Build API for article misalignment. This will have a few underlying components:
- Language-agnostic quality model: map any Wikipedia article to [0-1] score where 0 = no content and 1 ~ featured article.
- Language-agnostic importance model: per last quarter's work (T272175#6894768), use pageviews as basic proxy for importance while other more flexible filters (e.g., topic, country, occupation) are developed. Normalize to [0-1] range.
- Language-agnostic misalignment model: difference between quality and importance.
- Language-agnostic misalignment metric: summarization of intensity of misalignment in a project.
The API then has two pieces:
- Precomputed scores for all wikis
- Function for computing article misalignment for single articles on the fly