Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

WineGraph: A Graph Representation For Food-Wine Pairing

Zuzanna Gawrysiak, Agata Żywot, Agnieszka Ławrynowicz

TL;DR

This work tackles the scarcity of structured food–wine pairing data by introducing WineGraph, an extended FlavorGraph that integrates wine data into a heterogeneous graph to enable taste-based pairings guided by sommelier-defined rules. It builds Aroma descriptors from large-scale food and wine reviews using tokenization, 1–3-grams, UC Davis wine wheel mappings, and 300-dim word2vec/TF-IDF embeddings to form aroma vectors for foods and wines. A rule-based pairing procedure and dynamic graph integration via metapath2vec create top-k wine pairings and embed them in a 300-dimensional space, enabling enriched food–wine recommendations. Evaluation with Normalised Mutual Information shows that adding wine data does not degrade clustering quality and can improve it, demonstrated via burrito pairing examples and cluster visualizations. The work delivers a neural-symbolic framework for joint representation of food and wine in a graph, suggesting paths toward richer culinary knowledge graphs and more precise pairings.

Abstract

We present WineGraph, an extended version of FlavorGraph, a heterogeneous graph incorporating wine data into its structure. This integration enables food-wine pairing based on taste and sommelier-defined rules. Leveraging a food dataset comprising 500,000 reviews and a wine reviews dataset with over 130,000 entries, we computed taste descriptors for both food and wine. This information was then utilised to pair food items with wine and augment FlavorGraph with additional data. The results demonstrate the potential of heterogeneous graphs to acquire supplementary information, proving beneficial for wine pairing.

WineGraph: A Graph Representation For Food-Wine Pairing

TL;DR

This work tackles the scarcity of structured food–wine pairing data by introducing WineGraph, an extended FlavorGraph that integrates wine data into a heterogeneous graph to enable taste-based pairings guided by sommelier-defined rules. It builds Aroma descriptors from large-scale food and wine reviews using tokenization, 1–3-grams, UC Davis wine wheel mappings, and 300-dim word2vec/TF-IDF embeddings to form aroma vectors for foods and wines. A rule-based pairing procedure and dynamic graph integration via metapath2vec create top-k wine pairings and embed them in a 300-dimensional space, enabling enriched food–wine recommendations. Evaluation with Normalised Mutual Information shows that adding wine data does not degrade clustering quality and can improve it, demonstrated via burrito pairing examples and cluster visualizations. The work delivers a neural-symbolic framework for joint representation of food and wine in a graph, suggesting paths toward richer culinary knowledge graphs and more precise pairings.

Abstract

We present WineGraph, an extended version of FlavorGraph, a heterogeneous graph incorporating wine data into its structure. This integration enables food-wine pairing based on taste and sommelier-defined rules. Leveraging a food dataset comprising 500,000 reviews and a wine reviews dataset with over 130,000 entries, we computed taste descriptors for both food and wine. This information was then utilised to pair food items with wine and augment FlavorGraph with additional data. The results demonstrate the potential of heterogeneous graphs to acquire supplementary information, proving beneficial for wine pairing.
Paper Structure (8 sections, 2 equations, 4 figures, 5 tables)

This paper contains 8 sections, 2 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Materials and Methods
  4. Data Preparation
  5. Pairing Procedure
  6. Incorporating Pairing Into Knowledge Graph
  7. Experimental Results
  8. Conclusions

Figures (4)

  • Figure 1: Visualization of the WineGraph using t-SNE projection.
  • Figure 2: Graph embedding with metapath2vec on WineGraph. Random walks traverse through various paths and gather nodes of different types (sample paths are shown in the left part of the figure).
  • Figure 3: Flavour profiles for wine pairing generated for burrito + guacamole
  • Figure 4: Sample clusters.

Theorems & Definitions (2)

  • definition thmcounterdefinition
  • definition thmcounterdefinition