Author here: We hope to fix this w/ the techniques described in the post! The search engine will first collect a high-recall set of candidates which is then passed to a high-precision ranker. This should help you w/ your cuisine and dish queries misclassifications. It will also introduce semantic understanding into the query engine meaning if you type "French" it will not give you "French fries" but French cuisine.
This seems like a pretty good write up, but I do wonder if it’s necessary? I use grub hub and Uber eats fairly often. If I want to search for a restaurant I would hope it’s just doing a match on exactly what I search with minor spelling checks.
If I’m looking for a dish or cuisine, why not display it in a visual fashion like Uber eats? There’s usually not that many options to actually sift through.
For example the first graphic of the embedding space has a very clear mixture of restaurant names and food items. Feels a bit messy.
