Research talk:Automated classification of article importance/Work log/2017-05-22
Monday, May 22, 2017 edit
Today I'll follow up on communication with WikiProject Africa, and continue developing the view rate pipeline.
WikiProject Africa edit
We got some feedback from a member of WikiProject Africa, asking some good questions about the project. A couple of articles were specifically mentioned, Colonialism and Born in Africa. I decided to load in the dataset and check our data for these articles.
Unsurprisingly, Colonialism is in the 99th percentile for both inlinks and views, with 2,810 inlinks and 52,649. The "Born in Africa" album article is 12th percentile in inlinks (it has 3 inlinks), and 16th percentile in views (116). Their predicted ratings are therefore not too surprising.
When I first trained a model for WP:Africa, I mentioned the lack of performance for High- and Mid-importance articles. Based on the scatterplot above, it is fairly straightforward to see that their ratings are not closely related to number of views or inlinks, although for Top- and High-importance they do to some extent favor articles that are visited more often.