Ziko, again, we are not talking about machine translations; Google
doesn't have machine translation for Bangla, Malayalam, Tamil etc.
yet. This is about translation memory.
One of the things about MAT, whose use in the professional translator
community is still debated but most popular for translations of
time-dependent things like news, is that the original is often a very
rough translation that requires a _lot_ of editing. The biggest
problem is not the toolkit itself (with some exceptions - punctuation
and templates, for example) but the translators who do not bother to
use it properly, creating poor translations with lots of spelling
mistakes and leaving behind a wasteland of poor quality articles.
GTTK can be used as a force of good if someone puts in the appropriate
time and effort; when used _properly_ by a careful, knowledgeable
translator who gives ample time for proofreading, articles created
with it should be virtually indistinguishable from any other article.
It is my thought that the huge problem here is lack of engagement with
communities. Essentially, Google swooped down and started dropping
large amounts of poor quality content on our projects without engaging
the people from those communities. The people in Google's contest also
didn't engage the communities, nor did they respond to requests to
improve their content.
-m.
On Wed, Jul 28, 2010 at 7:18 AM, Ziko van Dijk <zvandijk(a)googlemail.com> wrote:
2010/7/28 Nathan <nawrich(a)gmail.com>om>:
Just to be sure I understand...
It's good that you ask, indeed. :-)
No, it's not about free software, and the Wikimedians are not too
snobby or lazy to correct poor language. That is what I frequently do
in de.WP and eo.WP, and I suppose Ragib and many others as well. The
point is: The machine translated articles are often so bad that I
simply don't understand them. I *cannot* correct them, because I don't
know what they are saying.
Kind regards
Ziko
What's happening here is that human
beings, using a software tool, are translating
articles from the
English Wikipedia into a variety of other languages and posting them
on the comparatively small Wikipedia projects in these languages. The
articles, of unknown intrinsic quality, are usually mid to low quality
translations.
In the projects with an active community, some have rejected these
articles because they are not high quality and because the community
refuses to be responsible for fixing punctuation and other errors made
by editors who are not members of the community. In the projects
without an active community, Wikimedians (who may not speak any of the
languages affected by the Google initiative) are objecting for a
variety of other reasons - because the software used to assist
translation isn't free, because the effort is managed by a commercial
organization or because the endeavor wasn't cleared with the Wikimedia
community first. Some are also concerned that these new articles will
somehow deter new editors from becoming involved, despite clear
evidence that a larger base of content attracts more readers, and more
readers plus imperfect content leads to more editors.
What I find interesting is that few seem to be interested in keeping
or improving the translated articles; Google's attempt to provide
content in under-served languages is actually offending Wikimedians,
despite our ostensible commitment to the same goal. Concerns like
bureaucratic pre-approval, using free software, etc. are somehow more
important than reaching more people with more content. It all seems
strange and un-Wikimedian like to me. Obviously there are things
Google should have done differently. Maybe working with them to
improve their process should be the focus here?
_______________________________________________
foundation-l mailing list
foundation-l(a)lists.wikimedia.org
Unsubscribe:
https://lists.wikimedia.org/mailman/listinfo/foundation-l
--
Ziko van Dijk
Niederlande
_______________________________________________
foundation-l mailing list
foundation-l(a)lists.wikimedia.org
Unsubscribe:
https://lists.wikimedia.org/mailman/listinfo/foundation-l