Page MenuHomePhabricator

CX2: Don't apply unmodified content restrictions to translations published to the user namespace
Open, MediumPublic

Description

Based on the amount of unmodified machine translation (MT), Content translation can prevent from publishing (T190283) or add the published article to a tracking category (T190798) based on different limits. This is great to encourage users to review the translations. However, in some cases users publish an initial rough version on their user namespaces and these mechanisms can get in their way.

On the one hand, users should be able to publish a rough MT version in their user namespace to edit it further with other tools (as described in this comment) and pointing the attention of the community (including them in the tracking category) to review those does not make much sense.

On the other hand, the user namespace provides also a good opportunity for trying how the limits work to adjust them (e.g., try if a lightly edited machine translation is prevented from publishing) without the risk of publishing low quality test content in a more relevant namespace. So we want to provide a way for users to still be able to optionally experience the limits in the user namespace if they want to.

Proposed solution

If the user publishes the content under the user namespace (and only in that case!):

  • Do not prevent from publishing even if the unmodified content exceeds the threshold. The article should be published and the error should not be shown to the user.
  • Do not add the article to the tracking category. Warnings will still be show to encourage reviewing, but the content is not added to the tracking category if it is finally published.
  • Provide an option to enable/disable the application of the limits.

Event Timeline

Pginer-WMF triaged this task as Medium priority.Feb 6 2019, 11:14 AM
Pginer-WMF moved this task from Needs Triage to CX2 on the ContentTranslation board.

I agree. The warnings should probably still be shown, but the publishing shouldn't be prevented. Sometimes it's easier to fix the issues after the page is published in some way.

This came up also at https://nl.wikipedia.org/wiki/Wikipedia:De_kroeg/Archief/20200410#Feedback_vertaaltool in comments by user Bertux.

~~~~~

Hello Developers !!!

The proposal written above is not enough.

Who wrote this proposal ?

The person who wrote the proposal did not understand the problem.

You should discuss in [[:mw:Talk:Content translation]] not here.

Thank you !!!

--~~~~

I wrote before that it is inappropriate to consider this problem here at this stage.

I refrain it.

That is because, generally, members of Wikimedia Phablicator did not actually experience this trouble.

Therefore, the discussion here should be discontinued and it should be discussed in [[:wm:Talk:Content translation]].

All Wikimedia Phablicator members should have a MediaWiki.org account. Why don't you write your opinion on [[:wm:Talk:Content translation]] ?

I will point out specifically that the discussion at this stage is inappropriate.

The text ''If the user publishes the content under the user namespace (and only in that case!):'' in the proposal may reflect the desire to not be held responsible for content tranlation (and its developers) when bad translations are created.

However, content translations will not be criticized even if the created article has inconveniences due to conten translation bugs (eg, the previously mentioned category problem). That's because most critics haven't used content translaion in the real world, and at best base it on what is written in MediaWiki. Therefore, it is only the users who are criticized.

This condition indicates that the author of the proposal does not understand the actual situation described above.

Thank you !!!

--~~~~

To show what the criticisms of content translation look like, I will show you a previous discussion in Japanese wikipedia:

[[:ja:Wikipedia:井戸端/subj/コンテンツ翻訳ツールにおける日本語への機械翻訳をツール側で禁止する提案]]

Perhaps, few members can understand Japanese, but you can get an overview using machine translation.

In this discussion, participants mention the warnings about unmodified text produced by content translations.

However, in my experience, this warning is quite random and even if I replace the strange word order generated by machine translation until I (a native Japanese speaker) feel it natural, the warning does not disappear.

However, the participants in the discussion argue that this function is reliable.

In other words, the participants in the discussion are discussing with reference to the Javanese discussion without actually using content translation.

In addition, proposal writers dwarf the problem to just waring against unmodified text.

In [[:wm:Talk:Content translation]] there has been reported many bugs in wider ranges in the function that prohibits publishing .

Therefore, the discussion here should be discontinued and it should be discussed in [[:wm:Talk:Content translation]].

Phabricator is the bug tracking and project management system used by the Wikimedia Foundation. It is used to capture tickets to improve our products that developers pick and implement. Tickets are created capturing ideas and proposals from different channels, including the Content Translation talk pages, but also user research observations and more.

If there is a relevant piece of information that could improve the proposal, please share it. I'm sorry but from all your comments above I cannot figure out which is the problem you find with the current proposal and what you think should be different. Are you suggesting to keep the warnings when publishing under the user namespace? to remove them from all namespaces? to improve the way those warnings are shown? Unless we have a clear description of which is your proposal, we cannot do much to support it.

Thanks!

I acknowledge that some tasks are better done outside the content translation tool, but this can easily be abused since it's possible to move userspace drafts into article namespace pretty easily. Therefore, I'm against not putting those articles in the tracking category. If it's possible to change whether the articles are put in the tracking category depending on the user (with administrators given the right to change that), that would be fine though.

@HaussmannSaintLazare: Please stop assigning this task to yourself and please refrain from adding disruptive comments. Work is being planned in Wikimedia Phabricator. Follow T215403#6139593 and bring up general questions not related to the topic of this task in other places. Thanks for keeping this constructive.

This is really necessary because the translation tool is quite buggy. Being allowed to have it saved into your personal space would solve many issues.