WikiLoop: Big Tech's Open Knowledge contributions
summary
Google pilots WikiLoop to feed machine learning and boost editor productivity by building concrete tools and data pipelines. It ships WikiLoop Double Check, a vandalism review app that uses ORES scores, an API and random sampling to avoid reinforcing model bias and hits 100,000 reviews after outreach with a 30% user growth. It also ships a civility check via Perspective API prototype from interns, logs false positives and negatives, and plans an open civility dataset while hitting API rate limits.
AI-written summary · about
chapters (9)
- 0:00WikiLoop and team intro
- 1:11Agenda and program overview
- 1:34Virtual circle for open knowledge
- 2:38Three aspects of WikiLoop projects
- 7:38Community engagement and tool naming
- 13:39Demo: Double Check tool
- 19:43Civility Check for editors
- 28:33Q&A
- 29:39Open discussion
related talks
42mData, Privacy, & Trust in Open Source: 10 Lessons from Wikipedia
35mThe Most Human Documentation
53mThe joy of volunteering with open technology and culture
21mÜber Tech Lead On Code Revi
29mWhat is LocalWiki, and why is it so much fun? Let's edit it!
27mThe bots that help run Wikipedia
32mHumans In the Loop: Collaborating in the Age of AI
18mBelieve it or not: Designing a Human-AI Partnership for Mixed-Initiative Fact-Checking