I actually went to the web site just to see, and it's worse than I thought. Even their Western menu, the stuff random Westerners think is "Chinese food" is presented as JPEGs of photographs (sometimes out of focus) of the physical menu, which is itself strewn with typographical errors and mysterious annotations.
So, to get even the bad text an actual patron has in the physical restaurant you need to scrape the site, download the images, and successfully OCR from low resolution out-of-focus photographs. It's not impossible but good luck to you, and at the end the results will still be pretty unsatisfactory. "Frind pok" is actually what they wrote, they meant Fried Pork, but that's not an OCR error it's really what they paid to have printed.
However while the restaurant's manager might care, as I understand it her husband is the hard core chef who ensured it's a success, why should he give a shit? Presumably the errors are just in the text for stupid barbarians like me - many of them don't even order from his real menu anyway. His taste is what matters, nobody comes to the restaurant because of the typography or web site design, they come to eat his food.
The last one is a problem because real family restaurants do want to raise prices/be more upscale too, but none of their customers will let them because they expect banh mi to be $3.
Additionally, I have it in my head that not bothering to fix menus shows a certain admirable pragmatism. "Frind pok" is not correct but it is correct enough.
So, Google and these other companies, the option is - build it yourself and try to do better, or buy data from the companies that do this at varying degrees of quality, or don't have menu data at all. Except the last option, people _want_ menu data, it's one of the most common things people want to know about a restaurant.
Instead eveyone is busy building their own little feudal kingdom and they call it platform even if it's actually just a toll booth
So I’m sure you could sign up yelp and google and Uber eats and everyone else for a common data standard, but you’d then still have to go chasing the restaurants to go put that information in a system somewhere.
We haven’t even really been able to convince businesses to put their opening hours online, it’s still such a problem that one of my interview questions at google in 2018 was “name as many ways as possible you might be able to discover a business’s opening hours online”
Menus are about an order of magnitude more complex than that - it’s a tough thing to get restaurants to do.
Yet their hours are wrong, and different, on google and bing.
if I own a business there is no singular place I know of where I can publish my opening hours. My laundrette here can't publish the fact that they will be closed due to sickness except by placing a sheet of A4 at the door.
"A structured representation of food or drink items available from a FoodEstablishment."
Clearly the issue is that no one has convinced FoodEstablishments of the business case to spend tech dollars publishing their menus.
More forward-looking restaurants manage this all themselves as part of their digital strategy, but it’s still a small percentage and disproportionately located in the US.
I’m not saying Google doesn’t also scrape or use other sources, I’m sure they do, but this is one of those situations where the whole system is broken. Tbh one of the bank shot benefits of having all of these digital delivery services is that some restaurants are using aggregators that can also publish menu data.
As far as the authors idea about markup for menus, that’s great, but highly improbable for a bunch of reasons: most restaurants don’t update their menu frequently, dishes are often difficult to represent structurally, POS systems are often modeled differently than the printed menus, etc.