3 ms·

So the menus rotate about once per month for most restaurants. And all of a sudden diners have to scroll through a ton of menus to find something they want. We

So the menus rotate about once per month for most restaurants. And all of a sudden diners have to scroll through a ton of menus to find something they want. We can't use item-to-item collaborative filter because we always have a sparse data problem on the new menus. So instead we are trying to create NLP embeddings on the menus and then ranking new menus based on similarity to old menus the diner liked.

Terretta 6y ago
Does this dovetail with "tasting menu spenders" interest in exploring novel tastes? For some categories of interests, the best approaches branch out rather than converge in. Also, many find perusing menus a positive part of the experience, like choosing a place to go on a trip.