This project aims to develop and test a classifier to identify good feedback from a corpus of unmoderated/non-featured article feedback (v.5) posts.
The design goals of AFTRank are the following:
- reduce the workload of feedback moderators by prefiltering feedback not worth being moderated:
- automatically hide or decrease the score of bad feedback without the need of hitting AbuseFilter
- increase the score of unmoderated good feedback
- correct the bias towards high-traffic articles produced by the current UI of the FeedbackPage:
- allow moderators to explore the large amount of feedback from low-traffic articles that hardly receives any attention
- surface feedback from articles with an expected higher feedback quality but low moderation activity