RE: specification

Hi Philipp, hi all,

regarding the SKOS-XL mapping, I'm recapping here and adding a few things
following your last mail of 2013 and the things setup in
http://www.w3.org/community/ontolex/wiki/Linking_to_SKOS

1) The LexicalEntry was a nice solution, almost totally inferable with
current specifications/reasoners (apart from the end of the chain, but I'm
positive for the future :-D) and IMHO much close to the "intention" of the
object. 
However, I do agree with you that simply it is not acceptable to use any
heuristic, like having the canonical form of an ontolex:LexicalEntry become
the chosen label for a concept. In short: not a viable solution.

2) on the contrary, telling that all forms are label does not seem to me
equally viable. Reasons in the wiki page above, however, in short: no
support from inference (i.e. for adapting to SKOSXL, we have to add all the
necessary triples and just reuse the ontolex:form object as a skosxl:Label),
and I don't see a strong intensional binding between the two objects.

So far, IMHO it is better not to state any strong axiomatic relationship
between the two. This does not prevent people to make any resource both a
ontolex:Form and a skosxl:Label, as the two classes do not need to be
disjoint.

One note. I previously mentioned the fact that there was another potential
issue in mapping to SKOSXL: maybe SKOSXL were not such the "independent
labels" which were considered (see considerations about "being a label of
something" in Philipp email). It is in fact very common practice in SKOSXL
that, if two concepts C1 and C2 have morphologically identical labels, yet
they have to be two different URIs. All the organizations I dealt with,
follow this approach.  In this sense, though reified (mostly for purpose of
metadata, like date of creation, modification etc..), the labels continue to
be "the labels of something".
This obviously is another difference with respect to linking concepts to
same items in a lexicon. If I have a lexicon to reuse, hopefully I would
like to have concepts expressed by the same objects, be linked to them if
possible.
By looking on the SKOSXL specification, I found it clearly stated that
having the same literalForm does not imply being the same label. However,
this sentence does not imply that "reusing" labels is forbidden.

I thus asked on the SKOSXL mailing list, as the specification was not about
this aspect:
http://lists.w3.org/Archives/Public/public-esw-thes/2014Jan/0000.html

The answer is confirming the common practice of having own labels for
concepts (see last reply by Vladimir Alexiev, and I add that in many cases,
the labels do not have a URI at all). 
However, Antoine Isaac also confirmed that it is not forbidden by the
specification to follow the case of shared labels. This leaves room for the
kind of Ontolex/SKOS interaction we were envisioning.

See you all in half an hour on the call!

Armando






> -----Original Message-----
> From: Philipp Cimiano [mailto:cimiano@cit-ec.uni-bielefeld.de]
> Sent: Thursday, January 9, 2014 4:32 PM
> To: public-ontolex@w3.org
> Subject: specification
> 
> Dear all,
> 
>     I have started writing the official and final specification of the
lexicon
> ontology model, see here:
> http://www.w3.org/community/ontolex/wiki/Final_Model_Specification
> 
> Feedback and contributions are welcome.
> 
> I have decided to start with 5 modules and one chapter discussing
> recommendations for using ontolex with SKOS (not part of the model).
> 
> I propose we have a call tomorrow to discuss two things:
> 
> 1) Structure of the document: I propose to merge translation and variation
> into one module, we can discuss about a shortcut property translation as
an
> inference by a property chain
> 2) Metadata Module: I propose to define a simple vocabulary of 10
> properties or so to keep complexity low and there ease adaptation
> 3) Linking to SKOS: my proposal given the discussion is not to link to
SKOS at
> all but simply give a number of recommendations on how to use ontolex and
> SKOS togeteher
> 
> Access details are here:
> 
> http://www.w3.org/community/ontolex/wiki/Teleconference,_2013.10.01,_
> 15-16_pm_CET
> 
> 
> Regards,
> 
> Philipp.
> 
> --
> 
> Prof. Dr. Philipp Cimiano
> 
> Phone: +49 521 106 12249
> Fax: +49 521 106 12412
> Mail: cimiano@cit-ec.uni-bielefeld.de
> 
> Forschungsbau Intelligente Systeme (FBIIS) Raum 2.307 Universität
Bielefeld
> Inspiration 1
> 33619 Bielefeld

Received on Friday, 10 January 2014 13:36:16 UTC