Semi-mechanizing the DTTP translations

[Pierre Slamich]

Yes, although we might be finished by then ;-)
Thanks to the method we're reviewing and correcting around 1000 strings per
day at the moment.

sincerely,
Pierre

On Tue, Dec 18, 2012 at 4:06 PM, Hannie Dumoleyn <
lafeber-dumoleyn2 at zonnet.nl> wrote:

>  Hi Pierre, Redmar, and all who are interested,
> Would it be an idea to brainstorm on this in #ubuntu-translators? Perhaps
> in January 2013?
> I agree with Redmar that the msgmerge is a good method, especially for
> huge documents. The only snag is that you still have to approve the fuzzies
> offline before uploading the file back to Launchpad. We use this method for
> the Ubuntu Manual "Getting started with Ubuntu" (Lucid > Maverick > ....>
> Raring) and with success.
> Redmar, sorry for not yet having tested your popsort :(
> Regards,
> Hannie
>
> Op 18-12-12 00:51, Pierre Slamich schreef:
>
> Hi Hannie, Hi Redmar,
> Thanks a lot for the tips: we're interested in using your approach, and
> more generally it might be interesting expending the msmerge approach to
> all teams that are already underway for the DDTP, and the Google one to the
> teams that need to get started.
>
>  - For the Google Translator Kit approach, I guess we could extend the
> mock project we did for fr_FR to other languages (and streamlining our
> process by using Bazaar) by creating a global team responsible for the DDTP
> Mock project and including in this team one member from each language team
> responsible for uploading the machine translated po for his or her language.
>
>  - For the msmerge approach, do you already have a project to handle this
> ? Is there any advantage in msmerging raring against releases older than
> quantal to get more modified strings ? How many strings have you been able
> to recover using that approach ?  It might be neat to generate the msmerged
> po for all languages ? Importing them as actual translations (not fuzzy)
> into a mock project like the Google Translate one would show them as
> suggestions for the actual DDTP as well.
> The translator would thus be able to pick the human translated one when
> available or to build on the machine translated one otherwise.
>
>  Can we try to schedule some time to coordinate on this so that we can
> use both approaches and try to onboard all the other languages teams once
> we have a rock-solid process ?
>
>  Pierre
>
> Pierre Slamich
> pierre.slamich at gmail.com
>
>
> On Mon, Dec 17, 2012 at 10:30 PM, Redmar <redmar at ubuntu-nl.org> wrote:
>
>> Hi Pierre,
>>
>> I've actually tried a similar approach for Dutch using msgmerge, which
>> might also be worth checking out. When you merge the translations of an
>> older version of ubuntu into the current version (msgmerge
>> quantal_ddtp.po raring_ddtp.po -o merged_ddtp.po, for example), there
>> will be a lot of 'fuzzy' translations for strings that are similar (for
>> example, meta packages for different programs, debugging symbols etc).
>> These fuzzy often only need a few small changes (eg program name) to be
>> accepted, which can really speed up translations. And you don't have to
>> worry about google putting in a weird translation, since it is all based
>> on earlier translations done by a human.
>>
>> On a related note, if any of you work on ddtp-translations offline, I
>> have written a python program that can sort entries in ddtp po-files
>> based on the popularity of the package. This way, the most popular
>> packages will be at the top of the po file, and you are always sure you
>> are working on the most important packages first.
>>
>> You can get the code here:
>> bzr branch lp:~redmar/+junk/ddtp_popsort
>>
>> It has a small readme file, please let me know if something is unclear
>> or not working for you.
>>
>> Regards,
>> Redmar
>> --
>> Ubuntu Dutch Translators
>>
>>
>> Hannie Dumoleyn schreef op ma 17-12-2012 om 17:58 [+0100]:
>>  > Hello Pierre,
>> > This is a very good idea! I have just uploaded the first part of the
>> > incomplete Dutch translation (900kb) to GTT.
>> > Thanks,
>> > Hannie
>> >
>> > Op 17-12-12 12:55, Pierre Slamich schreef:
>> >
>> > > The DDTP represent around 50 000 strings to translate * 140
>> > > languages. On very good weeks, a typical translation team translates
>> > > 500 strings (see UWN for examples weekly figures).
>> > >
>> > >
>> > > Would take a lot of weeks (years?) with highly motivated volunteers
>> > > of a large translation team, working non-stop, at their best to get
>> > > done with it.
>> > > Thus we had the idea to delegate initial translation suggestions to
>> > > Google Translator Kit and review translations with humans to speed
>> > > the process.
>> > >
>> > > We successfully did an import for circa 40 000 French strings  (yup
>> > > you read that right) this week-end in a mock project called DDTP
>> > > Automation (https://translations.launchpad.net/ddtpautomation).
>> > > To keep it short, the translations from this project appear as
>> > > suggestions in the French DDTP, and can be reviewed by actual
>> > > translators.
>> > > We've started using them, and it turns out that a lot of them are
>> > > actually useful and are speeding up the translation process a lot.
>> > >
>> > > We detailed the (somewhat) tedious process in English at
>> > > http://lite.framapad.org/p/ddtpUbuntu
>> > > Questions and inquiries welcome.
>> > >
>> > > Pierre
>> > >
>> > >
>> > > ---
>> > > pierre.slamich at gmail.com
>> > >
>> > >
>> >
>>
>>
>> --
>> ubuntu-translators mailing list
>> ubuntu-translators at lists.ubuntu.com
>> https://lists.ubuntu.com/mailman/listinfo/ubuntu-translators
>>
>>
>
>
-------------- section suivante --------------
Une pièce jointe HTML a été nettoyée...
URL: <https://lists.ubuntu.com/archives/ubuntu-fr-l10n/attachments/20121219/ecb81689/attachment.html>