[WMCEE-l] Content Translation tool for CEE Spring?

Michal Matúšov kubof at i-espero.info
Sat Jan 31 16:33:52 CET 2015


Hi Mykola and all

(I am sorry for very long response, it is full of technical affairs and I
hope I gave some light into the presented issues)

The Content Translation extension basically provides 2 tools for
contributors:

   - language independent: the translation interface with 3 columns for
   original article, translated article and some tools (like preview of linked
   articles); this interface automatically map links, categories and files
   - language dependent: additional tools like machine translator,
   translation memory, dictionary etc. - need additional settings for ALL
   language pair

Content Translation can be set for "source language without machine
translation". That's the case for Danish (from English and Swedish) now.

As I understand, for both tools, language independent and dependent, is
during translations created translation memory for the particular lang pair
(similar for Translate extension used for translating particular page, you
know the "Translate this page" link on top of pages). This memory can be
developed without machine translation and be used for next translations,
sometimes even so great, that whole sentences are translated from there.

Machine translator can be added for particular language pairs, if there
exist some good. This is checked in advance by Language Team by standard
form http://hexm.de/8iyt

If there is machine translator enabled, the Content Translate check the
amought of text which wasn't edited. Contributions made by Content
Translation are marked by tag "contenttranslation" and so can be tracked. I
suppose, that new tag for non-proofreaded content can be added by
developers.

Quality of machine translation vary significantly between language pairs
(even between different directions of the same language pair!). If there is
no good translator, we should omit it (or help to create it ;). For some
languages, like Slovak-Czech it is very great!

Now Content Translation publish in default into user namespace, this can be
changed (or leaved, if the local community prefer).


*But there is another issue to talk about.* Usually it is required that the
local community make request for the tool, not make request for all wikis
in whole region. See more about the requirements
https://www.mediawiki.org/wiki/Content_translation/Languages

Personally I consider this tool very useful (after fixing some newborn
ills). For informing users about this tool, to help evaluate and make
feedback for its developers (and hopefully to help with improving the
machine translator Apertium) we (WUG Esperanto kaj Libera Scio) plan to
make contest on Esperanto Wikipedia (thanks for the related session in
Kyjiv, it was useful for this!).

Regards

2015-01-31 15:42 GMT+01:00 Mykola Kozlenko <mycola-k na ukr.net>:

> Hi,
>
> I have some concerns about "tool works good", as machine translation tools
> will never be perfect.
>
> We have quite a lot of experience with article contests in Ukraine, and
> machine translations are usually the worst things you can find. The issue
> is that machine translations from close languages (e.g. Russian or
> Belorussian to Ukrainian) cannot be spotted immediately as structure of
> sentences is usually correct. However, when you start reading, you will
> often find some crazy mistakes, mostly completely inappropriate words (e.g.
> an article about biology translated using military terminology). In many
> cases we end up getting negative feedback from readers who say that some of
> the articles are unreadable due to machine translation mistakes and that
> they prefer reading original articles instead (say, Russian or Belorussian
> in our case). Thus in previous contests machine translated articles were
> disquali fied unless they were proofread by the end of the contest.
>
> While I agree that good translation tool is helpful, we need to get some
> protection from users who will do a machine translation without
> proofreading. Is there a way, for example, to prevent a user from saving
> the article in main namespace before he has proofread the entire article?
> Or probably can we tag translations which were not proofread by humans?
> Otherwise we are afraid of being flooded by badly translated articles.
>
> Best regards,
> Mykola aka NickK
> WMUA
>
> --- Оригінальне повідомлення ---
> Від кого: "Zana Strkovska" <777.zana na gmail.com>
> Дата: 30 січня 2015, 18:11:22
>
> If this tool works good, it will be very useful
>
> Regarads,
> Zana
>
> 2015-01-30 1:34 GMT+01:00 Anna Koval <akoval na wikimedia.org>:
>
> Hello CEE. :)
>
>
> Happy news to share. :)
>
>
> I had a very productive talk with WMF's Language Engineering team today.
>
>
> *Learned more about their content translation tool.
>
>
> *Told them about CEE Spring.
>
>
> *Added a request to Phabricator to add the tool to CEE languages as a beta
> feature.
>
>
> https://phabricator.wikimedia.org/T88030
>
>
> If you are on Phabricator, please add yourself to the task. If you are not
> on Phabricator, you can create an account by signing in with your Wikipedia
> credentials.
>
>
> The language team is very excited about the potential of their tool on our
> project and they will be looking for testers.
>
>
> You can read more about the content translation tool on the Wikimedia blog
> and watch a video demonstration of it there too.& nbsp;
>
>
> https://blog.wikimedia.org/2015/01/20/try-content-translation/
>
>
> :)
>
>
> --
>
>
> Anna Koval, M.Ed.
>
> Manager, Wikipedia Education Program
>
> Wikimedia Foundation
>
> +1.415.839.6885 x6729
>
> akoval na wikimedia.org
>
> education.wikimedia.org
>
> _______________________________________________
> WMCEE-l mailing list
> WMCEE-l na tools.wikimedia.pl
> http://tools.wikimedia.pl/mailman/listinfo/wmcee-l
>
>
>  _______________________________________________
> WMCEE-l mailing listWMCEE-l na tools.wikimedia.plhttp://tools.wikimedia.pl/mailman/listinfo/wmcee-l
>
>
> _______________________________________________
> WMCEE-l mailing list
> WMCEE-l na tools.wikimedia.pl
> http://tools.wikimedia.pl/mailman/listinfo/wmcee-l
>
>


-- 
KuboF Hromoslav (Michal Matúšov)
------------- ��al��ia ��as�� ---------------
HTML pr��loha bola vy��at��...
URL: <http://tools.wikimedia.pl/pipermail/wmcee-l/attachments/20150131/23674b73/attachment-0001.html>


More information about the WMCEE-l mailing list