Table 1 Relation between recipe co-occurrence based pairing ranking and chemical overlap based ranking.

From: FlavorGraph: a large-scale food-chemical graph for generating food representations and recommending food pairings

Ingredient 1

Ingredient 2

Co-occur Prob.

Chemical overlap

Recipe co-occur. ranking

bamboo_shoot

water_chestnut

0.532

94

Parsnip

Turnip

0.524

101

Oregano

Basil

0.515

159

Raspberry

Blackberry

0.503

116

Nutmeg

Cinnamon

0.472

139

Miso

Sake

0.507

4

Anchovy

Caper

0.415

2

Saffron

basmati_rice

0.394

6

ham

swiss_cheese

0.349

2

tortilla_chip

Avocado

0.317

5

Chemical overlap ranking

Apple

Strawberry

0.021

189

Bean

green_bean

0.117

178

Banana

Apple

0.090

173

Pineapple

Apple

0.081

169

Apricot

Apple

0.069

169

Cocoa

Apple

− 0.178

167

Orange

mandarin_orange

0.054

166

Papaya

Apple

− 0.049

165

Ginger

Pepper

− 0.026

164

Lemon

mandarin_orange

− 0.066

164

  1. Co-occurrence probability is normalized point-wise mutual information (NPMI) between pairs of ingredients. NPMI ranges from − 1 to 1, with − 1 (never occurred together), 0 (independent of each other), and 1 (co-occurred perfectly).