The Pirates of Silicon Valley is a movie about the creation of Apple, and Microsoft, and about the hard work of these pioneers of computer crafting (because it was a craftsmanship at that time), and the not always orthodox methods they were using. This biographical movie went through the emotion of the first results in these newborn companies, and the intent to be out of the system: “Better to be a pirate than be in the Navy”, used to say Steve Jobs, founder of Apple Inc.
This spirit of liberty (or rebellion) was represented by the early team of Machintosh computer by applying a pirate flag (with a rainbow apple as eye patch) on the building they were working, as tells Andy Hertzfeld, one of the team members. Why I am talking about this? because many people asked me from where I would “gathering data from various sources”, as I wrote in a previous post.
The answer is in what the U.S. Copyright Office says:
Copyright law does not protect recipes that are mere listings of ingredients. Nor does it protect other mere listings of ingredients such as those found in formulas, compounds, or prescriptions. Copyright protection may, however, extend to substantial literary expression—a description, explanation, or illustration, for example—that accompanies a recipe or formula or to a combination of recipes, as in a cookbook. (last revised 2/12)
In other words, the list of ingredients, or a mere listing of data, like cooking time or temperatures, is not copyrightable, while the explanations in a cookbook, or the text of a publication about cooking falls under the copyright law. A word to the wise is sufficient…
We were saying few time ago that what Computer Science applied to Cooking misses dearly is a database of recipes. Actually, young researchers from GTI joined me in this adventure, strongly pushed by the idea… and by their boss, Josep Blat, who has always been supportive.
So now we are three working on this project: a technician and a PhD student, plus an experienced PhD student that in not directly involved in the project. Their background is mainly related to Human-Computer Interface, which sounds perfect to me because this in a certain way, completes my “hard computer science” AI approach. Some other student briefly joined us, and now moved to Yahoo! or to other (founded) projects.
The first task we are dealing with is to create a base to work with: the famous recipe database. I will start gathering data from various sources, Miguel is proposing a database structure, and Alan will be in charge of the parsing . In few weeks, and after few script, we will have some results to computer. Stay tuned!
Parsing is the process of analysing a string of symbols, according to the rules of a formal grammar. For our task, the aim of parsing is doing a is syntactic analysis of recipes, i.e. identifying all the ingredients, the different steps of the cooking process, etc.
During another visit at Alicia, in Manresa, I had the occasion to spy a kitchen from the inside, and learn from it.
It al started with a chat with Jaume Biarnés on my future projects. After launching with enthusiasm a serie of new projects, some of them utopian but fun, some of them very concrete, he proposed me to go and see how a professional kitchen works. That’s how I met Sergio, the chief cook at La Fonda, one the the restaurants of the St Benet complex, the others being Món and L’Angle (the latter with one Michelin star).
This is the kind of restaurant that works with fixed menus. This means that the clients have a choice over 2 or 3 dishes, and that the cooks can prepare a lot of the food in advance. This is called la mise en place: to anticipate part the work in the hours before lunchtime. This lowers the rush at the moment of serving, which generally happens in those restaurants that work à la carte: with a vast selection of dishes, that have therefore to be prepared at the moment.
Working with a lot of dishes implies having a lot of chefs to cook them when the clients make the orderz. This is the kind of rush you see in restaurants shown in chef Ramsay’s Hell’s Kitchen, but with a reduced amount of special effects and drama. On the opposite, working with fewer dishes, allows you to better organise your work: Sergio explained me that in this kind of restaurants, a single chef can deal with 30-40 clients.
I came back from that day with Sergio with a lot of notes, that took me months to elaborate. This experience gave me more insight of the real work, and where some process can be optimized; because that’s what we’re talking about: optimizing processes in the food industry.
As long as it seems that I am too busy right now to work myself on the Culinary AI project, we decided to give some space to students to help us developing part of the work. This is amazing for me, because it opens the possibility to confront my ideas with younger ones, and to clarify them… almost in the culinary sense!
We are then looking for students finishing they undergrads studies and well-disposed toward applied research. The idea is to develop a Computer Science prototype as final result of the project, but also to push the student in developing other analytical skills, more focused on the use of technology related to food. Just a couple of examples:
To create a structured database of recipes
This comes from the remark the I haven’t found a well done recipes database, structured and with methodised information. The challenge is twofold: fist, an historical and cultural research should provide us with a reasonable set of recipes to record, then a correct format has to be found and developed to record correctly the data.
To analise nutrient to provide an alternative aliments grouping
To use the well known “Food guide pyramid” is restrictive under many points of view, mainly because it fails to capture many details of a diet. It has been revisited recently, but some aspects are still to be researched. The student will use the USDA nutrients database to identify groups of aliments with similar characteristics via clustering analysis techniques.
Finally, in the context of the very recommendable Master in Intelligent Interactive Systems (MIIS), a student already applied to do her master project working or recommender systems. What a crowd!
First, our collaboration with Alícia went on, focusing on common projects to carry on. A study of metabolic diseases seems to provide both a common ground and very interesting technical developments. This will be the direction we will push the next applications, also because there is a concrete possibility to improve the quality of life of many patients.
Second, I wrote, in collaboration with people from GTI (Interactive Technologies Group) at University Pompeu Fabra, a research proposal for RecerCaixa. It is a one year program that provides some founds to help young researchers from Catalonia in developing their projects. The aim of RecerCaixa is to finance innovative scientific ideas with a social impact. That’s here that can come in an intelligent system specially aimed at providing solutions for people with a metabolic disease. Our proposal describes an adaptive tool that assist people following a nutrition therapy, i.e. following a diet adapted to a patient’s needs. Nutrition therapy is the most common, inexpensive, and efficient therapy that people with a metabolic disease can follow.
Writing down a full project, thinking about all its implications on society, or counting the number of people involved, helped a lot in realising what we were dealing with. To talk about your ideas, and to write them down for others to read are two faces of the same medal that do not look the same at all! Mainly, it forces you to focus on the objectives you have and on their practical feasibility. This is already a lesson learnt almost for free, regardless if we win the call or not.
Third, I had some more field research experiences. More at Alícia, where we dissected the mechanisms of a big restaurant, and more at Suite 7, where I learnt the difficulties of being chef for one night. But let’s keep these stories for another day…
Apparently there are no recipe database available out there. And by database we intend “a collection of information organized in such a way that a computer program can quickly select desired pieces of data”. This definition highlights the main property of a DB, i.e. each piece of information is separately labelled, as bottles in a winery.
This must sound a frustrated post… but it is! Moreover, looking around there are some stuff, but it is generally written in HTML (to be read in a browser evidently), but this sounds so old. What about using a pretty classification in XML and then dynamically build a web page on top of this piece of information? I am not talking about weird thing: this is commonly done on Internet, almost in every page you visit (if they don’t use some more advanced technique). This is even true for for commercial DB. So I ask: Why to pay 300$ (minimum) for a DB that is of no practical use? The truth is that commercial databases are no more than a web page embedded in a Access file. (translation: don’t waste your money, crawl the web).
I have the feeling that the food industry is really far away from IT. Once, chef Jaume Biarnés told me that there isn’t a web page that professionals can consult. Internet is the Paradise of free expression, so everyone is writing his own blog about cooking. This is wonderful, but this also means that for the professional user there is nowhere to consult a standard version of a classical recipe. Professionals of food industry modify standard recipes in astonishing ways, but they need to consult what the “fathers” said in order to improve it, and this needed piece of information should be the same for every cook, so they can start from a common ground and use the same language.
Before the WWW era, there were books that were classical, used in every cook school. Chefs can cite them right away, build on them, and communicate from this common language. Spanish cuisine took off in the last 20 years because of communication between chefs and information sharing. But it was before the WWW era, which apparently has still its progress to do in terms of clear and quick information sharing.
Surfing through the Net for information on taste, it encountered many discussions over cilantro. Apparently there is a very animated diatribe between cilantro haters vs. lovers, that has been exacerbated by the fashion of adding indiscriminately cilantro to dishes.
One of the ideas of my research -or a side effect of it, if you prefer- is to understand human taste for food. A piece of software is able to track patterns in user’s choices: when applied to food, this pattern is nothing more than the taste of a gourmet, a valuable information for any kind of food industry. This kind of information, even if tracked in a very different way, has commonly used, e.g. the cilantro fashion is justified by the fact that 81% of the world population is cilantro-lover.
What hides behind the love (or hate) for cilantro? Apparently the “haters” are unable to detect chemicals in the leaf that are pleasing to all those who like the herb. The chemicals that give the cilantro its typical fresh and spicy taste are found also in other fruits and herbs, like lemon grass and thyme, but there are also unsaturated aldahydes, found out George Preti, and this compound makes cilantro haters think of soap when they eat it.
To feel a flavour while ignoring others is what happens for cilantro haters because “it’s possible that they have a mutated or even an absent receptor gene for the receptor protein that would interact with the very pleasant smelling compound”, suggests Charles J. Wysocki of Monell Center for basic research on the senses of taste and smell. This lack of receptor can be genetic, as a study on 41 pairs of identical twins and 12 pairs of fraternal twins seems to point out.
The physical capacity of feeling odors and flavors is another central aspect of the whole experience of eating. Even if no a lot of publicity is given to these biological and genetic factors, I wonder how they bias the data about user’s taste, for example, and how these results have to be integrated in the training of an artificial intelligence.
Coriander (aka cilantro or Chinese parsely) is the common name of the plant Coriandrum sativum, a member of the carrot family Apiaceae or Umbelliferae. The young leaves are the herb called cilantro, while the older leaves and seeds are called coriander – although the herb is commonly referred to by both names.