Page MenuHomePhabricator

Allow using ContentTranslation only for autoreview, patroller, and sysop at trwiki
Closed, DeclinedPublic

Description

Hello,

Please disable ContentTranslation for users who don't have autoreview, patroller, or sysop rights.

Community discussion: İçerik çevirmeni (Robot çeviri) aracının kısıtlanması teklifi

Thanks

Event Timeline

Change 891408 had a related patch set uploaded (by Stang; author: Stang):

[operations/mediawiki-config@master] trwiki: Restrict ContentTranslation to autoreview/patroller/sysop

https://gerrit.wikimedia.org/r/891408

Change 891408 merged by jenkins-bot:

[operations/mediawiki-config@master] trwiki: Restrict ContentTranslation to autoreview/patroller/sysop

https://gerrit.wikimedia.org/r/891408

Mentioned in SAL (#wikimedia-operations) [2023-02-23T14:19:36Z] <ladsgroup@deploy1002> Started scap: Backport for [[gerrit:891408|trwiki: Restrict ContentTranslation to autoreview/patroller/sysop (T330363)]]

Mentioned in SAL (#wikimedia-operations) [2023-02-23T14:21:38Z] <ladsgroup@deploy1002> stang and ladsgroup: Backport for [[gerrit:891408|trwiki: Restrict ContentTranslation to autoreview/patroller/sysop (T330363)]] synced to the testservers: mwdebug2002.codfw.wmnet, mwdebug1002.eqiad.wmnet, mwdebug1001.eqiad.wmnet, mwdebug2001.codfw.wmnet

Mentioned in SAL (#wikimedia-operations) [2023-02-23T14:30:46Z] <ladsgroup@deploy1002> Finished scap: Backport for [[gerrit:891408|trwiki: Restrict ContentTranslation to autoreview/patroller/sysop (T330363)]] (duration: 11m 10s)

Anerka subscribed.

Hi,
It caught my attention that users without autoreview rights can still use ContentTranslation. Could you check & fix it?

Some examples from last few days;
ContentTranslation (new user)
ContentTranslation (old user without autoreview rights)
ContentTranslation (old user without autoreview rights)

Thanks for let me know. I'm not within these groups on trwiki, and I found I could reproduce this issue: an article published published by me. I could receive this notice said I could not publish

image.png (859×832 px, 92 KB)

But when I clicked the "Publish" button, an article is created successfully
image.png (274×1 px, 32 KB)

I wanted to share some statistics over a longer period of time to provide some perspective, and get input from the Turkish Wikipedia community that could be very useful to improve our tools for the future.

Below you can see three graphs for Turkish Wikipedia capturing translation activity from January 2019 to June 2023:

  • Published translations. While for most months the number of published translations is below 2K, there are some peaks of activity during July 2021 (7.39K translations), December 2021 (3.1K translations) and February 2022 (3.5K translations).
  • Percentage of deleted translations. While for most of the months the number of deleted translations are below 20%, there are three months (coinciding with the three peaks of activity mentioned above) where the percentage of deleted translations is much higher (80% of the published translations being deleted). This means that most of the additional translations for those months above the usual activity were deleted.
  • Translations by user edit count. While for most of the months translations are published by users with different levels of experience, for the above three peaks of high activity with low quality translations most of the translations were published by users with 1K edits or more. In the last graph you can see the green line (users with 1K-10K edit count) and orange line (10K+ edits).
monthly-translations-at-top-10-wikipedias-2023-07-12T15-26-00.522Z.jpg (376×705 px, 25 KB)
monthly-rate-of-deleted-translations-2023-07-12T15-26-28.018Z.jpg (400×705 px, 35 KB)
monthly-translations-by-user-edit-count-bucket-2023-07-12T15-31-14.363Z.jpg (376×723 px, 38 KB)

The above data seems to suggest that the issues of low quality articles are not generalized across the board, but they seem to concentrate on specific moments of high activity, and they don't seem driven particularly by inexperienced editors.
Based on that, a general limit for inexperienced users is not expected to be effective, and we may need to concentrate our efforts in understanding what was driven those peaks and identify measures to prevent those situations.

People more familiar with the activity on Turkish Wikipedia may be able to share more context about what happened during those months (July and December 2021, and February 2022). Was there any contest, vandalizing activities or any other unusual event that could explain a spike of editing activity?

As part of our plans to improve the current system of limits (T251887), we considered mechanisms that could help with problems we heard from other communities about campaigns/contests (T331023), but we want to hear more about the particular case of Turkish Wikipedia. Feel free to share any additional thoughts, questions, ideas, interpretations of the data or further context.

Thanks!

Stang removed Stang as the assignee of this task.Dec 24 2023, 2:28 PM
Stang subscribed.