Re: properties as nodes etc.

>[Danny Ayers
>>
>>  Presumably the edge (==arc) can have as an attribute the URI of a property
>>  (e.g. dc:creator), but that this wouldn't be unique. The edge itself would
>>  have an identity status akin to that of bNodes, only existing in relation
>to
>>  the nodes to which it is connected.
>
>I think you could actually do it this way -  A predicate is by definition a
>resource, and you could create an instance of a property class as another
>resource and use that for your predicate.  Why not?  Then that edge would
>have its own URI, and you could make statements about it.
>
>Of course, if you did that for each triple you could end up with an awful
>lot of resource URIs, but then you probably won't need to do it for most.
>
>>  Is there any good reason within RDF MT why the edge shouldn't have it's
>own
>>  (instance) URI?
>>

Yes. It would mean that instances of properties could have other 
properties independent of the properties of the properties 
themselves. Which would be nightmarishly complex, for no obvious gain 
in utility. Moreover, it would greatly complicate the graph syntax, 
again with no obvious utility. I have never see a plausible use case 
for this.

>I have assumed that it is because the MT is supposed to match the M&S so far
>as possible, and the M&S doesn't support it explicitly.  Anyone who actually
>knows want to amplify this?

Sure. The key distinction you have to bear in mind is between a 
property on the one hand, and a particular arc, on the other. Many 
arcs can be labelled by the same uriref, so one cannot identify any 
particular arc with the property itself. (Philosophers of a language 
have a word for this, "trope". So for example, Loves might be a 
property, but the particular Loves between Marc Anthony and Cleopatra 
was one trope of Loves. In RDF, properties have urirefs, but tropes 
don't. )

To talk about properties is easy in RDF: just use the uriref as a 
node label. Bear in mind that it is the uriref which singles out the 
particular 'thing' in all cases, so if you want to say something 
about a 'thing', use that uriref as a node label. If the 'thing' is a 
property, then putting it on an arc is a way of saying that something 
else has that property, but its the same property whether its on the 
node or the arc. RDF acts this way all the time, eg consider the 
closure rule for rdfs:Range:

ex:age  rdfs:Range  ex:number .
Jenny ex:age _:y .

-->

_:y rdf:type ex:number .

In the first triple, the property ex:age is on a node, in the others 
its on an arc. Same property in all three triples, though.

I really doubt if you want to be able to refer to a single arc. The 
kind of thing that the trope-theory people worry about are English 
sentences like "Marc Anthony's love of Cleopatra was responsible for 
his losing the battle", which seems to be referring to a particular 
instance of a property. To hack round this, introduce a subproperty 
(rather than an instance) to do the work, eg in this example we could 
say that there was some special property that caused the battle to be 
lost, and which was a subproperty of Loves and which holds between MA 
and Cleo, but not anybody else: their special lovin' property. This 
follows the English idiom used in things like 'The love we have is 
unique'. Then you can do it all in conventional RDF.

Pat Hayes


-- 
---------------------------------------------------------------------
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, 25 June 2002 12:10:05 UTC