Decoding Global Palates: Unveiling Cross-Cultural Flavor Preferences Through Online Recipes

Area of Science:

• Food Science
• Computational Gastronomy
• Cross-Cultural Studies

Background:

• Cross-cultural food choices are complex, influenced by cultural nuances and flavor preferences.
• Understanding these preferences at a molecular level can inform food choices in diverse contexts.

Purpose of the Study:

• To analyze cultural flavor differences and identify consistent patterns across China, the US, and Germany.
• To explore the potential for data-driven insights into cross-cultural flavor appreciation.

Main Methods:

• Mapped ingredients to flavor compounds using an authorized database.
• Applied machine learning to 75,000 online recipes (25,000 per culture) to classify by country.
• Utilized user interaction data (recipe ratings) to differentiate appreciated and less appreciated recipes.
• Performed exploratory analyses on ingredient networks and flavor compound clustering.

Main Results:

• Machine learning classifiers accurately distinguished recipes by country (77% accuracy).
• Models predicted recipe appreciation within cultures with up to 66% accuracy.
• Cross-cultural prediction showed modest accuracy (max 54%), suggesting shared flavor patterns between Chinese and US cuisines, with German preferences differing.
• Exploratory analyses supported machine learning findings, highlighting opposing sweet/savory trends between US and German users.

Conclusions:

• Molecular flavor analysis reveals significant cultural distinctions and potential shared preferences in food.
• Data-driven approaches can uncover nuanced cultural flavor profiles.
• Findings have implications for developing culturally sensitive food recommender systems.