[Wikipedia-l] Volapük Wikipedia is the 15th largest?

[Francis Tyers]

On Fri, 2007-10-12 at 00:21 +0200, User 32X wrote:
> Hello,
> 
> Sabine Cretella schrieb:
> > Francis Tyers schrieb:
> >> Machine translation has a useful, and under-used place to play in the
> >> development of smaller Wikipedias.
> >>
> > I completely agree to this statement :-)
> 
> You overestimate the power of machine translation.
> I've just translated http://en.wikipedia.org/wiki/Serbin
> with Babelfish into German, if I wouldn't already know
> what that article is about, I wouldn't have known it after
> reading the German "translation".

I'd like to make a point here:

1) Neither English nor German are smaller Wikipedias, in fact both are
quite large!

Machine translation between distant languages is often poor. Not to say
that SYSTRAN (as available through Babelfish or Google) is the best you
can get, but even with a local installation and customised
dictionaries/rules there is still a fairly low limit on quality.

On the other hand, between closely related languages, machine
translation is often very good, as I have demonstrated with a previous
post. You might not be aware, but machine translation between
Spanish-Catalan and Spanish-Galician is already in production use, in
the first case at El Periódico de Catalunya, and the second place at El
Voz de Galicia. Maybe you don't care about smaller languages, but if
that is the case, why bother to post?

If you look at the current statistics for page size and quality, the
Spanish language Wikipedia is ahead of these other two languages. With
machine translation this gap could be decreased (in some cases
significantly).

So to wind up, you've taken a microscopic test and extrapolated to an
untenable conclusion. If you choose to revisit the issue, I would
recommend you do it from the point where we set off, that of
under-resourced, smaller Wikipedias.

Yours,

Fran

> Best regards,
> 32X
> 
> _______________________________________________
> Wikipedia-l mailing list
> Wikipedia-l at lists.wikimedia.org
> http://lists.wikimedia.org/mailman/listinfo/wikipedia-l