In this first installment of our interview series with the "Flavor Network" research group, Linda Chon interviewed Professor Tim Tangherlini, a computational folklorist in the department of Scandinavian. Professor Tangherlini leads an innovative project that uses data science to trace the evolution of Nordic flavors over the past 200 years. By analyzing historical cookbooks and flavor compounds, his team aims to uncover how Scandinavian cuisine has transformed and what it can reveal about social, cultural, and economic changes in the region. Forthcoming interviews with his students will follow this month.
What led to your interest in this project? Why the interest in flavor?
I thought YY Ahn's work on flavor networks was a very interesting approach to a systems-level understanding of cuisine. I'm very interested in aspects of everyday life and the things that influence us on a day to day level that we often don't even recognize as having such an important influence on us until they're taken away." Flavor networks became important at the end of the 18th century as autocratic empires started to crumble, leading to democracy movements and the broad adaptation of democracy in Scandinavia, as well as the ability of people to imagine things like cooking as more than just subsistence cooking. Things like transportation, reading, literacy, and the emergence of a broad middle class - groups of people not just scraping by - coincide with the development of food culture. We have no idea what we will find - Scandinavian cuisine might just be quite bland and marked by pickled herrings and meat based sauces. But more likely we will find that with the emergence of accessibility to global markets through shipping and trade, that things like different types of spices start to creep into the cooking and cuisine becomes less of a privilege of the aristocratic and upper classes and becomes something that people recognize as a shared part of the culture. Our goal is to try to understand what Scandinavia tasted like from the end of the 18th century to the mid 20th century. Are there moments when certain things change? It's a project that integrates folklore and the culture of everyday life, and data science. Can we leverage data science to understand how it tasted?
Could you briefly describe what a "flavor network" represents? How do you leverage computing to capture how different flavors interact?
Every ingredient in a recipe is made up of different molecules and those molecules acting individually or in concert with each other create flavor compounds. We can match ingredients to flavor compounds. Things like heating can change the interaction of ingredients to create different flavor compounds. One of the students on this project is a chemical engineer interested in flavor. For each ingredient we're going to extract or match the ingredients to flavor compounds, and for each step of the recipe we will look at the transformation of flavor compounds in that recipe. Professor YY Ahn's work examines the pairing of complementary flavors rather than contrasting flavors in Asian vs European cuisine. We're hoping to see something like that – some change over time – using data methods like change point detection. Change point is built into the system, as we're taking cookbooks from different periods that were in broad circulation and using them as a proxy. For each recipe, we'll be mapping ingredients to flavor compounds and looking at how the recipe combines different ingredients to create a flavor profile. We might have an overrepresentation of recipes. Essentially it's a list of flavors and how they relate to each other and which flavors seem to be most common or of the highest degree in the flavor network.
How does this project fit into your work on social networks and computational folkloristics?
It's all related to foodways. Foodways is an area of study in folklore that involves aspects of vernacular culture and its circulation across social networks. It's not using a social network model, but a flavor network model instead. It's very folkloristic in that sense, as it's focused on what people would be likely to eat and prepare as part of everyday culture. Networks are built into the analysis.
What are the difficulties of research in the experience of taste, given the subjectivity of that category? What kinds of primary sources do you use?
We're using flavor as a proxy for taste (there's a difference between flavor, which is made up of chemical compounds, and taste, which is more subjective) and using cookbooks as a proxy for what people were actually making. Since we're using lists of ingredients and mapping them to flavor compounds that we know from contemporary ingredients, there is no reason to think that they taste the same as they do today. So there are some really fundamental assumptions that we have had to make, and they are hard but solvable. It would be better to use a more systems level view of the overall flavor space than if you were to sit down with one cookbook and make some small number of recipes. We can work with all the recipes and ingredients and flavor molecules comprehensively because of the data science approach. There are other problems - where do we get the cookbooks, how much are Nordic publishers borrowing from other cookbooks? Is it an accurate representation of Norwegian or Danish or Swedish cooking? Norway was a possession of Denmark and Sweden until the Danish bankruptcy. Instead of becoming a free and independent nation, it became a protectorate of Sweden until the 20th century. So we get all the social, cultural, political history folded into this problem. It's an interesting problem to think about changes in Scandinavia through this approach - things like agricultural practices changing, the emergence of a middle class, food becoming associated with different levels of status, which might be addressed as downstream tasks. It wouldn't be possible without a data science approach because there are so many recipes and ingredients and ways of combining ingredients, which create different flavor molecules. And we also have a wonderful model from YY and his group.
I was able to take a brief look at Professor Yong-Yeol Ahn's work on flavor networks and food pairing, where he breaks down the different types of ingredient combinations that can be found in different regions of the world based on shared flavor compounds. Is your own project focused on a specific region(s) within Scandinavia?
No, our project is more on a regional level as opposed to global, Denmark, Sweden, Norway. We know how to work with these languages (as opposed to Finland and Iceland). One of the undergrads reads Scandinavian languages quite thoroughly, so they make a good target set. We want to come to an understanding of flavors in Scandinavia over a long period of time, as well as the rise of new Nordic cuisine in the past two or three decades. It's a throwback to when people are necessarily using local ingredients but with the intrusion of wares from colonies, particularly Denmark and Sweden bringing in spices, which is related to social and economic status.
If you're able to, can you share any progress you have made so far? What stage of the project are you currently at, and what direction do you hope to take your project going forward?
We're a small team and the undergrads are busy, but they're very dedicated. So far we have selected the cookbooks, parsed out the recipes, and we're using some large language models to convert the lists of ingredients from Scandinavian languages to English to map ingredients to a database of flavor molecules that we've also been able to acquire and adjust for this project. We're now working on extracting ways ingredients are combined in each recipe to figure out flavor molecules we need to map for each recipe. We have 8-12 cookbooks working with, there are some challenges with machine translation, but it does a good job when given some hints as to what we are actually looking for. Two students are working on fine tuning the large language model and retrieval augmented generation of our output that we're then going to map to the flavor molecule space. As of last week we're getting together a list of cooking terms used throughout these cookbooks that can be reduced to a small number of procedures or techniques that change the flavor compounds within the ingredients that are subjected to this technique.