What does this research mean for the field?

A combined framework of aroma kit-assisted rapid sensory methods, electronic sensors, and K-nearest neighbors (KNN) machine learning can perfectly distinguish the sensory profiles of French and Chinese Marselan wines. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

June 1, 2026Applied Food Research0 citationsOpen Access

Leveraging Aroma Kits and Machine Learning: A Novel Rapid Sensory and Electronic Profiling of French and Chinese Marselan Wines

View Full Paper

SWShiqin WangBeijing Normal University - Hong Kong Baptist University United International College YCYuehua ChenBGI Group (China)HDHuanlin DuBeijing Normal University - Hong Kong Baptist University United International College

Key Points

This research aims to characterize the sensory profile differences between French and Chinese Marselan wines using advanced analytical methods.
Selected seven commercial Marselan wines from China and France for analysis.
Evaluated odor and flavor characteristics using aroma kits, electronic sensors, and machine learning methods.
Applied techniques like Check-All-That-Apply, Polarized Projective Mapping, and electronic nose and tongue.
CATA and aroma kit showed strong similarity with PPM-UFP (RV = 0.895), confirming effectiveness in profiling.
KNN-based methods achieved perfect classification accuracy (1.0000) in distinguishing French and Chinese wines.
French Marselan wines exhibited fewer animal-related attributes than Chinese wines, reflecting regional variations.

Abstract

Marselan is a cultivar experiencing rapid growth in China's wine industry, yet the sensory profile differences between French and Chinese Marselan wines remain poorly characterized. To address this, seven commercial Marselan wines were selected for analysis, comprising five wines from three representative Chinese regions (Ningxia/CN, Shanxi/CS, Xinjiang/CX) and two from France. Odor characteristics were evaluated using Check-All-That-Apply (CATA) with the aid of an aroma kit (Le Nez du Vin, France), Polarized Projective Mapping with Ultra-Flash Profiling (PPM-UFP), and electronic nose (E-nose) analysis. Flavor characteristics were assessed using PPM-UFP and electronic tongue (E-tongue) techniques. Results demonstrated strong similarity between CATA with the aroma kit and PPM-UFP (RV = 0.895). Furthermore, CATA showed substantial agreement with E-nose analysis (RV = 0.753), while PPM-UFP exhibited even higher similarity with both E-nose (RV = 0.826) and E-tongue (RV = 0.761), confirming the effectiveness of both modified methodologies. In E-nose data analysis, K-nearest neighbors (KNN)-based methods KNN, principal component analysis (PCA)-KNN, and linear discriminant analysis (LDA)-KNN achieved perfect classification accuracy (1.0000), significantly outperforming support vector machine (SVM)-based (0.619-0.881) and back-propagation neural network (BPNN)-based (0.1429-0.7143) approaches. All four analytical methods successfully distinguished French and Chinese Marselan wines. French Marselan wines exhibited fewer animal-related attributes compared to Chinese ones, differences that may reflect regional variations in climate and winemaking practices. The findings support that aroma kit-assisted rapid sensory methods with electronic sensors and machine learning provide a robust framework for wine sensory profiling. This study provides valuable insights for quality improvement of Chinese Marselan wines and demonstrates the considerable potential of aroma kits in enhancing rapid descriptive sensory analysis.

AI에게 질문

Bookmark

View Full Paper

Cite This Study

Wang et al. (Fri,) studied this question.

synapsesocial.com/papers/6a1d224302fbce9130638037 https://doi.org/https://doi.org/10.1016/j.afres.2026.102227

AI에게 질문

Bookmark

View Full Paper