Page MenuHomePhabricator

Identify articles that should be de-prod'ed.
Open, Needs TriagePublic


In English Wikipedia, proposed deletion (prod) is often overruled when it was inappropriate. Could we build a model to detect when a prod'ed article should probable be de-prod'ed?

Here are some features that Aymatth2 suggested:

  • article age
  • number of editors
  • number of confirmed editors
  • number of other language versions
  • presence of talk page
  • number of project tags
  • characteristics of the nominating user
  • length
  • number of citations
  • number of inbound links

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald Transcript

How would we get some good labeled data for this? Is there a log event when an article is Prod'ed that we can look for?

Category:Past proposed deletion candidates has talk pages of articles that survived proposed deletion. For getting articles that were actually deleted, it should suffice to go through the deletion log looking for PROD or a link to Wikipedia:Proposed deletion.

I would also like to suggest daily views as an input in this model.