W3C home > Mailing lists > Public > public-xg-urw3@w3.org > July 2007

RE: [URW3 public] Re: [URW3] ... three questions based on the last telecon

From: Mike Pool <mpool@convera.com>
Date: Mon, 16 Jul 2007 06:54:38 -0400
Message-ID: <6B1CD4A11E730446897E4715B119338A431777@vnmail.convera.com>
To: <Peter.Vojtas@mff.cuni.cz>, "Ken Laskey" <klaskey@mitre.org>
Cc: <public-xg-urw3@w3.org>

Hi, Peter,

See my responses below within [mjp][/mjp] tags.


-----Original Message-----
From: Peter Vojtas [mailto:Peter.Vojtas@mff.cuni.cz] 
Sent: Monday, July 16, 2007 4:24 AM
To: Ken Laskey
Cc: public-xg-urw3@w3.org; Mike Pool
Subject: [URW3 public] Re: [URW3] ... three questions based on the last

Dear colleaguess (sent to public list and separately to KL and MP),

as I have pointed in the ontology page in Top Level comments by P. 
Vojtas is there a mistake?, see

by W3C standards, basic information unit is a triple (subject, 
predicate, object) which can be true or false in a structure (to avoid 
discussion whether it is a sentence or proposition, w3c uses statement).

An RDF triple is designed to represent something, i.e., propositions,
but an RDF statement is just another means of expression.  If you write
a sentence in RDF triple terminology and I write the same one in KIF, in
virtue of what are they equivalent?  In virtue of the fact that they
represent the same proposition, I would claim.

Perhaps we want to argue that we should use RDF triples to represent
these propositions, and in terms of representational syntax, say, for
example, as you do below, that triple3 hasUncertainty Y.  (That's along
the lines of what I'd do but I'd argue that we want to allow for a
formalism in which 'hasUncertainty' can link an uncertainty value to any
well-formed formula (in a full first order logic) and that we may need
another relation, hasUncertaintyAccordingTo, that could be a ternary
relation linking wffs, agents and uncertainty values.)   I'm a little
squeamish about constraining ourselves to triples from the outset as
they notoriously lack sufficient expressiveness without appealing to ad
hoc "reified relationships" (which also seem to pull us out of our
first-order world).  (To take a simple example, I can't
straightforwardly represent the fact that Joe was a Microsoft employee
in 2001 using triples, so I think it's a bit dangerous to assume from
the outset that triples will be adequate for the representational needs
we might have. ) 

I think that the representational adequacy of RDF, and the triple
vocabulary, for uncertainty reasoning should be one of the things that
we talk about.  So our ontology and representational approach shouldn't
assume at the outset that it is adequate for representation.


see e.g. http://www.w3.org/TR/rdf-primer/ for following example

http://www.example.org/index.html has a creation-date whose value is 
August 16, 1999

here we can use reification for another writing asigning an identifier 
to the statement

ex:triple1	rdf:type	rdf:Statement
ex:triple1	rdf:subject 	http://www.example.org/index.html
ex:triple1	rdf:predicate 	http://purl.org/dc/elements/1.1/creator
ex:triple1	rdf:object 	http://www.example.org/staffid/85740
ex:triple1	ex:creator	http://www.example.org/staffid/85741


especialy in our ontology discussion the triple

urw3:Sentence  	urw3:hasUncertainty  urw3:Uncertainty

should be by my opinion rewritten as

urw3:triple2	rdf:type	rdf:Statement
urw3:triple2	rdf:subject 	urw3:Sentence
urw3:triple2	rdf:predicate 	urw3:hasUncertainty
urw3:triple2	rdf:object 	urw3:Uncertainty
urw3:triple2	ex:creator	urw3:Mitch

and instance ( consider also an "instance"  ex:triple1 
urw3:hasUncertainty  urw3:0.9)

as, e.g.

urw3:triple3	rdf:type	rdf:Statement
urw3:triple3	rdf:subject 	ex:triple1
urw3:triple3	rdf:predicate 	urw3:hasUncertainty
urw3:triple3	rdf:object 	urw3:0.9
urw3:triple3	ex:creator	urw3:Peter
urw3:triple3	urw3:tool	urw3:Bayes

Yes, this seems reasonable to me, but I don't think it goes far enough
to handle all our representational needs.

For example, I may also want to say:
((urw3:triple3 AND urw3:triple4) OR urw3:triple9) hasUncertainty .1

Or that the probability that Joe worked at Microsoft in 2001 is .3, or
that according to Fred the probability that Joe worked at Microsoft in
2001 is .2 and according to Sally it is .4.


The example with the german sentence (Mathias can help) is very usefull 
(words morgen and Morgen are problematic) because it shows what can 
happen. MP assigns an uncertainty to his own translation by expression 
"if I've translated correctly". Nevertheless, by my opinion "Es regnet 
morgen" is a problematic sentence because "Es regnet" is about present 
and "morgen" with lower case m in the begining means tomorrow, it is an 
adverb. So a problem accurs, what to do with a gramaticaly wrong 
sentence (note that the word sentence I use here in the linguistical 
sense, which in the W3c terminology can by defined appropriately by 
corresponding triples, by defining predicates like subject, verb, 
object, mode (manner), place and time). I would say either "Es regnet 
heute morgen" or "Es wird morgen regnen".

I would just chalk this up to my trying to use Babelfish to get a German
sentence.   A malformed sentence can just be ignored, I think.

Sorry for such a long mail, concluding I would like to say, please let 
us use w3c terminology (arguing for necessity of an extensions of 
standards we need be compatible with current standards). Next, the above

example shows we need to define our own prefix and rdf vocabulary for 
uncertainty ontology.

Thanks for your response, much appreciated.  I now fully understand the
point you were making in the meeting.

Greetings Peter

>     *From:* Mike Pool
>     *Sent:* Friday, July 13, 2007 3:25 PM
>     *To:* public-xg-urw3-request@w3.org
>     *Subject:* three questions based on the last telecon.
>     Hi, all:
>     We've been having some great discussions during the meetings and
>     like to pick up a few threads that came up in the last meeting.  I
>     reread these as I was trying to write up the minutes:
>     1)
>     Peter suggested that we use w3c standards as our guide for the
>     meaning of 'proposition'.  Peter, do you know if this is defined
>     somewhere by the w3C.  Could you point us to the definition?
>     2)
>     I argued that propositions, in the sense of the meaning of a
>     sentence that is invariant through all the paraphrases and
>     translations of the sentence, rather than assertions or sentences
>     the kinds of things that hold probability values. 
>     Kathy noted in the meeting that a problem with this definition is
>     that a system might assign different uncertainty values to 2
>     different logically equivalent sentences.   I can imagine that
>     is possible, but where it occurs I would think it nothing more
>     a weakness in the system, not in the definition I've suggested.
>     example, I might misunderstand 'Es regnet morgen' as 'it will rain
>     this morning' rather than 'it will rain tomorrow' (if I've
>     translated correctly) and assign it a different probability value
>     than that which I'm assigning to 'it will rain tomorrow'.  But I
>     think that anyone who observed my doing this would point out that
>     it's a contradiction, i.e., that since these things have the same
>     meaning, I'm obligated to assign them the same probability value.
>      In other words, it is in virtue of their representing the same
>     proposition that I'm obligated to assign them the same probability
>     value.    So, I think this only helps to underscore the fact that
>     when we explore our intuitions, we believe that propositions are
>     real p.v. holders.
>     3)
>     Anne, you said at one point that "not all beliefs can
>     be represented as numerical values" and that it "glosses over
>     inconsistencies - beliefs may be logically incompatible".  I was
>     intrigued by the suggestion, can you say more?
>     Again, thanks all for a stimulating telecon on Wednesday.
>     in advance if this address is not the right forum for these
>     discussions.
>     Best regards,
>     Mike Pool
>     --------------------------------
> Ken Laskey
> MITRE Corporation, M/S H305 phone: 703-983-7934
> 7151 Colshire Drive fax: 703-983-1379
> McLean VA 22102-7508
Received on Monday, 16 July 2007 10:54:48 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 19:50:54 UTC