Network analysis reveals culinary fingerprints that distinguish world cuisines
Researchers have developed a method to characterise and compare global cuisines by analysing how ingredient types are combined in recipes. The approach achieved 95% accuracy in identifying cuisines from recipe samples, offering food scientists new tools for understanding culinary traditions, developing culturally appropriate products, and creating data-driven approaches to recipe adaptation and innovation.
Understanding the underlying structure of different cuisines has practical implications for the food industry. Manufacturers developing products for international markets need to understand what makes each culinary tradition distinctive. This research provides a quantitative framework for identifying the essential ingredient combinations that define a cuisine, which could inform product development, recipe reformulation, and the creation of fusion foods that respect traditional flavour profiles.
The methodology could also support nutrition research by revealing how different cultures balance ingredient types, and assist in adapting traditional recipes for dietary restrictions while maintaining authentic character.
About the study
In the study – The networks of ingredient combinations as culinary fingerprints of world cuisines – published recently in npj Science of Food, an international research team analysed data from CulinaryDB, a repository containing 45,661 recipes across 23 cuisines spanning five continents. Rather than examining individual ingredients, the researchers grouped them into 20 broader categories including vegetables, spices, additives, herbs, meat, dairy and seafood.
Each cuisine was then represented as a network where nodes correspond to ingredient types and links describe how frequently pairs of types appear together in recipes. This approach captures not just what ingredients a cuisine uses, but how it combines them.
“Investigating how different ingredients are combined in popular dishes is crucial to uncover the principles behind food preferences,” the authors note. They found that cuisines differ “not only in the popularity of ingredient types and range of recipe sizes, but also in the structural organisation of ingredient-type combinations.”
Key findings
The network approach revealed clear structural differences between culinary traditions. European cuisines tend to distribute ingredients across different types and favour smaller recipe sizes. Asian cuisines prefer larger recipe sizes and often emphasise one dominant ingredient type.
“For example, European cuisines typically distribute ingredients across different types, whereas certain Asian and South American traditions emphasize one dominant type, such as vegetables or spices,” the authors explain.
Indian cuisine presented the most distinctive profile, appearing as a star-shaped network centred on spices. The researchers note this configuration “effectively captures the predominant centrality of Spices in Indian cuisine, which are an essential component to most of its recipes.”
Italian cuisine stood out for the central role played by herbs alongside spices, a feature shared only with Thai cuisine.
The study also found an exponential relationship between the number of recipes within a cuisine and the number of ingredients used. This suggests that culinary creativity typically builds upon established combinations rather than exploring entirely novel pairings.
Machine learning validation
To test whether these networks truly capture culinary identity, the team trained machine learning classifiers to identify cuisines from recipe samples. The full network representation achieved 95% classification accuracy, substantially outperforming a simpler approach based only on ingredient frequencies, which reached 79% accuracy.
Even simplified network representations retaining only the most significant connections achieved 87% accuracy, demonstrating that the essential character of a cuisine can be captured with relatively few key ingredient pairings.
Cultural and historical patterns
When the researchers clustered cuisines based on their network similarities, the results reflected both geographical proximity and historical connections. Asian cuisines formed a coherent group further subdivided by latitude. Indian cuisine stood alone, confirming its distinctive approach to ingredient combinations.
Spanish and Portuguese cuisines grouped with Latin American countries, reflecting former colonial relationships. The authors suggest that the presence of North African and Middle Eastern cuisines near Iberian traditions may reflect “the long-lasting effects of medieval Islamic colonization of Southern Europe.”
Limitations and future directions
The researchers acknowledge that their analysis relied on a single data source and examined cuisines at a single point in time. Future work could track how culinary traditions change over time and incorporate information about food processing methods.
“More broadly, our study offers novel insights into the structure of world cuisines, enabling data-driven approaches to their characterization, cross-cultural comparison, and potential adaptation,” the authors conclude.
Reference
Caprioli, C., Kulkarni, S., Battiston, F., et. al. (2025). The networks of ingredient combinations as culinary fingerprints of world cuisines. npj Science of Food, 9, 242. https://doi.org/10.1038/s41538-025-00588-4
