Page MenuHomePhabricator

Apply a distinct tag to edits made using Content Translation 2 during its beta
Closed, ResolvedPublic

Description

We have a contenttranslation tag for CX edits generally, but there won't be any way to tell from the change tags which edits are made with CX2 and which with CX1. It is possible to tell this from the Content Translation–specific tables, but we're trying to centralize all our analysis in the Analytics Data Lake, where those tables aren't available.

So can we add one more tag to changes made with CX2? We can stop doing this once we've fully transitioned and go back to just looking at the umbrella contenttranslation tag. I don't have a strong opinion about what the extra tag should be named, but how about contenttranslation-v2?

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptAug 8 2018, 6:17 PM
Neil_P._Quinn_WMF renamed this task from Apply a distinct tag to edits made using Content Translation 2 during it's beta to Apply a distinct tag to edits made using Content Translation 2 during its beta.Aug 8 2018, 6:19 PM
Neil_P._Quinn_WMF edited projects, added ContentTranslation; removed CX-analytics.

Hah, okay, I guess something can't be in ContentTranslation and CX-analytics at the same time.

@Amire80, do I need to do anything special to get this triaged so it will get implemented soon? :)

Neil_P._Quinn_WMF triaged this task as High priority.Aug 8 2018, 6:24 PM
Neil_P._Quinn_WMF moved this task from Triage to Tracking on the Product-Analytics board.

It's probably an easy thing to do. It will be somewhere around here: https://phabricator.wikimedia.org/diffusion/ECTX/browse/master/api/ApiContentTranslationPublish.php$175

I haven't touched that code for a while, however, so I defer to @Nikerabbit for estimation.

@Nikerabbit, will we be able to do this soon? :)

The proposal looks good to me, I added the ticket to the quarter board (Language-2018-July-September).

One question is about whether it is possible (with a reasonable effort) to apply the tag retroactively. That is, to add the contenttranslation-v2 tag to those translations already created with version 2.

If it is not trivial, we can create a separate follow-up ticket to avoid increasing the scope of this one.
If it is hard, we can live with a few untagged CX2 translations.

Change 455916 had a related patch set uploaded (by Petar.petkovic; owner: Petar.petkovic):
[mediawiki/extensions/ContentTranslation@master] Introduce a distinct CX2 tag

https://gerrit.wikimedia.org/r/455916

One question is about whether it is possible (with a reasonable effort) to apply the tag retroactively. That is, to add the contenttranslation-v2 tag to those translations already created with version 2.
If it is not trivial, we can create a separate follow-up ticket to avoid increasing the scope of this one.
If it is hard, we can live with a few untagged CX2 translations.

After discussion with the team, we plan to only tag the new translations for now.

Change 455916 merged by jenkins-bot:
[mediawiki/extensions/ContentTranslation@master] Introduce a distinct CX2 tag

https://gerrit.wikimedia.org/r/455916

Etonkovidova closed this task as Resolved.Sep 3 2018, 5:20 PM
Etonkovidova added a subscriber: Etonkovidova.

@Petar.petkovic
The tag ContentTranslation2 tag is alwayas displayed along with ContentTranslation - was it intended?

The link for ContentTranslation2 works - redirects to https://www.mediawiki.org/wiki/Content_translation/V2
ContentTranslation2 tag is added to http://cx2-testing.wmflabs.org/index.php/Special:Tags.

@Petar.petkovic
The tag ContentTranslation2 tag is alwayas displayed along with ContentTranslation - was it intended?

Yes, that was an intention. To quote @Neil_P._Quinn_WMF:
For ease of analysis, I think we should set both tags on CX2 revisions: contenttranslation marking it as a CX edit generally, and contenttranslation-v2 making it as a version 2 edit specifically. That way, the contenttranslation tag will cover the first history, and we can then stop setting and querying the v2 tag once v1 has been fully replaced.

Petar.petkovic removed a subscriber: gerritbot.

@Petar.petkovic, @Pginer-WMF I just queried the data, and so far I only see two edits with this tag in the production cluster: this one on the Spanish Wikipedia and this one on the Romanian Wikipedia.

Does that seem roughly right to you?

@Petar.petkovic, @Pginer-WMF I just queried the data, and so far I only see two edits with this tag in the production cluster: this one on the Spanish Wikipedia and this one on the Romanian Wikipedia.
Does that seem roughly right to you?

Roughly yes. Access to version 2 is not very visible yet, so I don't expect many published articles. The work around T196435 can be useful to confirm data (you can check the graph here). That is based on the internal translation database, and for the recent period it shows 4-6 translation published. It would be good to check with @Amire80 if there may be some discrepancies to identify their cause (e.g., I assume that only translations published to the main namespace are counted, but we ma need to confirm).

We did a small outreach campaign in June asking 50 prolific users to publish an article with the new version and provide feedback, which may explain the spike in CX2 at the end of June. As we provide a switch for users to enable version 2 more easily (T194387) and do the corresponding announcements, we expect more people to use version 2 mor regularly.