This article exists as part of the online archive for HuffPost India, which closed in 2020. Some features are no longer enabled. If you have questions or concerns about this article, please contact indiasupport@huffpost.com.

Can A Computer Cook Up The Perfect Recipe?

Researchers in IIIT Delhi have created a system called FlavourDB that analysed flavour molecules in food, to identify ingredients that will pair well together.

DELHI—“We wanted to be able to generate new combinations of food and try them out, and one of the things that the data showed was that tea pairs well with a number of fruits, like guavas, and citrus fruits,” said Chef Akshay Malhotra, executive chef at ITC Sheraton in Delhi. “That made me think of the fruit flavoured iced teas that had glutted the market some years ago, but it also gave me an idea, to try making a fruit salad with a tea vinaigrette dressing.”

Malhotra, a chef who graduated from the Culinary Institute of America in 2006, and apprenticed at the Ritz Carlton in Naples, is talking about the data generated by a research team led by Assistant Professor Ganesh Bagler at IIIT-Delhi. Bagler’s team has carried out research in computational gastronomy to identify the molecules responsible for flavours, and this can be used in a number of innovative ways, including helping chefs figure out what combinations would taste good without endless, expensive, and wasteful, trial and error.

To this end, Bagler and his students have created a website and free Android app called FlavorDB, which anyone can use. The team used AI techniques like NLP to scan cookbooks and tasting notes and other resources, in order to generate data about different ingredients, and also used peer reviewed data to understand the molecular composition of these ingredients, and then, based on their earlier research, the team is able to determine how compatible different foods are.

“One of the surprising pairings that I didn’t think about is strawberry and mushroom, which have a very high compatibility score on FlavorDB,” said ITC’s Malhotra.

Andrew Boyers / Reuters

Can data scientists co-create with chefs?

Bagler and his students first made headlines in 2015, when they published research that scientifically broke down the unique use of spices in Indian cuisine. “We looked at flavour molecules, both olfactory (which you experience through your nose) and gustatory (which are experienced in the mouth) and started identifying the molecules,” Bagler said. “Once that was done, we looked at pairs of ingredients in a recipe to find shared molecules, to create a food pairing index.”

The team looked at random pairings of molecules as a baseline result, and compared that to cuisines around the world. What they found was that Western foods tended to use a higher number of complementary pairs than a random mix. On the other hand, the use of contrasting flavours was a definitive trait of Indian food. And while specific fruits and vegetables were highlighted in Western food, spices were more important Indian food.

“Now we’re looking beyond pairs to triangles and quadrangles, to determine the ‘fingerprint’ of food, so that people can look at these notes and compose them together like creating music,” Bagler said. However, for this to happen, he needs people in the food industry, from chefs to people working in FMCG companies that make chips and soft drinks, to get involved and bring their expertise on board as well. IIIT Delhi now has an annual Symposium on Computational Gastronomy to bring people together, which is where HuffPost India spoke to Bagler and ITC’s Malhotra.

Francis Dean via Getty Images

This isn’t molecular gastronomy

When people talk about science and food, the trend of molecular gastronomy is the first thing that comes to most people’s minds, but there’s a very big difference between that, and what Bagler’s team is trying to do. Molecular gastronomy has come to stand for transforming food using chemical and physical reactions — foams, spheres, flash-freezing and low-temperature cooking as some of the well-known applications. It’s also considered passĂ©; speaking to the Hindustan Times, Chef Srijith Gopinathan, executive chef of Taj Campton Place, San Francisco, USA said: “Molecular gastronomy is gone now. Maybe in India, it will last for a few more years. No harm in creating new things but not for the sake of it.”

What Bagler hopes to accomplish is much more behind-the-scenes, using machine learning to help reduce food waste, or develop more sustainable alternatives for recipes, by finding local ingredients that match desired flavour profiles.

And while ITC’s Chef Malhotra is showcasing recipes that will make it to his Sunday brunch, one of the ways to make a big impact is to change the ingredients in a bag of chips, or a cola can, and Bagler is keen to work with the industry as well. “Aside from making food sustainable, we also have been exploring the nutritional and health effects of foods and want to make people’s diet more healthy,” said Bagler.

Jaspal Sabharwal, formerly VP of franchise business in India and South East Asia, Coca Cola, who is on the board of Burger King in India, also believes that modern technology can have a huge impact on food development. He’s now the co-founder of a company called TagTaste, which works with clients to get feedback on food products at scale, and he said that today, as millenials become parents, they want to know more about the food that they are consuming. “These are a very data-oriented group,” he said, “and they consume everything through digitisation.”

“Thanks to the research that Bagler is doing, there is now empirical data about food, and it’s going to be personalised down to the individual level, which will be very important for this generation,” he added.

Chefs won’t be out of a job yet

With over 1.5 lakh (150,000) recipes now analysed, FlavorDB covers a wide range of ingredients that’s still growing, and Bagler plans to increase the complexity of what is being developed, from simple pairs to complex combinations of ingredients to understand how recipes truly come together. His excitement over the project is infectious, but he’s the first to sound a note of caution as well, saying that this is just the beginning.

While Bagler’s team can already predict what pairs of food work well together, that doesn’t mean that they’re ready to ask a computer to create its own recipes yet — rather, it’s only removing an element of trial and error in testing pairings. Even that can fall short at times, and require chefs to experiment.

For example, chocolatier Richa Chaudhary explored FlavorDB and saw that the pairing of tea and chocolate is quite harmonious. “Normally when people think about chocolate, they think about coffee, so this was really interesting to me,” Chaudhary said. “But when I actually tried it out the results were pretty bad. I put tea inside a chocolate and it didn’t work at all. But interestingly, when you drink tea and eat chocolate the tastes come together very well. So it’s not always straightforward.”

ITC’s Malhotra also agreed that knowing how to prepare ingredients and how to use them mattered a lot to chefs, and beyond flavours, there were other qualities like astringency and umami that a chef needs to be aware of, which will affect how they create dishes, compared to a computer that will always stick to the recipe. “FlavorDB can say that a lemon pairs with this ingredient, but how to use the lemon for the best effect, that’s something only a chef can do,” he said.

Close
This article exists as part of the online archive for HuffPost India, which closed in 2020. Some features are no longer enabled. If you have questions or concerns about this article, please contact indiasupport@huffpost.com.