Re: Problem with damlsParameter

On Friday, April 18, 2003, at 11:59  AM, Daniel Elenius wrote:

>> Well, we've had some discussion about this.  The designers of DAML+OIL 
>> have regarded this as a controversial and somewhat unsettled issue 
>> (that is, whether a property can be the range of another property).  
>> However, I think it's fair to say that, on a strict reading of the 
>> DAML+OIL spec, this usage is not legal.  So you are right.  However, 
>> we were not aware of it causing any problems, and consciously decided 
>> to leave it this way.
>
> I don't understand how ontologies where you allow properties as range 
> entities can be represented as graphs.

You reuse the label on a node and a property. The Graph is not the Thing.

> In the graph representation, the way I understood it, properties are 
> the (labelled) directed edges between the nodes (which can be classes 
> or primitive values). How can an edge point to another edge?

Well, if we're doing visual reasoning it, I find it hard to see how 
*not*. But such graph reasoning is, imho, to be avoided. If you check 
out the RDF Semantics doc, and the one on RDFS, you'll see that a 
property can be a class or an individual at the very same time.

> It doesn't make sense to me, and I can see why it doesn't make sense to 
> the JTP reasoner, it being very logical after all :)

No, it just means that you don't have the right axioms.

The case for DAML+OIL is tricky because, on some hands, it's a 
description logic, and they frown upon having Properties as individuals 
(though in the graph syntax for DAML+OIL, you'll see properties in 
subject and object  positions: consider subProperty relations). On other 
hands, it's supposed to build upon RDFS where that's not only kosher, 
it's heartily encouraged.

OWL does somewhat better in defining two major variants, OWL/DL and 
OWL/Full. The latter has the promiscuity of RDFS.

And none of this even touches on the fact that you can put an "opaque" 
URI anywhere you like. I.e., if you have http://someProperty in one file 
but don't *declare* it ot be a property, you "can" do whatever you like 
with it. This is part of the motivation for the RDFS approach.

[snipped the rest as I'm having trouble attending to it right now but 
think the rest is important]

Cheers,
Bijan Parsia.

Received on Monday, 21 April 2003 13:03:05 UTC