I think this exchange with an expert of languages on-line can be of interest to us. Discussion comes from the Unicode mailing list.
Best
Glenda.
At 16:29 25/05/2012, Philippe Verdy wrote:
2012/5/25 glenda.jerome <info@execlub.org>:
> At 00:15 24/05/2012, Philippe Verdy wrote:
>>
>>
>> http://blog.wikimedia.org/2012/05/21/introducing-designs-for-the-universal-language-selector/
>> Enjoy :-)
>> -- Philippe.
>
>
> Philippe,
>
> could you please comment a little further on your comment? It would be of
> real help to people working on the Wikidata project. I personally have no
> idea why it would be a good/bad idea as I see cons and pros to the concept.
> The real issue we have is to associate a term, its display, its exact
> meaning and its value to a linguistic community. My fear is that a term may
> have orthoghraphic and meaning variances and that such a simplified approach
> may errode this diversity. This is not an area where we are expert, but I am
> aware that what may decided for a MediaWiki project may influence many
> things.
>
> Thank you!
> Glenda
I did not want to be obtrusive when posting this URL. But the current
project in the Wikimedia foundation certainly has many valuable
contributors available that could criticize the concepts and help
defining it more precisely.
But if the intent is to create associations of terms across languages,
the difficulty is several orders of magnitude higher than doing that
only within the same language.
For now there are manual tuning to help create interwiki links, but we
all know that these links are not (and cannot) be completely
transitive, include for the reason that articles on each wiki have
different structures and coverage of their subject (and it's irrealist
to have as many articles with exactly the same coverage across all
languages). So articles are splitting their subjects differently :some
wikis will group several subjects with similar names on the same page,
some will be so developed that they will be separated.
In my opinion, the only safe interlingual classification that works is
the one used in Commons (but Commons does not cover all subjects found
in other wikis, it is poor about history, linguistic, politics, music,
cinema, books and comics, and advanced sciences for example, because
it is difficult to illustrate many concepts distinctly with images,
audio or videos.
So instead, I would build a dedicated search engine that can search
through existing interwiki links and find associations, and possibly
develop some automatic translation tools, or text extractors that can
be used to compare contents from the various projects.