Last Saturday night, I
ran into my supervisor on one of the many random bridges in
Amsterdam. I told him about the long-overlooked study of formal
cooking, and was glad to hear that his envisioned range of
possibilities and research avenues for this emerging field coincided
with my own. Which isn’t so surprising, since I must've gotten
these ideas somewhere, and since he has been applying the
(linguistic) modeling approach he’s been working on for some two
decades to other domains, like music and reasoning.
So what can you
do with formalized and hierarchically structured recipes? Well, you
can generate new recipes and parse ingredients to form recipes. Let’s
start this week with the idea of generating recipes. Suppose you have a large number or
recipes, in tree diagrams, and you’ve formalized and uniformized
all ingredients, appliances and cooking procedures. What you can do
then, is extract parts of the recipe-trees, and combine them,
possibly in completely novel ways, with other extracted tree-parts.
Formally, a context-free grammar should be able to deal with this (hence the trees -- and I welcome discussion on whether cooking is context-free or maybe only regular or even context-sensitive).
Next, we can see how likely certain combinations are. Frying
the water and soaking the celery overnight doesn’t sound like a
likely way to make soup. Frying onions and soaking beans overnight
more so. But the model can find larger sub-recipes that occur across
recipes too: frying onions, adding spices, adding vegetables, adding
water seems like a very common string of operations where the result
of the previous operation is combined with another ingredient. So we
can assume that the cook has Gestalt principles and stores this
ordered set of operations, or 'sub-recipe' as a whole, in order to make
the computing more efficient.
Enough theory for now,
after all: theory without her data is nothing. Next week I’ll say some more
about how you can parse ingredients to make a soup with whatever you
have lingering around in the cupboard. Informally, of course, because
the algorithm has to be patented and made into an app so I can fund
my own post-doc.
This week, we add to
the corpus of soups the next in the list of souvereign states, Albania. It’s again a bean soup, from this site. I
promise I won’t always add bean soups; that would skew the sample
and be boring. I also promise I won’t always follow the
alphabetical list. I already got some good suggestions for yummy
soups that will break this emerging pattern, and I’m happy to do
so.
Interesting about this
soup is (again) its bi-componential nature, where you cook things on
the one hand (in this case, beans, carrots and onion), and make a
paste of other ingredients, by frying, on the other. Here, the paste
does not revolve around the mint and coriander, as in the last
recipe, but contains bell pepper, oregano and some flour to give the soup it's gravy-like texture. The procedure is pretty
much the same otherwise. Notice how the flour is baked with the paste
here, instead of being added to the whole of the soup later on (which
I recall from other soups). Also notice that, as opposed to the
Afghan soup, we boil some vegetables (carrot & onion) with the
beans, giving a richer taste to the beans, which remain clearly the
head of this group. Perhaps we can think of the carrot and onions as
modifiers. Yes. That’s a nice thought. They modify the beans, and
apparently the Afghan cuisine allows bare, unadorned bean-phrases
(BPs) whereas the example of Albanian cuisine shows a richer BP.
We’ll see in future soups how this issue of complexity works out for other soups
containing BPs.
The other day, I had
lunch with my Macedonian colleague. When I mentioned that I had an
Albanian bean soup on the menu for this week, she started talking
about a Macedonian soup her mother makes. Interested in the
possibility of Kochbund-effects, known for the Balkan-region, I
decided to do some elicitation. Not completely to my surprise, it was
nearly the same recipe, but with slight variations. The pepper is
cooked with the beans, rather than added to the fried paste, and (and
this one I really loved) the paste has to make a sizzling sound when
you add it to the BP, otherwise it’s not hot enough (feature
checking or something similar, mayhaps?). She continued about a Macedonian
liver soup, which sounds totally delicious and like something I
should make on one of those cold winter days ahead of us.