RE: [SKOS] SKOS-XL & label relations

Hi Antoine,

Thanks for these comments. I've been trying to avoid getting into too much
detail at the moment, because I want to focus on the main questions of scope
-- do we move all support for label relations into XL? Do we publish XL as
part of the recommendation? 

Having said that, here's a few remarks on the finer points...


> Section 3: identity of xl:Label. I think I miss the sentence that was
> in
> the previous draft [2] to explain that different instances of xl:Label
> can have a same literalForm
> > Two different lexical label resources may share the same literal
> form.
> > In other words, if two lexical label resources share the same literal
> > form, they are not necessarily the same lexical label resource. The
> > data model has been defined in this way, to enable lexical label
> > resources to be part of concept schemes or other aggregations, and
> for
> > their membership in such aggregations to be part of their identity.

Yes, something like this would be good.

> Section 3.4.2:
> > @@TODO raise a SWD issue to consider relaxing domain of
> skos:inScheme?
> 
> +1, for making possible the story on XL literals as well as other
> useful
> situations (e.g. for SKOS extensions with newly defined entities are
> not
> skos:Concept but would require to be attached to a specific KOS)

Actually, when I was doing the checking on Sean's schema, I realised that
the SKOS Reference does not assert any domain for skos:inScheme at all. I
can't remember if this was a mistake or a deliberate choice, but I think
it's the right thing to do. So in fact this is not an issue.
 
> Section 8:
> I wonder whether seeRelation is too generic: seeLabelRelation, as was
> proposed before, is huglier, but less prone to misunderstandings...

I know, I made the name more generic because you can use it to point to
either xl:NaryLiteralRelation or xl:NaryLabelRelation. This needs more
discussion.

> Note that in the current version there will still be the problems about
> the use of seeRelation, as Tom pointed in [3].
> I am especially sceptical about the refusal to create any integrity
> condition
> > This means that there does not necessarily have to be any
> > correspondence whatsoever between the SKOS lexical labels of a
> > resource, and the labels involved in an associated XL literal or
> label
> > relation. The example below, which is consistent with the SKOS+XL
> data
> > model, illustrates this.
> 
> Do you really think we cannot at least constrain one of the labels to
> belong to the concept to which the relation is linked? That does not
> appear very useful from an inference perspective but may make the story
> a bit easier to swallow...

Unfortunately, the open-world assumption makes things complicated here. The
problem is, if you make some formal constraint on the data model such that
there has to be a correspondence between labels on a concept and labels in a
relation to which the concept is linked, you can't then do any consistency
checking, all you can do are draw some rather vague and useless inferences.

For example, let's say we have the formal constraint that, if C
xl:seeRelation R and R xl:literalRelated L, then L must be a label of C, the
problem is that L could be either preferred, alternate or hidden.

In other words, given the following graph...

<C> xl:seeRelation [ xl:literalRelated "foo" ] .

... the constraint does not make this graph inconsistent, nor does it
support the inference of any concrete triples. All we can infer is either
<C> skos:prefLabel "foo", or <C> skos:altLabel "foo", or <C>
skos:hiddenLabel "foo", but we can't say which.

So I think the only way to handle this is via usage conventions and best
practices, I don't think there's any value in formal constraints on the data
model.  

Cheers,

Alistair.

> [1] http://www.w3.org/2006/07/SWD/SKOS/xl/20080414
> [2] http://www.w3.org/2006/07/SWD/wiki/SkosDesign/SKOS-XL
> [3] http://lists.w3.org/Archives/Public/public-swd-wg/2008Mar/0083.html

--
Alistair Miles
Senior Computing Officer
Image Bioinformatics Research Group
Department of Zoology
The Tinbergen Building
University of Oxford
South Parks Road
Oxford
OX1 3PS
United Kingdom
Web: http://purl.org/net/aliman
Email: alistair.miles@zoo.ox.ac.uk
Tel: +44 (0)1865 281993

Received on Friday, 2 May 2008 10:05:22 UTC