[SKOS] notations, label and their range (was Re: some thoughts about the OWL WG comments)

>> I still don't get it: we say that skos:notation works with typed 
>> literal, as in [1]
>>> This property is used to assign a notation to a concept as a typed 
>>> literal [RDF-CONCEPTS 
>>> <http://www.w3.org/TR/skos-reference/#ref-RDF-CONCEPTS>].
>> But in fact for the most common case (a concept having one notation), 
>> skos:notation would be used with plain literals? I'm really not 
>> convinced by what we are going to propose here...
> Antoine
> Are you not convinced because we haven't stated it clearly enough? Or 
> not convinced by the notion that skos:notation might be used with a 
> plain literal? 

I am not convinced because:
First I am not aware this was ever stated, actually. To me until Guus' 
mail, skos:notation was to be used only with typed literals, and if 
people wanted to use plain literals they would use private use language 
sub-tags [BCP47] with skos:prefLabel. If I read [1] that's really the 
feeling I have. And I worded the SKOS Primer to promote this practice [2].

Second, even though I recognize the interest of having one property for 
all notations (plain or typed literal) I'm not much in favor of this. 
For implementors it might make things more difficult, to anticipate both 

> The suggestion is that we temper the original wording:
> [[
> This property is used to assign a notation to a concept as a typed 
> literal [RDF-CONCEPTS 
> <http://www.w3.org/TR/skos-reference/#ref-RDF-CONCEPTS>].
> ]]
> which states that typed literals are used for skos:notation (but note 
> that there are no semantic conditions, so this would just be 
> convention anyway).

I guess there was a typo in your new wording, which is the same as the 
old one. Even if I usually trust your arguments I won't buy such one ;-)

More seriously, I'm more and more intrigued by the use we make of these 
"by convention". In fact I have re-read the labelling section in the 
reference, and found [3]

> By convention, RDF plain literals are always used in the object 
> position of a triple, where the predicate is one of |skos:prefLabel|, 
> |skos:altLabel| or |skos:hiddenLabel|. However, there is nothing in 
> the RDF or OWL Full semantics which prevents the use of a URI or a 
> blank node to denote an RDF plain literal. 

What should implementors conform to? In theory, they should anticipate 
SKOS labels to come either with a "normal" plain literal or an OWL 
individual that claims to be one.
But I would understand very well that implementations comply only to the 
"conventional" SKOS. Cases where URIs and blank nodes would be used as 
object of labelling triples will (*and should*) be rare, and at first 
sight I'd say accommodating them is a pain.
In this specific case, therefore, unless RDF tools by deafault normalize 
the data by generating "true" RDF literals from blank node and 
individuals that they get we might be heading to trouble. And I guess 
this won't happen anyway, as such objects may not be attached to any 
literal string.
So we remain with two forms of possible "conformance", the real one and 
the "conventional" one, in which let's say 95% of users will be interested.
Until now I could leave with that. But there are 11 "by convention" in 
the SKOS reference, putting interoperability at risk in term of 
"conventional" practices.
I would not expect all to be followed by a majority of implementors as a 
default, of course. There are also some KOS-specific semantic 
constraints (on skos:broader and skos:narrower paths, for instance) that 
are introduced as conventions and that do not harm much 
interoperability. But the fundamental "representation" conventions may 
be more dangerous to introduce.

>> By the way cc Norman Gray, as this conflicts a bit with what I've 
>> previously written to him
> It would be very useful to have Norman's comments on this.

Well at the beginning he was using plain literals with notations, but 
with the Reference as it is now worded I really did not have any 
difficulties convincing him that he was wrong :-/


[1] http://www.w3.org/2006/07/SWD/SKOS/reference/20081001/#L2064
[2] http://www.w3.org/TR/skos-primer/#secnotations
[3] http://www.w3.org/2006/07/SWD/SKOS/reference/20081001/#L1581
>     Sean
> -- 
> Sean Bechhofer
> School of Computer Science
> University of Manchester
> sean.bechhofer@manchester.ac.uk
> http://www.cs.manchester.ac.uk/people/bechhofer

Received on Tuesday, 14 October 2008 09:22:30 UTC