Prof Ganesh Bagler is the pioneer of ‘Computational Gastronomy,’ the emerging data science that blends food with Artificial Intelligence. Trailblazing research from his lab has established foundations of this niche area that deals with food, flavours, nutrition and health. In an exclusive interview, Prof Bagler shares his audacious dream of transforming the global food landscape through data-driven innovations
What does computational gastronomy involve? What inspired you to take up this unique career path?
Computational Gastronomy blends food and data with the power of computation for achieving data-driven food innovations.
With training in physics and computational techniques, as a teenager I have had aspirations of being an astronomer. Far from being interested in culinary nuances, I have been rather agnostic about food. Thus, it has been a serendipitous journey from astronomy to gastronomy.
My journey into Computational Gastronomy started in a classroom. In the winter of 2014, I chanced upon the thought of investigating the food pairing pattern in the Indian cuisine while teaching the ‘Complex Networks’ class in IIT Jodhpur. The idea was to probe the food pairing in Indian recipes vis-à-vis its Western counterparts which had been shown to be following a uniform blend of ingredients.
Starting from there, to the discovery of unique positioning of spices in Indian recipes, all the way to laying the foundation of a new niche known as Computational Gastronomy, it has been an exhilarating journey.
What do you mean by culinary fingerprints of cuisines and taste prediction algorithms?
Culinary fingerprints refer to unique features that make a cuisine special. Traditionally, a chef or a culinary enthusiast would have an intuitional idea of culinary features that distinguish a cuisine from the rest of the cuisines.
By looking at the cuisines through the lens of data, Computational Gastronomy extracts the quintessential features of regional cuisines labelled as ‘culinary fingerprints’. Which ingredient categories (dairy, vegetable, spices etc.) dominate a cuisine? Which ingredient combinations feature the most prominently in the recipes of a cuisine? What flavour molecules characterise the recipe composition of a cuisine? Culinary fingerprints capture the essence of these questions to present a data-driven view of cuisines revealing their architecture.
Food ingredients are composed of flavour molecules that lend them their taste and odour. A large number of these molecules have been characterised with their taste and odour. The research done on taste predictions algorithm in our lab, as well as by many other laboratories across the world, focuses on predicting the taste of an organic compound. We have devised state-of-the-art machine learning algorithms for predicting the bitter and sweet taste of small molecules. This research will be of value for creating low-calorie artificial sweeteners that would help tackle the problem of obesity and diabetes.
Is India ready for it? How can computational gastronomy be applied to the Indian food sector?
The crux of Computational Gastronomy is in leveraging meticulous data compilation on all aspects of food and their analysis for bringing about food innovations. If there is anything that I have learned about Computational Gastronomy in the last six years of my journey, it is that this is the future of food.
In the last couple of decades, many aspects of our lives have been transformed by the application of data and computation – be it tagging on Facebook or logging into Macbook. Artificial Intelligence is entering our lives in a big way. The question is whether and how this revolution is going to change food and cooking which are primarily seen as an artistic endeavour. My wager is that data-driven analysis of food will dramatically change the way we look at food in decades to come.
Beyond the international food giants, the Indian food sector (can’t name the companies which have used and are consulting with me due to confidentiality reasons) has responded enthusiastically and has woken up to the potential of Computational Gastronomy. However, I must say, this is primarily restricted to large FMCG and F&B companies.
In the light of Covid-19, do you think a data driven approach to food can help chefs prepare healthy food for their guests?
We have gathered scientific evidence for the health impacts of food ingredients from around 38,000 research articles. This compilation is available in the form of DietRx database. Chefs can creatively use ingredients that are known for their benevolence to create recipes that are healthier. I am also a co-founder of Ayusla Café, a start-up incubated at IIIT-Delhi Innovation and Incubation Center, where myself and Rishi Agarwal are trying to create immune-boosting recipes by dipping into the Ayurveda knowledge base.
Tell us about your FlavorDB and DietRX projects?
Created in my laboratory (Complex Systems Lab, IIIT-Delhi), FlavorDB is the world standard for the rich data on flavour molecules found in natural ingredients. It also provides a way to find ingredients that share flavour molecules, and therefore can be used as a basis for creating uniform or contrasting food pairing, as desired. Many culinary professionals such as the Michelin star Chef Garima Arora have reported the use of food pairing principles for creating recipes.
As explained above, DietRx is a compilation of health impacts of 2,222 dietary ingredients. By investigating these data, we have shown that culinary herbs and spices have a broad-spectrum benevolence against a range of disorders. Eventually, this resource is expected to be of value for dietary recommendations and for designing healthier diets.
How do flavours and food pairings work together?
Food pairing is a measure of the extent of sharing of flavour molecules across the ingredients of a recipe. The taste of a recipe ultimately depends on the combinations of ingredients that go into it. Food pairing attempts to capture this notion.
What advice would you give chefs working in the hospitality sector? How can they use computational gastronomy to satisfy guest palates?
Computational Gastronomy is at an early stage of its evolution. By dissecting food into its constituent elements (traditional recipes, ingredient flavours, nutrition, and health impacts) it offers data-driven solutions for food innovation. Food and beverage companies are particularly poised to gain from this apart from chefs and culinary enthusiasts.
Chefs could use FlavorDB for exploring the flavour basis of ingredients and for exploring novel food pairings. Similarly, they could use DietRx for creating healthier recipes. They could use RecipeDB, a structured repository of over 1,18,000 recipes from across the world to identify recipes that fit into their criteria (cuisine, ingredients, processing technique, nutritional value).
Thanks to the power of permutations and combinations, the potential number of recipes that can, in principle, be generated is astronomically large. Aligned with this thought, in my lab, we are working on developing novel recipe generation algorithms. Having created a structured compilation of worldwide recipes, we hope to build Artificially Intelligent programmes that would capture the intuition of an expert chef to create recipes that have not been realised yet. This would lead to interesting culinary experiments for chefs in collaboration with a computer. However, this is a work in progress and you will have to wait for this dream to be realised for a while!