Ganesh Bagler, a scientist at IIIT Delhi, is working on the data analytics of food, which is grown into a new area called computational gastronomy. Ganesh Bagler has been trained in Physics and Computer Science. He did a PhD in Computational Biology, an interdisciplinary training. Presently, he is working on various themes also called complex systems at IIIT Delhi.
Scientist Ganesh Bagler tells Restaurant India how chefs at restaurants can make the best use of the data using computational gastronomy.
When I use the word computational gastronomy, it encompasses analysis and investigation of data pertaining to food, flavours and health - that’s the thematic area.
After collecting the recipes and food data we did the food pairing for all the regional cuisines from across the world. It was published in the International Conference on Data Engineering; it was honoured with the best paper award.
We have been looking at patterns in terms of what kind of flavour molecules are combined in the traditional recipes from across the world. Previously, I was working on various other topics like ethel code of the discovery and brain network, etc. While teaching a class on IIIT group tour on complex networks, I picked up this paper which was looking at the flavour network in western cuisines, largely the European and American.
This paper was investigating a specific hypothesis given by Chef Heston Blumenthal from the UK – ingredients that are similar to each other in terms of their taste, flavours and odour tend to go with each other in a recipe. If there’s a bunch of ingredients like a basket of vegetables or dairy products, etc, Chef Blumenthal suggested any of these ingredients which have similarity among themselves in terms of taste, flavour and odour have a higher chance of being selected or being used in a traditional recipe.
The paper which I picked up and was teaching about it was investigating this hypothesis and confirmed that indeed the recipes of Western cuisines, American and European, tend to combine with those ingredients that are similar in terms of their flavour molecules.
I, then, started investigating on Indian recipes. At that time, I, along with my team, took 2,543 recipes from Tarla Dalal’s website and found that these recipes are composed of around 192 unique ingredients for which we could find the flavour molecules. We also gathered that flavour data was not available for certain ingredients, for example, hing or asafoetida. We investigated the pairing pattern during which, to our surprise, we found the Indian cuisines have dissimilar ingredients in their recipes. This means a typical Western recipe has a uniform blend of ingredients whereas a typical Indian recipe has a contrasting blend of ingredients. There is a diversity of ingredients present in the Indian recipes. It’s kind of a potpourri of flavours.
This ended up being a discovery in the novel food pairing phenomenon in Indian cuisine. It was highlighted as an emerging technology in the famous MIT tech review.
We finished it in 2015.
What I ended up doing after that:
Other than completing my other investigations, I continued working on this point knowing this particular area has a lot of potential both for asking fundamental questions such as why we eat, what we eat, can we design new recipes, as well as some applied questions.
We first compiled recipes from across the world, including India. And the number of recipes which we have now are approx 1,60,000 and these are themselves made up of a large number of ingredients - to the tune of around 5,000 unique ingredients. It’s a structured data; we have mentioned what kind of ingredients is used, from what source it comes from, are there some synonyms connected. We have data from 24 world regions, 71 countries.
The second was the work on the world’s first comprehensive repository of flavour molecules for natural ingredients. These ingredients which are used in the recipes be it tomato, chilly, mango, banana, etc., you’ll find all kind of ingredients.
Obviously, flavour molecules are not extracted and reported by food chemistry people for all ingredients, but whatever we could get, which are 1000 approximately, we created a huge repository of the flavour molecules of the ingredients. Earlier, they were scattered around; not only flavour but also an individual molecule. Let’s say what kind of odour and taste it is associated with, for example, one could be bitter and other could be sweet. We call these as flavour percepts and have recorded them all together. This was published in a well-known bio-medical journal called .
This will bring far more value to the chefs for looking into what kind of a property a particular ingredient might have, and also about food pairing. We also did an in-built food pairing.
A couple of months back, we made an Android app known as Flavour DB. I made it because chefs can get help from the app. I am associated with Indian Federation of Culinary Association (IFCA); Chef Manjeet Gill who is the president of IFCA has been an admirer of our work.
The information of different molecules, in natural ingredients as well as synthetic ones, is present. We have been asking a couple of questions which are linked to it like - is it possible for us to build an artificial intelligence algorithm for classification of a molecule in a computer. We have already put up this paper for review called bio archive, and we are making code and data for the same which was not available until date.
India is facing an epidemic of diabetes. In general, lifestyle and the dietary practice that people have been following has triggered the Type 2 diabetes at an early age. Therefore, a large population in India, potentially, can become diabetic, given the number of the patient reported so far and the anticipated reports by public health experts. Because sugar happens to be one of the critical factors in the diet we have in India, the question would be can we find out natural or natural-like molecule which is sweet but at the same time not adding calorie to the food we consume.
We, inevitably, consume food which, somehow, interacts with our body to give rise to desirable and, sometimes, undesirable effects. Because of over-consuming or taking it in a proportion which is not prescribed, we may end up having disorders. That’s where the question of what to eat and what not to eat lies when it comes to health and nutrition. The food itself is fairly complex in terms of what it constitutes micro or macronutrients; on top of it, body mechanism is vastly complicated.
Therefore, many researchers are studying the effects of red wine on health and more food ingredients. Often, people are coming up with contradictory conclusions on the effects of Red wine on health; they might be referring to a generic health indication or a specific disease.
Given this premise, within the purview of computational gastronomy, we wanted to compile all these scientifically published evidence in Medline. Doing it manually would not have been possible. We applied artificial Intelligence tools and created text-mining protocol - a protocol which will make sense of human-readable sentences.
This aggregates and will tell if the ingredient used is having a positive or negative impact on health.
We created Diet Rx – the idea was to create a repository of all the researches. It has 2,222 ingredients and effects on health, positive and negative, altogether.
Yes, we did food pairing of 2500 Indian cuisines as well. We picked up the cuisines from eight regions of India. These regions were decided on the basis of the geography as well as culture. We studied on the cuisines from South India, Maharashtra, Gujarat, Jain, Punjabi, Mughlai, Rajasthani and Bangla.
The food pairing can help you in designing Mughlai recipes. I am not claiming that we have an algorithm or technique but if you take an existing recipe and want to tweak it then food pairing data, through its analysis, can help you come up with the possible alternatives. It won’t tell you the existing recipe is good or it should change. We have taken it as a standard – the ground truth; we have assumed that it’s been cooked from centuries and the population, at large, likes it.
How it will help the industry? The chefs can think of tweaking the recipes, or even they can come up with a recipe from ground zero (starting from the scratch). One can think of food beverage pairing; given a recipe what kind of beverage can go with it, whether it is alcoholic or non-alcoholic; the industry has been interested in knowing this.
We look at the recipe, ingredients and flavour molecules and try to come up with various patterns which define them. You can call them quintessential features of a cuisine. That’s useful for the marketers in the food industry.
Given a certain bunch of ingredients, say a basket of tomatoes, onions, etc. if you randomly put them all together, even such recipe will have some food pairing. Ingredients have some similarity in terms of their flavour molecules. So, compared to such a randomized version, the real recipe of Western cuisine, tend to have a uniform positive pairing.
On the contrary, below zero or below the random cuisine, Indian recipes have lesser similarity when compared to the randomized bunch. That’s why we call it negative pairing. There’s nothing negative in terms of bad.
Copyright © 2009 - 2024 Restaurant India.