Make the Machine translation stricter than what it is right now by 45% in Kurdish Wikipedia (ku.wikipedia)
Discussion is here
Description
Description
Details
Details
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Content Translation: Set MT threshold to 45% for Kurdish WP | operations/mediawiki-config | master | +1 -0 |
Related Objects
Related Objects
Event Timeline
Comment Actions
Hi @Balyozxane, thanks for taking the time to report this! What do "stricter" and "45%" mean exactly? Can you provide steps to reproduce a problem, please?
Comment Actions
Hi @Aklapper
What I mean by machine translation is the built-in machine translation on the Content Translation tool. People have been abusing this tool by publishing low quality translations which are basically unmodified machine translations. We have the same problem on ku.wiki. After reading the Translation quality help page, I discussed the issue with @UOzurumba who suggested making the Machine translation stricter by maybe 10%. When I discussed the issue with local contibutors, we came to an agreement that we should start with a higher restriction and lower the restrictions per user feedbacks.
To reproduce the problem, go to ku.wiki Content translation and start a new translation and see how much you need to change to publish an unmodified translation. A user was able to publish a translation by removing a word, and adding two prepositions. Please, see the table at Communtiy page, the original text, machine translation, aforementioned user's translation and my edits (more like rewrite) to the said translation are given side by side. Also you can reproduce the same results by going to cxdebugger tool and typing 1868548 as the Translation ID.
What does stricter mean? Add another 45% to whatever it is right now per Translation quality help page. Maybe it is similar to phab:T321819 I hope user UOzurumba and the Language Team can provide with more details.
How did we decide on 45%? When I use the content translation on mobile, the robot usually comments by saying "You've changed the text 40%" or sometimes lower other times higher. I believe 45% to be a little high but it should give us what needs to be changed to get the correct rate.
I hope this is clear enough. Thank you!
Comment Actions
Change 890947 had a related patch set uploaded (by KartikMistry; author: KartikMistry):
[operations/mediawiki-config@master] Content Translation: Set MT threshold to 45% for Kurdish WP
Comment Actions
Change 890947 merged by jenkins-bot:
[operations/mediawiki-config@master] Content Translation: Set MT threshold to 45% for Kurdish WP
Comment Actions
Mentioned in SAL (#wikimedia-operations) [2023-02-22T08:12:14Z] <kartik@deploy1002> Started scap: Backport for [[gerrit:890947|Content Translation: Set MT threshold to 45% for Kurdish WP (T324941)]]
Comment Actions
Mentioned in SAL (#wikimedia-operations) [2023-02-22T08:14:10Z] <kartik@deploy1002> kartik: Backport for [[gerrit:890947|Content Translation: Set MT threshold to 45% for Kurdish WP (T324941)]] synced to the testservers: mwdebug2002.codfw.wmnet, mwdebug1002.eqiad.wmnet, mwdebug1001.eqiad.wmnet, mwdebug2001.codfw.wmnet
Comment Actions
Mentioned in SAL (#wikimedia-operations) [2023-02-22T08:22:56Z] <kartik@deploy1002> Finished scap: Backport for [[gerrit:890947|Content Translation: Set MT threshold to 45% for Kurdish WP (T324941)]] (duration: 10m 41s)