AI Model Connects Consumer Ratings with White Wine Chemistry
Researchers have developed a machine-learning model linking wine ratings to chemical profiles.
Key Points
- • The AI model connects consumer wine ratings with chemical profiles.
- • Data from 64 wines and 145 chemical compounds were analyzed.
- • Certain chemical markers correlate with consumer perceptions of quality.
- • Future enhancements may include text analysis from user reviews.
A new machine-learning model developed by researchers in Denmark and Germany has successfully linked consumer ratings of white wines to their chemical profiles, transforming the approach to wine quality assessment. Published on July 18, 2025, this study utilized data from the Vivino app, which aggregates user ratings for wines, providing an extensive consumer perspective often overlooked in traditional evaluations.
The model was trained on a dataset of 89 white wines, 64 of which had corresponding ratings from Vivino. It examined 145 volatile organic compounds and analyzed numerous chemical properties, including ethanol, density, and various sugars and acids. Researchers found that specific esters were associated with lower ratings, whereas terpenoids, reducing sugars, and lactic acid correlated with higher scores.
These findings challenge the conventional subjective assessments commonly employed in the wine industry, which typically rely heavily on expert opinions. Moving forward, the researchers aim to enhance the model by integrating text analysis of consumer reviews and expert evaluations to refine the assessment process further. This approach promises a more objective and comprehensive understanding of factors influencing wine quality.