Page MenuHomePhabricator

DE2 - Ratio of Abstract-created articles
Open, Stalled, HighPublic

Description

  • Definition:

Average ratio between monthly net Abstract-integrated articles and new organically-created articles, both measured with a 30-day survival rate, across 3 wikis

Where:

  • Integrated articles = articles created through Abstract Wikipedia and rendered on a language wiki
  • Organically-created articles = articles created directly by editors on that wiki using existing tools
  • 30-days survival rate = articles not deleted or overwritten after 30 days of their creation
  • Wich wiki(s) is not confirmed as yet.
  • Technical Implementation/Data Source

Blocked on T421837

  • Baseline

0

  • Target

TBD after the three target wikis are defined

  • Dashboarding

Blocked on implementation

  • Approval

TODO @MMiller_WMF

Event Timeline

This comment has been deleted.
Miriam added a subscriber: LMorgantini-WMF.

Some updates:

We finalized with the Abstract team a new definition: Ratio between net Abstract integrated articles and net organically-created articles, tracked monthly, per wiki . Target=1.

Implementation of this metric (and, consequently, dashboarding), is not possible at the moment. The Abstract team is building the framework T421837 for individual wikipedias to integrate abstract articles. Only after that's finalized, we will be able to instrument that and  start querying and collecting data. The timeline for this is likely end of Q4. @mforns is meeting Abstract engineers next week to get a more precise estimation and agree on the earliest time at which DPE can start working on the instrumentation

Miriam changed the task status from Open to Stalled.May 1 2026, 5:05 PM

Changing status to stalled until we can proceed on the implementation.

@mforns checking if you have more clarity around timelines here? Thank you!

Miriam updated the task description. (Show Details)
Milimetric renamed this task from DE2 - Net new Abstract Wikipedia articles to DE2 - Ratio of Abstract-created articles.May 13 2026, 2:12 PM

@SCherukuwada @Miriam

We want to square off more specificity here:

  • wiki size should be representative enough; also 3, or 4 or 5?
  • determine quality of article e.g. higher than stubclass?

I chatted with Marshall yesterday. Let's go with 3 wikis. For quality measure, can we work in language around survival rate? Maybe something like, "Average ratio between monthly net Abstract integrated articles and new organically-created articles, both measured at a 30-day survival rate, across 3 wikis"