Description
Description
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | None | T254724 Experiment: homepage content suggestions in the KaiOS app | |||
Resolved | nshahquinn-wmf | T271816 Finalize strategy for automatic generation of top-read articles list |
Event Timeline
Comment Actions
We now have a clear method for selecting trending articles:
- Get the most-viewed articles in the country, properly accounting for traffic to redirects
- Remove articles where less than 10% or more than 90% of the traffic came from computers (as opposed to mobile devices)
- For each day, remove any articles that were shown in the previous 7 days.
- Manually remove a set of bad recommendations (e.g. "Main_Page", "News", "Pornography", "XXX_(film_series)") if they show up.
I've backtested this method from 1 February to 14 March and the results look reasonably good.