Re: [RDF-Concepts:109] What is the expressive power of RDF?

From: Graham Klyne <GK@ninebynine.org>
Subject: [RDF-Concepts:109] What is the expressive power of RDF?
Date: Wed, 19 Feb 2003 16:05:36 +0000

> Peter,
> 
> With reference to your comment:
>    http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0154.html
> 
> I accept a need for editorial revision, and have recorded it with id 
> 109-ExpressivePower [**].  The purpose of the rest of this message is to 
> try to ensure that I properly understand your concerns.
> 
> [**] for my own tracking purposes:  when the scope of the issue is 
> established, I'll ask Brian to allocate a WG tracking ID.
> 
> I think the problem can be described thus:
> [[
> There is incorrect wording describing the expressive power of RDF.
> 
> A formal description would be:
> "The expressive power of RDF is equivalent to the binary
> existential-conjunctive subset of first order logic".
> 
> Any informal explanations should be consistent with this.
> ]]
> 
> 
> To further help me understand your concerns, can you clarify to me why you 
> regard the following representations are not legitimate answers to the 
> questions you raise:
> 
> [[
> 1. How can
>     takes(John,book,school)
>     be represented in RDF?
> 
>     <rdf:Description>
>         <rdf:type rdf:resource="ex:TakingEvent" />
>         <ex:taker rdf:resource="ex:John"/>
>         <ex:taken rdf:resource="ex:Book"/>
>         <ex:to    rdf:resource="ex:School"/>
>     </rdf:Description>

This is an encoding of a trinary relationship as several binary
relationships.  If you wish to admit such encodings, then I think that you
should be much more formal about ``expressive power''.  It may be that
under some suitable definition of ``expressive power'' RDF can express
n-ary relationships.  However, under other definitions of ``expressive
power'' the above encoding is not admissable.  For example, the above
encoding allows for TakingEvents that do not have a taker, but the trinary
takes predicate does not admit this possibility.

> 2. How can
>     loves(John,spouse(John))
>     be represented in RDF?
> 
>     <rdf:Description about="ex:John">
>        <ex:loves rdf:parseType="resource">
>           <rdf:Description>
>              <ex:spouse rdf:resource="ex:John" />
>           </rdf:Description>
>        </ex:loves>
>     </rdf:Description>
> ]]

This is not even an encoding, as it is missing the functionality of spouse.


You haven't addressed the second half of this comment.

> #g

Peter F. Patel-Schneider
Bell Labs Research
Lucent Technologies



> At 10:26 AM 1/30/03 -0500, Peter F. Patel-Schneider wrote:
> 
> 
> >RDF Concepts states
> >         The expressive power of RDF corresponds to the
> >         existential-conjunctive (EC) subset of first order logic [Sowa].
> >
> >How can
> >         takes(John,book,school)
> >be represented in RDF?
> >
> >How can
> >         loves(John,spouse(John))
> >be represented in RDF?
> >
> >How can the RDF and RDFS semantic conditions be represented in the
> >existential-conjunctive subset of first order logic?
> 
> -------------------
> Graham Klyne
> <GK@NineByNine.org>

Received on Wednesday, 19 February 2003 13:43:47 UTC