June 13, 2022
It was July 2021 when we decided to integrate our work on the prediction of vintage quality and prices with new applications of machine learning in the world of fine wines.
Pursuing our new plans, we came across the work of Professor Bernand Chen and of the research team of the University of Central Arkansas (UCA). In particular, we were attracted by an interesting paper called “Can wine reviews in Bordeaux reveal wine aging capability?”. K. Kwabla, F. Coulibaly, Y. Zhenis and Dr. Chen suggested that there were some correlations between the descriptors used by critics and the aging capacity of Bordeaux wines.
Time was running fast and the harvest in Europe was very close. As usual at this time, we were busy analysing the data about the growing season of the 2021 vintage, not only in Bordeaux (see the abstract of our Bordeaux 2021 report), but also in all the other regions we monitor, such as Burgundy.
The idea that the Saturnalia team came up with – which was enthusiastically shared by Dr. Chen – was to add a new element to the already existing database, made up of billions of satellite-derived data only for the Bordeaux area. The idea is to combine Saturnalia data about the vegetative growth and weather conditions with the extensive database of tasting notes collected and further elaborated through Dr Chen’s machine learning algorithms.
Wineinformatics, which incorporates data science and wine-related datasets, is the new data science research in wine. Dr. Chen is currently focusing on analysing wine reviews and finding the correlation with related topics, such as wine grade, vintage, price, terroir …etc. Dr. Chen developed the Computational Wine Wheel as the natural language processing tool for computers to understand the human language format reviews. With the help of the Computational Wine Wheel, region specific or research topics oriented wine reviews can be collected and processed as the clean dataset for data science research for answering the questions such as “What makes wine achieve a 90+ rating and be considered an outstanding wine?”, “What are the shared similarities amongst groups of wine?”, “What characteristics of classic (95+) wines show in 21 century Bordeaux?”, “Are wine reviewers reliable and consistent?”
Thanks to two of our proprietary indexes, the Saturnalia Evolution Index (SEI) and the Saturnalia Variation Index (SVI), we can leverage the reflectance derived through satellite sensors to measure the level of chlorophyll and water stress in the vines. Our AI-driven algorithms then elaborate our Vintage Score, a scoring of the potential quality of the wine. The peculiarity of the vintage scores is that there is no wine tasting involved, as it is based solely on objective sources which are skillfully mixed together, such as satellite data, weather data and orographic information related to the growing season.
The purpose of the collaboration is to leverage machine learning to evaluate the correlations between the taste profile, the vintages evolution and the optimal drinking window. The work of the two teams is proceeding and soon we will publish some sneak peeks on what seems to be a step forward in wine informatics and the fine wine industry.
© 2021 Ticinum Aerospace. All rights reserved