[SKOS] About links between SKOS-XL and SKOS reference (ISSUE-26)

Dear All,

Following the action:
> ACTION: Antoine will review Alistair's proposals w/r/t the 
> relationship between the existing solution and the extension [recorded 
> in http://www.w3.org/2008/03/18-swd-minutes.html#action06] 

First let me say that I do like the relation between XL [1] and the 
standard prefLabel/altLabel/hiddenLabel, including the semantics (they 
are actually the ones that I wrote on the whiteboard at the last F2F, 
let's be coherent with myself ;-)

I'm much less fond of the articulation between between XL and 
Reference's Label Relations [2]

As it is now, the recommended pattern for linking two labels resources 
(instance of xl:Label) together is still to attach them to a 
LabelRelation object, as in the example
> ex:fooRelation rdf:type skos:LabelRelation;
>   xl:labelRelated ex:bar;
>   xl:labelRelated ex:baz.
> ex:bar rdf:type xl:Label;
>   xl:literalForm "bar"@en.
> ex:baz rdf:type xl:Label;
>   xl:literalForm "baz"@en.

This is clearly against the idea that was behind considering labels as 
true individuals! The nice aspect of it was to be able to link them 
directly, like in:

ex:FOA ex:acronym ex:FoodAndAgricultureOrganization .

Your current solution would again amount to'reifying' the statement that 
links the two labels, which is to a great extent what people are 
uncomfortable about. Of course your solution still technically allows 
for creating the direct link I would like. But in the current 
formulation the added value of label-as-resources is not obvious for the 
use cases we had.

And actually when reading carefully your proposal it seems that you 
would really like to have people using always xl:labelRelated, in order 
to produce the 'standard' instanciation of the reified LabelRelation via 
the inference rule:
> uuu xl:labelRelated xxx . xxx xl:literalForm lll. --->> uuu 
> skos:labelRelated lll .


My first proposal would be to introduce an explicit xl:labelProperty (I 
assume it's not desirable now to use LabelRelation as localname) that 
users can then specialize, e.g. into the ex:acronym in my example.

We should also maybe remove the current xl:labelRelated that enables 
linking a label to the reification of a label link, or at least make it 
less central to the XL proposal. And consequently remove the 
specification of its semantics

This latter solution would allow to uncouple XL and the standard 
proposal for LabelRelation, and therefore to make both of them 
'extensions' of SKOS, if we think this is ok (and I do think this is ok).


Note that in case we still want to keep them linked, I would prefer to 
'replace' the inference rule
>  uuu xl:labelRelated xxx . xxx xl:literalForm lll. --->>  uuu 
> skos:labelRelated lll .

by a series of rules fitting the following pattern:
ccc skos:prefLabel xxx . xxx xl:literalForm mmm. xxx xl:literalProperty 
yyy. yyy xl:literalForm nnn.
--->>
ccc skos:seeLabelRelated [a skos:LabelRelation; skos:labelRelated mmm; 
skos:labelRelated nnn.]

which can generate skos:LabelRelation-information even if the xl:Label 
objects are directly linked, that is, linked without the mediation of a 
skos:LabelRelation object.

Best,

Antoine

[1] 
http://www.w3.org/2006/07/SWD/wiki/SkosDesign/SKOS-XL?action=recall&rev=2
[2] http://www.w3.org/TR/2008/WD-skos-reference-20080125/#L2914

Received on Tuesday, 8 April 2008 09:53:39 UTC