This task involves the work of defining and implementing a process for how the Tone Check model will be retrained.

Story

• As a new(er) volunteer who is in the midst of making an edit that causes Tone Check to become activated, I want to be confident the feedback I am being offered is aligned with the current state of Wikipedia policies and conventions, so that I can know I am making choices that align with what the people who will be reviewing/moderating/patrolling the edits I'm making expect.
• As an experienced volunteer who is concerned with ensuring new text that is added to Wikipedia is written in a tone that is aligned with project policies and conventions, I want for the Tone Check model to be perpetually kept in sync with relevant policies and guidelines as they evolve, so that I can be confident the feature is offering people to feedback that is aligned with what I/the project is expecting.

Requirements

To retrain the Tone Check model effectively, the Machine Learning Team will need the following...

• The text that caused each Tone Check to be shown
• For each span of text that caused a Tone Check to be shown, whether someone tapped to Revise what they had written or tapped Decline to continue on with what they originally wrote
  • Note: If someone chose to Decline revising, the reason they gave for that specific Tone Check instance.

Background

This need for this ticket was prompted by an offline conversation with @SSalgaonkar-WMF in the context of T389443.