W3C home > Mailing lists > Public > public-esw-thes@w3.org > May 2008

RE: multilingual labels in SKOS

From: Johan De Smedt <johan.de-smedt@tenforce.com>
Date: Sun, 1 Jun 2008 01:20:41 +0200
To: "'Peter Ansell'" <ansell.peter@gmail.com>
Cc: <public-esw-thes@w3.org>, "'Alistair Miles'" <alistair.miles@zoo.ox.ac.uk>
Message-ID: <001501c8c374$f1a51650$6402a8c0@TFJOHAN>

Peter, 

Typically, in each language there is one preferred term (hence my use of
prefLabel) per language for each concept.

This does not hold for non-PT (altLabel such as synonyms or near-synonyms),
a non-PT is language specific.
For related non-PT I intend using extensions of the skos:LabelRelation - it
is a laborious construction but it will work.

Thanks for the advice:
- making prefLabel-LANGUAGE a required and functional altLabel extension
(for each language) per concept should do it.
(I still will need some time to think this over and test it)

kr, Johan De Smedt. 
===================
-----Original Message-----
From: Peter Ansell [mailto:ansell.peter@gmail.com] 
Sent: Sunday, 01 June, 2008 00:50
To: Johan De Smedt
Cc: public-esw-thes@w3.org; Alistair Miles
Subject: Re: multilingual labels in SKOS

Strictly it would require non-standard code to validate, but you would
make up a temporary vocabulary which derived off of altLabel and
therefore be recognised by validators as part of the skos vocabulary.
Is there any need to denote prefLabel in each language within the one
concept?

It would be interesting if there were ways to add levels of
restriction for internationalisation to OWL natively. The only issue
currently is that the cardinality is restricted to a single statement
without regard for language... If you could have this cardinality
spread over the languages it would help.

For the immediate future you could quite easily just use altLabel with
a language tag to specify which language each altLabel is in. Do you
really need to restrict your system to only one in each language for
your system to work at all?

Peter

2008/6/1 Johan De Smedt <johan.de-smedt@tenforce.com>:
> Peter,
>
> I may conclude, I use indicate, the SKOS artifacts are not sufficient for
my
> use case.
> The post I made is to validate this.
>
> I do thing though it is also and RDFS/OWL issue.
> As indicated below, the OWL semantics easily allow to break any language
> specific modeling.
> So specifying properties like prefLabel-LANGUAGE always will need
> proprietary code to validate.
>
> It may be interesting for SKOS to promote this use case to OWL 2.
>
> kr, Johan De Smedt.
> ===================
> -----Original Message-----
> From: Peter Ansell [mailto:ansell.peter@gmail.com]
> Sent: Sunday, 01 June, 2008 00:07
> To: Johan De Smedt
> Cc: public-esw-thes@w3.org; Alistair Miles
> Subject: Re: multilingual labels in SKOS
>
> You are likely to have to create a prefLabel-LANGUAGE for each of your
> desired languages to get around it, trouble is though, that you can't
> get around the prefLabel cardinality if you derive these from it, so
> you would have to find another way. OWL doesn't care about literals
> really, it is more focused on ensuring that subjects and predicates
> are used "accurately", where accurately is decided by the ontology
> designer. Is there any real advantage to using prefLabel instead of a
> whole lot of alternates in your case anyway?
>
> Peter
>
> 2008/6/1 Johan De Smedt <johan.de-smedt@tenforce.com>:
>>
>> Dear,
>>
>> Neither of RDF-Schema or OWL seem to handle functional properties for
>> multilingual values very well.
>> Typically these could be used for skos:prefLabel.
>> (A prefLabel is desired to have cardinality 1,
>>  but extensions to SKOS to realize this do not work out in a multilingual
>> KOS)
>>
>> - Is this use case handled by SKOS ?
>> - is there an advised solution pattern ?
>>  e.g. making language specific sub-properties of skos:prefLabel that do
>> have cardinality 1
>>       (I do not like this because nothing prevents me making an
>> xx:prefLabel-en getting the value "dit is niet in de juiste
>> taal"@nl^^xsd:string)
>> - do we need to turn to inference rules to solve this ?
>>
>> thanks for advice on handling this use-case.
>>
>> Kind Regards,
>>   Johan De Smedt
>> =================
>> johan.de-smedt@tenforce.com
>> mobile: +32 477 475 934
>> =================
>>
>>
>>
>
>
Received on Saturday, 31 May 2008 23:32:27 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:38:59 GMT