- Confirm translations are ready
- List of trusted user groups
- Translate "Edit quality (20k sample)"
- Run prelabeling script
- Load revisions into labels.wmflabs.org
Description
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | Ladsgroup | T163007 Deploy ORES Review Tool for Latvian Wikipedia | |||
Resolved | None | T130213 [Epic] Edit quality models (damaging/goodfaith) | |||
Resolved | • Catrope | T192499 Deploy ORES advanced editquality models to lvwiki | |||
Resolved | Halfak | T163006 Train/test damaging & goodfaith models for Latvian Wikipedia | |||
Resolved | awight | T163005 Complete Latvian Wikipedia editquality campaign | |||
Resolved | Halfak | T162032 Edit quality campaign for Latvian Wikipedia |
Event Timeline
@Papuass, it looks like we don't have translations for lv yet in the labeling interface, so users will just see English.
If you want to get things translated, check out https://translatewiki.net. Specifically, we'll need translations for the basic interface and the damaging_and_goodfaith form.
We'll also need a list of "trusted user groups". These user groups are only given to users who are highly trusted. We'll filter their edits out of the labeling system so we don't waste people's time reviewing good work. For English Wikipedia, we use
- sysop
- oversight
- bot
- rollbacker
- checkuser
- abusefilter
- bureaucrat
One more thing. Can you provide a lv translation of "Edit quality (20k sample)"? We'll use this as the title of the edit quality labeling campaign.
I will work on translations.
Our trusted users are:
- sysop
- bureaucrat
- bot
- oversight (0 users)
- checkuser (0 users)
- patroller
- autopatrolled (Should we add these? This group is manually assigned for proven users.)
Edit quality (20k sample) = Labojumu kvalitāte (20k paraugs)
We should wait for half a week (probably untill Monday) until the translation gets added to the codebase. After that, I make the campaign and stuff.
(3.4)halfak@ores-compute-01:~/projects/editquality$ cat datasets/lvwiki.autolabeled_revisions.20k_2016.json | json2tsv autolabel.needs_review | sort | uniq -c 17337 False 2654 True (3.4)halfak@ores-compute-01:~/projects/editquality$ cat datasets/lvwiki.autolabeled_revisions.20k_2016.json | json2tsv reverted_for_damage | sort | uniq -c 19423 False 568 True
Looks like we'll be labeling 2654 revisions which is a smaller set than usual (usually ~5k), but I think seeing 568 that look like they were reverted for damage (best guess) is pretty good, so let's move forward. If we end up not having enough observations of damage from this campaign, we can always do another.
halfak@wikilabels-01:~/datasets$ sudo -u www-data /srv/wikilabels/venv/bin/wikilabels new_campaign lvwiki "Labojumu kvalitāte (20k paraugs)" damaging_and_goodfaith DiffToPrevious 1 50 --config /srv/wikilabels/config/config/ {'active': True, 'tasks_per_assignment': 50, 'created': datetime.datetime(2017, 4, 14, 16, 9, 37, 160265), 'name': 'Labojumu kvalitāte (20k paraugs)', 'wiki': 'lvwiki', 'view': 'DiffToPrevious', 'labels_per_task': 1, 'id': 56, 'form': 'damaging_and_goodfaith'} halfak@wikilabels-01:~/datasets$ cat lvwiki.autolabeled_revisions.20k_2016.json | grep '"needs_review": true' | wc # | sudo -u www-data /srv/wikilabels/venv/bin/wikilabels task_inserts --config /srv/wikilabels/config/config/ 56 2654 24591 300295 halfak@wikilabels-01:~/datasets$ cat lvwiki.autolabeled_revisions.20k_2016.json | grep '"needs_review": true' | sudo -u www-data /srv/wikilabels/venv/bin/wikilabels task_inserts --config /srv/wikilabels/config/config/ 56
The labeling interface is here: http://labels.wmflabs.org/ui/lvwiki/
Stats can be found here: http://labels.wmflabs.org/campaigns/lvwiki/?campaigns=stats
All looks good. I'll have translations deployed soon. For now, the UI seems to be falling back to English.
Translations are deployed! https://wikitech.wikimedia.org/w/index.php?title=Nova_Resource:Wikilabels/SAL&diff=1756578
Looks OK to my eyes.