Add damaging and goodfaith models for huwiki

[tgr]

Also update the list of trusted groups.

Bug: T185903

[codecov]

Codecov Report

Merging #142 into master will decrease coverage by 7.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #142      +/-   ##
==========================================
- Coverage   20.93%   13.91%   -7.03%     
==========================================
  Files          58       56       -2     
  Lines        1070      999      -71     
==========================================
- Hits          224      139      -85     
- Misses        846      860      +14

Impacted Files	Coverage Δ
editquality/codegen/generate.py	0% <0%> (-100%)	⬇️
editquality/utilities/generate_make.py	0% <0%> (ø)	⬆️
editquality/utilities/fetch_labels.py	0% <0%> (ø)	⬆️
editquality/utilities/merge_labels.py	0% <0%> (ø)	⬆️
editquality/tests/test_config.py
editquality/codegen/tests/test_generate.py
editquality/config.py
editquality/feature_lists/lvwiki.py
editquality/codegen/config.py	0% <0%> (ø)
editquality/codegen/util.py	0% <0%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 81fed22...5d11a50. Read the comment docs.

[halfak]

Adding these now will cause some warnings, but no big issues. Essentially, we'll have labels for some edits where "needs_review" is false with "reason" == "trusted_group"

[halfak]

1/10th of all edits are damaging? that seems wrong.

[tgr]

@halfak pointed out that I used the percentage of hand-labeled edits where I should have used the percentage of total edits so I redid the commits using the ratio between

 ack '"goodfaith": true' datasets/huwiki.labeled_revisions.40k_2016.json | wc -l

and

ack '"rev_id"' datasets/huwiki.labeled_revisions.40k_2016.json | wc -l

I'll update the docs once the PR passes review.

[tgr]

ack '"rev_id"' datasets/huwiki.labeled_revisions.40k_2016.json | wc -l gives 39678 so I am not sure if that's the correct denominator or I should have used 40000 instead (although it's not much difference). I suppose the missing 322 revisions are deleted ones?

[halfak]

I'd recommend using the 39678 as the denominator -- but it shouldn't make too much of a difference. I think you're right and 322 are revisions to deleted pages or they have been rev-deleted.

[adamwight]

Needs a rebase. The master branch has an updated merge_labels algorithm which simplifies the Makefile logic.

[adamwight]

This patch should drop the huwiki.reverted model too, AIUI.

[adamwight]

P.S., you should have privileges on this repo, and if the branch were under the org repo I could push my rebase: https://github.com/wiki-ai/editquality/compare/awight-huwiki

[tgr]

Closing in favor of #152 which uses a local branch (it did not seem possible to change the source for this one).