Re: Southampton Pub data as linked open data

Richard
>
> Bijan, Knud, Bernard, thanks for the clarification.
>
> I'm indeed surprised! Subclassing rdfs:label is okay in RDFS, and it 
> is okay in OWL Full, but it is not allowed in OWL DL.
The issue here is not that subclassing annotation properties is 
forbidden by OWL-DL, which is explainable and defendable. The real 
mistake of OWL, IMHO, is to have defined rdfs:label as a built-in 
annotation property, and not as a datatype property, and with no 
built-in alternative for dealing with names (beyond URIs). Defining 
rdfs:label as an annotation means that you can't build any OWL-DL 
inference whatsoever based on the value of rdfs:label, and that you 
can't define various types of "names", to which you would want to attach 
specific semantics, as subproperties of rdfs:label. For example 
skos:prefLabel and skos:altLabel are not defined as subproperties of 
rdfs:label, and if they were SKOS vocabulary would be OWL-Full instead 
of OWL-DL. But interfaces such as Tabulator need a value of rdfs:label 
for resource display, so in the description of a skos:Concept you have 
to repeat the value of skos:prefLabel in a rdfs:label field to make 
Tabulator happy. This is suboptimal at least.
>
> The RDF consumers I'm working on (RDF browsers and the Sindice engine) 
> don't care if you're in OWL DL or not, so I'm tempted to argue that it 
> doesn't matter much for RDF publishing on the Web. (IME, on the open 
> Web, trust and provenance are much larger issues than inference, and I 
> don't believe that the open Web will ever be OWL DL, so why bother.)
Well, the above example shows it's not a minor issue. DL-based inference 
does not necessarily mean involving arcane axioms and complex 
deductions. Even for Web publishing, interfaces willing to display 
"intelligently" the names (such as skos:prefLabel) of a resource need a 
minimal level of inference over the various type of names. The fact that 
those interfaces rely on rdfs:label for dispaly is indeed an issue. If 
OWL had a built-in owl:label datatype property, this would be much 
easier to deal with ... Is anything along thoses lines intended in OWL 2.0 ?

Bernard
>
> Others here will probably have different perspectives on this question.
Indeed :-)

Bernard
>
> Richard
>
>
> On 28 Jul 2008, at 17:01, Bijan Parsia wrote:
>
>> On 28 Jul 2008, at 16:23, Richard Cyganiak wrote:
>>
>>>
>>>
>>> On 28 Jul 2008, at 15:52, John Goodwin wrote:
>>>>> In an ideal
>>>>> world, John would declare pub:name a subproperty of
>>>>> rdfs:label, and the tools would infer the rdfs:label value...
>>>>> But most clients don't do that yet.
>>>>
>>>> Am I allowed to declare something as subproperty of rdfs:label?
>>>
>>> As far as I know, yes.
>>>
>>>> I'm
>>>> guessing this is one of those things that is allow in RDF, but not in
>>>> OWL DL?
>>>
>>> I would be surprised if that is the case.
>>
>> You're surprised.
>>
>>> What makes you think so?
>>
>> The spec? :) But also you can try one of the species validators.
>>
>> (rdfs:label is an annotation property and you are not allowed to 
>> subproperty annotation properties in OWL DL)
>>
>> http://www.w3.org/TR/owl-semantics/syntax.html#2.1
>> """Properties relate individuals to other information, and are 
>> divided into four disjoint groups, data-valued properties, 
>> individual-valued properties, annotation properties, and ontology 
>> properties"""
>>
>> Then if you look at the rest of the grammar, you'll see where 
>> annotation properties are allowed.
>>
>>> Can anyone else comment on this?
>>>
>>> (FWIW, foaf:name is a subproperty of rdfs:label.)
>>
>> And hence, not OWL DL.
>>
>> Historywise, this sort of annotation is a kind of metamodeling. At 
>> the time, the WebOnt working group (at least the DL contingent) 
>> wasn't sure how to handle this (it's not a standard feature of 
>> logics, esp. if you give it a strong semantic reading a la OWL Full). 
>> So the compromise was to forbid this.
>>
>> In OWL 2 (DL), you can get this sort of effect two ways, annotations 
>> (which are under discussion and being explored) or by punning classes 
>> and individuals (which won't actually help you with the built in 
>> vocabulary).
>>
>> Typically, subpropertying rdfs:label isn't really a *domain modeling* 
>> thing, but an attempt to spec a *presentational* issue (i.e., many 
>> UIs exploit rdfs:label, and one wants to indicate which properties 
>> should show up in the UI). Thus, there's a bit of tension there.
>>
>> HTH.
>>
>> Cheers,
>> Bijan.
>
>
>

-- 

*Bernard Vatant
*Knowledge Engineering
----------------------------------------------------
*Mondeca**
*3, cité Nollez 75018 Paris France
Web:    www.mondeca.com <http://www.mondeca.com>
----------------------------------------------------
Tel:       +33 (0) 971 488 459
Mail:     bernard.vatant@mondeca.com <mailto:bernard.vatant@mondeca.com>
Blog:    Leçons de Choses <http://mondeca.wordpress.com/>

Received on Tuesday, 29 July 2008 12:58:31 UTC