tl;dr
The WMF Research team is conducting a second labeling campaign to evaluate the Revert Risk model for Wikidata. Please assist by visiting this link and labeling each revision as Keep, Not Sure, or Revert.

Hi folks,

The Research team at WMF is currently conducting a second labeling campaign to evaluate the Revert Risk model for Wikidata. As you might already know, this is part of the ongoing efforts to develop a new generation of machine learning models supporting patrolling work on Wikimedia projects.

In pursuit of this goal, the Research team at WMF created the Revert Risk models to help identify content that "might be reverted". The Revert Risk models for Wikipedia are already in production, and the focus now is on evaluating the model for Wikidata. Many of you participated in the first labeling campaign (thanks to you all!), and now we seek your assistance once again for this second campaign to gather more data.

Kindly visit this link and label each revision in one of these three categories: Keep, Not Sure, Revert. Please note that "Not Sure" should be used in all cases where the Keep or Revert labels are unclear to you. you.

If you have any questions, please contact Diego via e-mail: diego(at)wikimedia.org or on Meta (Diego (WMF).

Thanks,

-Mohammed

On Thu, Aug 10, 2023 at 9:27 AM Mohammed Sadat Abdulai <mohammed.abdulai@wikimedia.de> wrote:

Hi everyone,


As you're likely aware, the Wikimedia Foundation's Machine Learning and Research teams have been working on migrating from ORES to Lift Wing — a new open-source machine learning infrastructure. This shift brings a host of new capabilities and simplifies the process of retraining models over time. (For more details, see the previous announcement)


Lift Wing has already been trained using a dataset comprising reverted and patrolled edits. However, it would be extremely helpful to have additional new training data to help the model get even better at detecting problematic edits on Wikidata. Therefore, we need your help.


How You Can Contribute

The Research team built a tool to make your involvement easy and effective. This tool allows you to label new training data quickly and efficiently. You can find the tool here: Annotation Tool. It will show you an edit and ask you if you would keep or revert the edit. You can skip any you are not sure about.


By participating in this process, you're helping enhance the accuracy of the bad edits detection system on Wikidata, making it more robust and reliable.


If you encounter any issues or want to provide general feedback, feel free to leave us a note on this ticket phab:T341820.


Cheers,

--
Mohammed S. Abdulai
Community Communications Manager, Wikidata

Wikimedia Deutschland e. V. | Tempelhofer Ufer 23-24 | 10963 Berlin
Phone: +49 (0) 30 577 116 2466
https://wikimedia.de

Grab a spot in my calendar for a chat: calendly.com/masssly.

A lot is happening around Wikidata - Keep up to date! Current news and exciting stories about Wikimedia, Wikipedia and Free Knowledge in our newsletter (in German): Subscribe now.

Imagine a world in which every single human being can freely share in the sum of all knowledge. Help us to achieve our vision!
https://spenden.wikimedia.de

Wikimedia Deutschland — Gesellschaft zur Förderung Freien Wissens e. V. Eingetragen im Vereinsregister des Amtsgerichts Charlottenburg, VR 23855 B. Als gemeinnützig anerkannt durch das Finanzamt für Körperschaften I Berlin, Steuernummer 27/029/42207. Geschäftsführende Vorstände: Franziska Heine, Dr. Christian Humborg