Page MenuHomePhabricator

Migrate translation projects at Meta to the Translate extension
Closed, ResolvedPublic

Description

Frequently large translation projects are carried out in Meta: Updates from the WMF, Fundraising banners and editor appeals, surveys etc.

Now the volunteers are asked to translate the 2011 Editor survey. It is a very long document with many repetitive strings. It requires many comments and explanations about wording, as the discussions on Translators-l demonstrate.

The translation is done using the plain MediaWiki interface of Meta. This has many issues, because MediaWiki was never focused on translation. The issues are many:

  • there's faulty RTL support for RTL languages.
  • there's no Translation memory to make working with repetitive strings efficient.
  • using the browser built-in spell-checking is hard in this environment.
  • there's no easy way to review and proofread translations and compare them to other languages.
  • there's no structured way for the original document authors to comment on the source strings.
  • in large translation projects the alignment of the original strings with the translated ones may get lost because of wrong order of line breaks.
  • There's no way to automatically and comprehensively check the translation status of every language. Then translator manually adds his language, manually marks it as "progress", "proofread" and "ready". A Meta admin must manually check the statuses and manually copy it to the target.
  • There's no automatic way to notify the translators that the source string has been updated.

The Translate extension for MediaWiki solves all these issues. It is successfully and extensively used in Translatewiki.net for translating MediaWiki and its extensions and several other projects.

Please start using the Translate extension for translation projects in Meta. This will make the translation much more efficient and save a lot of time for everyone. Thank you.


Version: unspecified
Severity: enhancement

Details

Reference
bz28021

Event Timeline

bzimport raised the priority of this task from to Unbreak Now!.Nov 21 2014, 11:31 PM
bzimport set Reference to bz28021.

ollinpostit wrote:

I agree with you. It´s very nice suggestion and I would like it also very much!

It's very nice to hear that there is interest for this extensions. I will certainly work towards this goal.

I would strongly support this feature as I translate whole texts one sentence by one sentence, so it would be logical to enable this feature. Besides, we can use machine translations from translating services like Google Translate and Microsoft Translator to help us in some forms of translations.

What are the issues (if any) that need to be addressed before this extension could be used on Meta?

The extension code has not been reviewed throughout. Some page translation features (for example page moving and page deleting) need testing and fixing.

Some of the procedures how we run the extension at twn are not suitable for maintenance free deployment of WMF. I'm thinking of things like the command line scrips and file based message group definition.

Is the extension mainly designed to work with i18n files or does it work fine with on-wiki MediaWiki messages as well? (The extension documentation doesn't seem to have any high-level overview).

I'm working to get this reviewed by Trevor and Roan ASAP. I've gotten the ok from the people who manage translations on Meta, too.

The ball slowly rolls.

Assigning this to Trevor for review.

In IRC today, I talked with Roan and Trevor about what would be reasonable expectations as far as reviewing and deployment. They indicated that if their EPM will allow the time, then they can do this early July before their RL sprint. So, assuming all goes well, this may be deployed in mid-July.

Re-assigning to their EPM.

Status update: Roan did some review and I addressed most of the reported issues. Roan is very busy.

Roan and Sam did a full pass over the extension. Roan left comments at www.mediawiki.org/wiki/User:Catrope/Extension_review/Translate. Looks like all of those have been addressed, in particular where it comes to Page Translation context.

Niklas' deployment access is in the works. Plan is to have Niklas deploy under guidance of a mentor (possibly Roan) as soon as review is done and Roan and Niklas have the time.

Latest comments from Sam (IRC): "I left a few TODOs for Niklas, but I don't see anything particularly wrong with the code. Robla wanted Tim to have a look over it.

Assigning to Rob, who wrote earlier today that Tuesday PDT there *might* be some time to push this further with Tim (crossing fingers).

We're currently working on getting Translate's Page Translation feature active on meta. We hope it will be there in the next few hours. I just trained Jon Harald Soby and Peter Gehres.

Configuration details are at http://etherpad.wikimedia.org/translate for engineering. Some details in words:

  • everyone will have translator rights
  • there is a group called translation coordinators that can make pages translatable (or untranslatable again)
  • no translation services like Translation Memory or Machine Translation (Google Translate, Microsoft Translator, Apertium) will be available. Main reason is that the apaches cannot make calls to the outside, and there may be licensing issues with the content providers.
  • Translations will be stored in a separate namespace (1198 for now)
  • Translation in English itself is disabled
  • Some special pages in the Translate extension are disabled, namely FirstSteps (translatewiki.net specific user registration page), ManageMessageGroups (only needed when file based message groups are used), ImportTranslations (no offline translation mode is supported), TranslationStats (graph creation disabled).

More later.

Extension was deployed on meta.wikimedia.org. There is one demo page active[1]. There are a few minor issues known. One should be fixed before production pages are added. As it is past 03:00 or 04:00 in NL and FI respectively, Niklas and I are going to sleep now...

  • NORMAL: renaming a page translation page generates too many log entries (http://userbase.kde.org/index.php?title=Special:Log&limit=500&type=pagetranslation )
  • NORMAL: after marking an updated page as translated, I came back at special:PT and saw the old version. Force reload gives the expected version.
  • CRITICAL: derivative pages appear to be updated only to the one but latest translation. May be a master/slave issue. To be researched.

[1] https://meta.wikimedia.org/wiki/Special:PageTranslation

aokomoriuta wrote:

I tried this extension on meta.
I found one bug about a summary for translation.

View below translation page's diff URL.
This revision(oldid=2984676) makes translation from "Markup" into "マークアップ" (Japanese).
https://meta.wikimedia.org/w/index.php?title=User:Siebrand/Pagetranslation_test/ja&diff=next&oldid=2984673
But the summary for this revision says 'Created page with "特別ページ"', which doesn't match this revision's change.
You can see there's the right summary 'Created page with "マークアップ"' on the old revision(oldid=2984673)'s, and 'Created page with "特別ページ"' must be on the next revision(oldid=2984686)'s.

There seems to be one gap between the translation content and the summary.

(In reply to comment #14)

There seems to be one gap between the translation content and the summary.

That is probably the issue marked CRITICAL in comment 13. We're going to look into that, and before page translation can be used "for real", we should have this fixed. It might be as easy as getting the content from the database master instead of the slave.

With r100004 deployed, this should be considered resolved. Thanks go out to all that have helped in getting this done!