Page MenuHomePhabricator

Identify Wikipedia communities to test the language-agnostic revert risk model for Automoderator
Closed, DeclinedPublic

Description

As described here, there are two models available to us to detect the likelihood that an edit should be reverted - a model supporting 47 languages, and one which is language-agnostic. As most large Wikipedia communities are covered by the former, including our pilot wikis, we need to spend a little time identifying which communities are not supported by this model, and would need to use the language-agnostic one. We could then test it with them.

Questions

Of the Wikipedia communities which are not in the languages listed here, which are the 10 with the:

  • Largest number of administrators?
  • Largest number of active editors?
  • Most edit reverts per day?

Event Timeline

Samwalton9-WMF renamed this task from Identify a Wikipedia community to test the language-agnostic revert risk model for Automoderator to Identify Wikipedia communities to test the language-agnostic revert risk model for Automoderator.Jul 18 2023, 10:00 AM
Samwalton9-WMF created this task.

We're going to pause this investigation until the Research team figures out which model they think is best.

KCVelaga_WMF changed the task status from Open to Stalled.Sep 3 2023, 8:38 AM

We're using the Language-agnostic model everywhere since the Multilingual model isn't ready.