Now that we have predictions for article translation for some language pairs, we could use these predictions for suggesting a 'morelike' API for missing articles. Given a wiki (source wiki), the user enters an article title that exists on the wiki. Our goal is to suggest similar articles that are missing from the wiki. Here's how it should work:
- Look up article's Wikidata ID;
- Look up article's title in other languages (for which we have predictions ready — see above link);
- Find similar articles in those languages using the 'morelike' API;
- Using the predictions for translation, filter out existing articles in the source wiki;
- Order articles by their predicted normalized ranks (normalized pageviews an article would receive in the next six months if created);
- Show these ordered article Wikidata IDs to the user.
- (Optional) We can also show our machine translated titles for these articles.