The Recipe Similarity Network
Michele Bellingeri, Axel Bidon-Chanal Badia, Marta Vila Rigat, Roberto Alfieri, Massimiliano Turchetto, Davide Cassi
Scientific Reports · DOI
Read original publicationNetwork science meets Catalan cuisine — a normalized similarity measure and clique-based community detection reveal the structural DNA of traditional and haute cuisine cookbooks, identifying Allioli and Joan Roca's signature Becada as characterizing recipes.
These are our reading notes and analysis. The original work belongs to its authors and publisher.
Approach
Three Catalan cookbooks analyzed through intersection graph theory: two traditional (622 + 100 recipes) and one haute cuisine from El Celler de Can Roca (57 recipes). A novel normalized similarity measure avoids size bias, and a clique-based community detection algorithm uncovers recipe families.
Key Findings
Traditional vs. Haute Cuisine: Traditional recipe networks are densely connected (connectance 0.82-0.96), while Roca's haute cuisine network fragments faster under increasing similarity thresholds — reflecting the specialized, unique nature of innovative cooking.
Characterizing Recipes: Network centrality independently identified Allioli (garlic, olive oil, salt) as the most central traditional Catalan recipe, and Becada amb brioix del seu salmís as Roca's most representative dish — which Joan Roca himself has called his signature. This validates the methodology.
Recipe Communities: A 7-clique at high similarity threshold reveals recipes sharing garlic, oil, salt as a core, with complementary proteins and starches creating diverse dishes from a common base — explaining how culinary traditions generate variety from shared fundamentals.
Universal Ingredients
Across all three cookbooks: salt, olive oil, water, onion, garlic form the backbone. Roca adds modernist ingredients (xanthan gum, agar-agar, truffle).