This week's update can be found on-wiki here: https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Updates/2021-03-25
--
Back in December 2020, we introduced the Abstract Wikipedia list to Aisha and Jade, Outreachy interns participating in Abstract Wikipedia / Wikifunctions data science.[1]
We want to share the exciting results of their work. During the past few months, Aisha and Jade developed a Vue.js web application to help people find user generated code across the MediaWiki Scribunto modules written in Lua. Their creative approach involved a complex data pipeline using Wikimedia Cloud Services which analyzed the usage and similarity of code across different language editions of Wikipedia and its sibling projects. This approach is intended to aid users as they consider the problem space of consolidation of certain types of code into pure, re-usable functions in the Wikifunctions project.
The application can be found here:
https://abstract-wiki-ds.toolforge.org
We hope you enjoy the following 3 minute demo video of the web-based tool:
https://www.youtube.com/watch?v=M31scvrfGdE
Here is the longer 41 minute audio discussion with Aisha and Jade and the rest of the Abstract Wikipedia team about their data science work:
https://commons.wikimedia.org/w/index.php?title=File%3AAbstract_Wikipedia_Da...
You can review the extensive data analysis and pipeline development and web-based tool material, including Jupyter notebooks (on the PAWS public notebook system) and code hosted in Wikimedia GitHub, starting from here:
https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Data#Get_important_module...
We want to thank Aisha and Jade for their excellent work during the internship, and we hope they return in the future.
If you'd like to collaborate on the data pipeline, further analysis, or the web-based tool, please do drop a line on the abstract-wikipedia mailing list and we'll see if we can connect on Phabricator and GitHub to grow this work further. We believe the approach taken here, if further generalized, may also be of interest as we look deeper at the natural language constructs across Wikipedias and sibling projects for Abstract Wikipedia, and as more folks explore topics like global templates as well.
We'd like to thank Srishti and all of the Wikimedia Foundation's Developer Advocacy unit for their support in Outreachy efforts and Wikimedia Cloud Services capabilities, as well as for the encouragement from Isaac in Wikimedia Foundation's Research unit to foster this collaboration with both Aisha and Jade - they formed a productive partnership and as the mentoring group we were delighted to work with Aisha and Jade.
We're delighted to have such a supportive global movement and knowledge network helping to increase the impact and awareness of this valuable initiative.
-The Abstract Wikipedia Team
https://meta.wikimedia.org/wiki/Abstract_Wikipedia
P.S.: We also wanted to let you know that an overview of how all the pieces fit together is now published in the April 2021 edition of the Communications of the ACM:
https://cacm.acm.org/magazines/2021/4/251343-building-a-multilingual-wikiped...
[1] https://lists.wikimedia.org/pipermail/abstract-wikipedia/2020-December/00033...