WikiLoop/2020 Year in Review
Spotlights
edit- With a new Feed Engine component for WikiLoop DoubleCheck, we enabled topic-based review feeds such as US Election 2020 Feed and COVID-19 Feed, to allow editors interested to focus on relevant articles they want to review.
- As more label data became available, we received requests from WMF's JADE and various third-party systems to supply training data from WikiLoop DoubleCheck to these systems. We built an infra and completed the integrations.
- We revamped i18n infra and scaled our i18n support from 6 languages to 23 languages, becoming one of the most internationalization-supported alternative Wikipedia review tools and scripts.
- We received independent coverage from external media such as FastCompany and The Wikipedia Signpost, and also got our blog post published on Google Search official blog "Keyword" and Google Open Source Blog through our engagement efforts.
- We crossed our first 100K user reviews, 1000+ anonymous editors and 400+ authenticated editors, 67 stars, 41 forks and 23 code contributors on GitHub. We achieved a site reliability of 99.6% for Prod instance in the first year we provide reliability monitoring.
- Our coalition conference brought 10+ institutions from industry, academia and open knowledge including Knowledge Futures Group@MIT, Credible Web Community Group@W3C, Scheme.org WikiData, GLAM, OKRG, Allen Institute for Artificial Intelligence, OpenRefine, Google Brain, YouTube, WMF AI team, etc.
Development
editWikiLoop DoubleCheck, an open-sourced app for Wikipedia counter-vandalism review
- User facing features
- Expanded languages supported from 6 to 23 languages
- Quick Keyboard Shortcuts
- UserPreference pilot
- Support DirectRevert
- Improved ReviewCard
- Improved Avatar
- Improved Leaderboard
- Infra & Data Pipeline
- Revamped review feed storage
- Launched Feed Engine for WikiLoop DoubleCheck
- Built a wikilink traverse pipeline to pull relevant revisions, and for discovery of relevant topics
- DevOps
- We launched a WMF-hosted version
- We established Heroku Monitoring, UptimeRobot Monitoring, and PM2 monitoring.
- We established logging through LogDNA
- We improved "Deployed on-to Heroku" button
- We set-up Continuous Integration testing on CircleCI and GitHub Actions
- Stability monitoring
- 99.6%(577.28ms avg) for Prod on Heroku, whole year
- 87.3%(379.04ms avg) for WMF-hosted (beta), since 2020-05-10
- 99.90%(477.46ms avg) for Canary, since 2020-07-13
- Analytics
- Revamped analytics metrics mechanism for server and client
- Build a statistic chart page
- Tech Integration
- We shared data with JADE, LDM
- We supply filtering with ORES
Pilot projects
edit- CivilityCheck: a project to evaluate the civility in the comments of Wikipedia discussions in order to address the problem of abuse that leads to declining editorship within the Wiki community. (by intern team of Deus Nsenga, Baelul Haile, David Ihim host and lead by Elan Houticolo-Retzler)
- CrossEditsPatternDetection: a pilot project to add functionalities for WikiLoop DoubleCheck to detect suspicious patterns of vandalisms across multiple edits (built by intern Haoran Fei)
Community outreach
edit- Launched a name vote campaign to get the name WikiLoop DoubleCheck. Target: WLDC users, 34% of users voted.
- WikiLoop communication campaigns combined drove 29% increase in WLDC users, and a 60% increase in content revisions using this tool between July and Aug 2020
- Established our pilot reward mechanism: a on-wiki barnstar, and built an automatic mechanism (media coverage)
- Established massive message outreach mechanism to communicate editors who signup to receive WikiLoop DoubleCheck communications.
- We started a Facebook Page and a Discord Community
- We connected WikiLoop DoubleCheck server to Discord channel
Coalition
edit- We started a WikiLoop Coalition Conference Call, a conference call bringing industry and academia to join the effort for helping open knowledge movement at scale.
- The conference since then received support and participation from institutions from Knowledge Futures Group@MIT, Credible Web Community Group@W3C, Scheme.org WikiData, GLAM, OKRG, Allen Institute for Artificial Intelligence, OpenRefine, Google Brain, YouTube, WMF AI team, etc.
Thank you!
editWe owe a huge thank you to all:
Eng: Xinbenlv(Lead), Elan HR. (TL and intern host), Chaoyue L., Hamdanil R., Ashwin R.
Eng Intern: Haoran F., Deus N., Baelul H., David I.
Comm Lead: Maria Cruz,
Comm: Monica W., Prad N.
Open Source Developer(github ID): langner, epicfaace, hrasyid, bhavyakaria, tzuhsiao, sohomdatta1, fmobus, aligoren, renamoo, pgrimaud, papuass, luisfors-g, he7d3r, grovina, dz-s, apatronl, antoinealb, alex-martelli, adehtiarov, MT-Wizard, FlorianKoerner
... and everyone who kindly helped and made this possible. Thank you!
We will continue to build data and tools for Wikipedia to boost productivity of editors and make editor-curated data more accessible to the open knowledge movement, and will continue to bring community and institutions together to support Wikipedia and open knowledge movements at scale. Please stay tuned for updates. Thank you!
On behalf of Project WikiLoop,
Macruzbar (talk) 01:12, 25 March 2021 (UTC)