|Declined||leila||T166186 [FY17-18] Community Health - Segment 3: Research on Harassment|
|Resolved||• DarTar||T171251 [Objective 3.1.2] Models for sockpuppet and toxic discussion detection|
|Resolved||• DarTar||T166190 Update page on Meta on indicators of toxic conversations|
Current plan is to do the following:
- Evaluate impact of context:
- Write a parser for the revisions XML to create conversation structure.
- Run parser on all revisions, so we have a structured view of all discussions in wikipedia talk pages.
- Evaluate the impact of crowd-sourcing annotations related to harassment with and without context of the conversation.
- Evaluate if we can find early indicators that lead to harassment.
- Build a visualization to view toxic a summary of comments by moth
- In real time as they come in, according to toxicity, showing which toxic comments are reverted, and which are still live.
- Requires developing a notion of which comments have been reverted (depends on parts of Evaluate impact of context)