Re: The RDF Approach to Indicating Language-In-Use

On Thursday, October 30, 2003, at 11:10  AM, pat hayes wrote:

>> From: Sandro Hawke <sandro@w3.org>
>> Subject: Re: The RDF Approach to Indicating Language-In-Use
>> Date: Wed, 29 Oct 2003 16:44:58 -0500
>>
>>>
>>>
>>>  > The *only* aspects of all of this that fall into the purview of 
>>> the
>>>  > Semantic Web are an importing mechanism and the translation from 
>>> a name to
>>>  > a namespace address.
>>>
>>>  Is owl:imports satisfactory as an importing mechanism?   (That is,
>>>  does OWL going to REC get you part 1?)
>>
>> As far as I am concerned, owl:imports is sufficient.  However, OWL 
>> going to
>> REC doesn't solve everyone's problems.  In particular, RDF is left 
>> without
>> an importing mechanism.
>
> I fail to see how an importing mechanism deals with the central issues 
> here. The main issue, to me, is what one should be able to do with a 
> URIref occurring inside RDF/RDFS/OWL/whatever in a document.

I don't see that it's what one should be *able* to do. I mean, you can 
pretty much do anything you like :)

>  Should there be a presumption that the Web can(/may/oughtto) be used 
> to retrieve some kind of information which might be useful to an agent 
> (human or software) in drawing conclusions concerning that URIref?

Ok, we've moved to what the *presumption* should be. But let's look at 
the three variants:

	*Can* the web be used to retrieve some kind of info which might be 
useful....

Well, as long as there *is* possibly useful retrievable info, then of 
course.

	*May* the web be used...

This depends on context. It's *not* permitted to look at an rdf graph, 
*then* add some extra web retrieved info, *then* draw some conclusions 
not entailed by the original graph, *and* claim that those conclusions 
are RDF entailed (or simple entailed), by the original graph. (You, of 
course, well know this :))

	*Should* the web...

I think the answer to this is a resounding "no". That is, I think bare 
RDF (or OWL) entailment is useful. I think adding more explicit (and 
more, explicit) import controls would be useful.

>  If not, URIrefs in OWL (other than in owl:imports) are just logical 
> constants, so why the hell are we obliged to use URI syntax in these 
> languages?

Well, it's useful for node merging when you *do* explicitly graph 
concatenation. It makes accidental name clashes rather less likely. 
That seems useful. There's interacting with other software.

I think the first is actually a pretty big deal. You can use the URI as 
a key to *ALL* sorts of documents (and to things within documents).

> If so, what protocols/assumptions are to be expected or invented to 
> support the nature of these sources and how to retrieve and use them? 
> I personally don't find the former position (importing does it all, 
> URIrefs outside imports are meaningless)

They clearly aren't meaningless, on any proposed position.

>  acceptable: it reduces the SWeb to conventional ontologies which 
> happen to be on the Web, which may well be useful but isnt the vision 
> of the SW that gets me excited.

What's that vision? I mean, I hear the "bootstrap vision" but I just 
don't even vaguely believe it. I sorta am interested in widespread, 
fairly interoperatable and shared KR. I suppose that's a bit boring.

> On the other hand, since our primary task is to produce some words, I 
> think that it is important not to say anything which would be 
> *inconsistent* with the conventional-ontologies-on-the-Web view, since 
> that is where the immediate industrial applications are.

And is a reasonable default, I think.

Applications will want to do more and other, but it's not at all clear 
to me what they *ought* to want to do.

Cheers,
Bijan Parsia.

Received on Thursday, 30 October 2003 12:44:02 UTC