Re: Southampton Pub data as linked open data

On Jul 30, 2008, at 3:07 AM, Richard Cyganiak wrote:

> Bijan,
>
> Wow, seems like I really pushed the wrong buttons!

No worries. I'll reset.

> My intention *was* to be controversial, but not *that*  
> controversial. I didn't mean this as a personal attack on you. I  
> meant it as an attack against a specific technical decision made in  
> the design of OWL.

That's fine, but then let's get into the actual technical detail.

> It is evident that I crossed a line here and for that I apologize,  
> and I accept the blame for the temporary deterioration of the thread.

And you should take the credit for elevating it back up. Good show.

> I think we agree on two facts:
>
> 1. Creating subproperties of rdfs:label puts ontologies into OWL  
> DL, which is undesirable for OWL reasoners.
>
> 2. Creating subproperties of rdfs:label enables RDF consumers like  
> Tabulator to do some useful presentational things.
>
> So there's a real conflict.

Yes.

> We don't want to force ontology authors to pick sides, so it would  
> be good to resolve it somehow.

Agreed.

> There are two proposals:

Ok.

> a) Accept that creating subproperties of rdfs:label was a bad idea,  
> and change the RDF browsers and existing ontologies to use either a  
> different property or a different mechanism,

I'd prefer a different mechanism for more reasons than the conflict,  
fwiw.

> b) Accept that subproperties of rdfs:label are a fact of life, and  
> change the OWL reasoners to automatically clean up those parts that  
> are outside of OWL DL.
>
> To me it is not self-evident that a) is the correct answer.

Nor to me!

> I would prefer b). I admittedly have not much clue about OWL, so  
> please forgive me if I display my ignorance once more, but wouldn't  
> it be a nice feature anyway if OWL reasoners were able to  
> automatically clean up ontologies to the desired level of Full, DL,  
> or Lite?

Pellet does a fair bit of this.  But can we move back yet another  
level first? What are the requirements? I.e., what is the exact  
functionality desired. If we can work that up clearly, then we can  
assess the technical solutions a bit better.

For example, the way that some clinical information systems use OWL  
is for form management. That is, they want to use OWL to manage very  
large collections of related forms. They way they do this, roughly,  
is to use OWL as a way of assembling relevant piece of information  
together and then use a "sanctioning" mechanism to describe how the  
information in the classes relates to the generated forms.

(For more details see:
	<http://rease.semanticweb.org/ubp/PUSH/search@srchDetailsLR?lrID=lr- 
lmua-stheidmann-1163408808058>.)

Clearly this needs a more sophisticated system that just rdfs:label.

> Would that be a workable alternative, instead of designing a new  
> mechanism for presentational annotations?

In my annotation space proposal, you can subproperty rdfs:label.  
Whether we succeed in getting it or something similar into OWL is a  
different question. Even then, there's still the question whether  
this is what we *should* do.

For you, what's the technical advantage of:
	name rdfs:subPropertyOf rdfs:label
over
	name displayFor AllAudiences.
(or something similar).

(Other than the very good reason that current systems support the  
former. I'm trying to understand the *design* rationale.)

> Any opinions appreciated.

Hope these help.

Cheers,
Bijan.

Received on Wednesday, 30 July 2008 02:32:47 UTC