Flavor Wheel Development from a Machine Learning Perspective

Area of Science:

• * Food Chemistry and Sensory Science
• * Application of Machine Learning in Beverage Analysis

Background:

• * The relationship between chemical composition and sensory perception is crucial for distilled spirits.
• * Understanding these links impacts spirit production, quality control, and consumer appreciation.
• * Existing knowledge requires deeper investigation into specific flavor and aroma profiles.

Purpose of the Study:

• * To investigate the complex relationships between chemical compounds and aroma descriptors in seven categories of distilled spirits.
• * To develop a data-driven aroma wheel for enhanced understanding and appreciation of spirit profiles.
• * To leverage machine learning for analyzing large-scale chemical and sensory data.

Main Methods:

• * Analysis of a dataset comprising 3051 chemical compounds and associated aroma descriptors.
• * Application of Principal Component Analysis (PCA) for dimensionality reduction.
• * Utilizing clustering machine learning models to identify descriptor clusters per spirit category.

Main Results:

• * Distinct clusters of aroma descriptors were identified for each of the seven spirit categories (Baijiu, cachaça, gin, mezcal, rum, tequila, whisk(e)y).
• * Machine learning effectively mapped chemical compounds to specific sensory attributes.
• * Development of a comprehensive aroma wheel based on the analyzed data.

Conclusions:

• * The study successfully established a link between chemical compounds and aroma profiles in distilled spirits.
• * The developed aroma wheel serves as a valuable tool for industry professionals and consumers.
• * Machine learning provides powerful insights into the sensory characteristics of complex beverages.