Request for feedback: Automatic article section recommendations (ar, es, fr, ja, ru)

Howdy.
The Wikimedia Research team is developing a new algorithm that automatically recommends sections for editors to add to articles. This algorithm uses machine learning to suggest sections that may be relevant to the article because other articles on similar topics contain these sections. Section recommendations are currently available for English, Spanish, Japanese, Russian, Arabic and French Wikipedias. This week, we are looking for experienced editors like you to take a look at these recommendations and provide feedback to help us improve them! If the recommendations are useful, we will use it in tools for editors who want to expand stub articles.

We have built a testing app that allows you to browse section recommendations. Once you’ve tried the tool on some articles in your language(s), please provide feedback on the project feedback talkpage about the quality of the recommendations. You can also provide quick feedback through the tool itself.

If you are interested in helping out, please read the instructions and get started!

Thank you,

Jonathan Morgan & Diego Saez-Trumper, Wikimedia Foundation Research

4 Likes

In general, I am very sceptical about suggesting possible contributions to contributors. But if it “works”, fine. Are there experiences that suggestions really make sense on the mid term?

1 Like

Ziko, we aren’t committed to using this for a specific product yet. But one idea we’ve been considering is that section recommendations could be useful for new editors (either in an edit-a-thon context, or as part of an onboarding workflow). We know that new editors tend to want to jump in and create new articles right away–and we know how problematic that can be for them and for the wikis. So we thought that perhaps we could instead point them towards expanding stubs, and give them some ideas of what kind of information to add to those stubs (via section recommendations). What do you think of this?

3 Likes

Yes, that is the problem with newbies… :-) I see your point.

1 Like