Re: lightweight reification

At 04:14 PM 5/30/00 +0200, Pierre-Antoine CHAMPIN wrote:
>Graham Klyne wrote:
> > >I must appologise for the diagram not being clear.  The intent was to show
> > >the whole statement being the subject of the 'asserted by' property which
> > >would, I think, be consistent with the RDF model.
> >
> > Ah, OK.  Consistent if the statement is reified.
>
>Why should we make a difference between reifying a triple and reifying an arc,
>that is :
>
>  ___________
>| S -(P)-> O|         S -(P)-> O
>|___________|             |
>       |              (assertedBy)
>  (assertedBy)             |
>       |                   v
>       v                Someone
>    Someone
>
>    (fig 1)              (fig 2)
>
>for me, both representation have the same meaning,
>even though M&S uses fig.1, I find fig.2 more readable
>
>NB : in fig.2 I'm not annotating property P ! I'm annotating this one arc
>(the one labeled with P, going from S to O... looks like a triple, doesn'it ?)

But I don't know how to represent fig 2 within the current RDF graph model 
... (P) is an arc, not a node, and the model supports only arcs that 
originate in node resources.

Fig 2 can be treated as a convenient shorthand for S-(P)->O, PLUS its 
reification, PLUS the arc 'assertedBy' from the root resource of the 
reification to 'Someone'.

I happen to quite like the direct arc-from-property approach, but it's not 
RDF as we know it.  The recent posting about:

    <property rdf:ID="id">...

suggests a way of representing this, which would map into a formal model 
consisting of quadruples -- almost the same as RDFM&S, section 5, except 
that the set called 'Statements' consists of quadruples of the form:

   {pred,subj,obj,ID}

where ID is a name for the statement that can be used to construct the 
subject (or object) of some other statement.  I see the approach of using 
hashes to name statements as an attempt to achieve the same effect within 
the current (triple-based) formal model.

#g


------------
Graham Klyne
(GK@ACM.ORG)

Received on Tuesday, 30 May 2000 14:25:30 UTC