said 'No, no, no.  I dispute that a = 3'  On
>the other hand, if I state that 3 = 4, there is an obvious need for
>others to be able to dispute that and say, '3 <> 4'.

Suppose A states that a=3 and B states that a=4 and C states that 
3=4. Then if we take what A says at face value, it follows logically 
that what B and C say amount to the same thing, since if a=3 then you 
can derive 'a=4' from '3=4' and vice versa.  Disagreement over the 
facts, and disagreements over meanings of names, come down to the 
same thing when facts can express identity between names.


PS Great story about Bertrand Russell. When 'principia mathematica' 
came out there was a lot of buzz in academia about this new formal 
logic stuff. One of the puzzling things was that you could derive 
anything from a contradiction. Apparently at a dinner party someone 
sitting next to Russell challenged him to prove that if 1=0 then he 
was the Pope, and he answered immediately: if 1=0, add one to each 
side of the equation, then 2=1; the Pope and I and two; therefore the 
Pope and I are one.

>>  Note that URIs in literals (e.g., in literals of datatype xsd:anyURI)
>>  are exempt from this merging. So the above text isn't quite right if
>>  you try to read it in full generality.
>>  Cheers,
>>  Bijan Parsia.

IHMC	(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32501			(850)291 0667    cell
phayes@ihmc.us       http://www.ihmc.us/users/phayes
Received on Monday, 12 April 2004 14:41:42 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:42:16 GMT