Page MenuHomePhabricator

Make the threshold for Chinese Wikipedia to prevent publishing 5% more strict
Closed, ResolvedPublic

Description

An initial adjustment to limit the publication of articles with more than 70% of unmodified machine translation (T246383) was reverted (T252371) because it became too strict.

This issue may be caused by the way modifications are computed (T251893). While the issue is investigated we propose to make the limits more strict but only by a reduced amount.

Considering the above, the proposal is to adjust the threshold for Chinese to prevent publishing when overall unmodified content is higher than 95%.

Event Timeline

Pginer-WMF triaged this task as Medium priority.May 14 2020, 3:26 PM

@Pginer-WMF I suggest to change the limit to 5% since there is no word segmentation in place before calculate the rate of unmodified machine translation. In T251893, the user who got blocked by the limit has an unmodified machine translation rate of 93%, but the translation is good enough to be published.

Pginer-WMF renamed this task from Make the threshold for Chinese Wikipedia to prevent publishing 10% more strict to Make the threshold for Chinese Wikipedia to prevent publishing 5% more strict.May 14 2020, 5:05 PM
Pginer-WMF updated the task description. (Show Details)

@Pginer-WMF I suggest to change the limit to 5% since there is no word segmentation in place before calculate the rate of unmodified machine translation. In T251893, the user who got blocked by the limit has an unmodified machine translation rate of 93%, but the translation is good enough to be published.

This is super useful input. Thanks!
Sure. Let's start with a 5% adjustment. I already updated the ticket description.

@Pginer-WMF Thank you. I will test CX for how the threshold holds up in Chinese when the change is deployed and I will share details with you if I find something.

Change 596916 had a related patch set uploaded (by KartikMistry; owner: KartikMistry):
[operations/mediawiki-config@master] Make the threshold for Chinese WP to prevent publishing 5% more strict

https://gerrit.wikimedia.org/r/596916

Change 596916 merged by jenkins-bot:
[operations/mediawiki-config@master] Make the threshold for Chinese WP to prevent publishing 5% more strict

https://gerrit.wikimedia.org/r/596916

Mentioned in SAL (#wikimedia-operations) [2020-05-18T11:10:23Z] <lucaswerkmeister-wmde@deploy1001> Synchronized wmf-config/InitialiseSettings.php: SWAT: [[gerrit:596916|Make the threshold for Chinese WP to prevent publishing 5% more strict (T252786)]] (duration: 01m 06s)

I have informed the community about this change.