Wikimedia Language engineering/Reports/2015/August

Wikimedia Language engineering monthly report for August 2015

Monthly updates from WMF Editing - Language team

Content Translation edit

Updates from the ongoing development and deployment for Content Translation.

Content translation was improved to create cleaner wikitext while publishing
Publishing failures because of parsoid rejecting HTML from translation was reduced to zero
Several bug fixes related to image, template, reference adaptations were done
Content translation now uses the element ids generated by RESTBase, that improved how we can align a newer article's sections with a translation of older version. This also addressed some bugs about translation of some sections being lost when CX not able to restore it against source article's sections
Machine translation engine selection and disabling used to apply only to the next paragraph. Now it also applies to the current paragraph.
Worked on the backend to provide suggestion to translate. Suggestions feature is soon coming to the Content translation dashboard

Details	Numbers
Articles created (during the month)	5147
Articles created till end of month	17708
Number of new translators (during the month)	1444
Highest number of articles created by one user (cumulative number till end of the month)	413
Highest number of translators for a Wikipedia (cumulative number till end of the month)	808 (es)
Total number of translators with only 1 published article (cumulative number till end of the month)	3769
Total number of translators with only 2 published articles (cumulative number till end of the month)	723
Total number of translators with only 3 published articles (cumulative number till end of the month)	338
Total number of translators with only 4 published articles (cumulative number till end of the month)	175
Total number of translators with >= 5 and <= 20 published articles (cumulative number till end of the month)	480
Total number of translators with >20 published articles (cumulative number till end of the month)	132
Articles deleted (during the month)	276

MLEB 2015.08 was released after a four month break since last release. It contains a lot of changes and it is last version planned to include support for the MediaWiki 1.23 LTS legacy release. It has already been downloaded 40 times only few days after the release and the previous release 2015.04 is a top 3 most downloaded releases with over 450 downloads.
The work on the Outreachy project by phoenix303 on improving Special:SearchTranslations of Translate is mostly finished. New features are available on Wikimedia sites in early September. We welcome feedback and bug reports on the new features and improved user interface.
Other extensions including Universal Language Selector and cldr had minimal updates to increase language support coverage and other small maintenance work.
Supported by volunteers and volunteer work, translatewiki.net keeps supporting translation of MediaWiki core and extensions. Thanks to our volunteer translators, MediaWiki gained a support for three new languages: Southern Kurdish, American sign language[1] and Adyghe. A new project, Crosswatch, was added for translation. The ElasticSearch version used by translatewiki.net was updated to 1.7 and its databases were reindexed to support the new translation search features.
Also by volunteer work, CLDR plural rule parsing code has been put into a new library. Work continues to make MediaWiki core use that library.

[1] An external JavaScript library is needed to display the script.

Translatable pages: Meta 2002 (+119); MediaWiki.org 2529 (+86); Commons 392 (+7)
Translate extension is installed on 196 wikis (+6), according to Wikiapiary

Apertium service (apertium-apy) has been updated to fix memory leak issue (T107270) with collaboration of upstream and Tech Ops team.
New campaign 'article-recommender-1' was enabled in ca, en, es, fa, fr, it, sw wikis.
cxserver registry configuration was moved from Puppet to cxserver. This avoids large configuration files in Puppet repository.