I've contemplated doing a similar thing but in a different domain (may take advantage of your library). I had an idea for seeding the recommendations by scrapin

I've contemplated doing a similar thing but in a different domain (may take advantage of your library). I had an idea for seeding the recommendations by scraping reviews from review sites and mapping their ratings onto likes/dislikes. Not sure how easy that would be though.