Hu - Neural Networks and Nebbiolo: Artificial Intelligence for Wine

NEURAL NETWORKS AND NEBBIOLO
ARTIFICIAL INTELLIGENCE FOR WINE
SHENGLI HU

SHENGLIHU @AI-FOR-WINE.COM Find more interactive visualizations, demos, other topics, technical details, and more at http://ai-for-wine.com .

Table of Contents

2 Deductive Tasting

3 Theory Knowledge

10 Terrior

12 Wine Auction

14 Wine Investing

15 References

Introduction

Everything about wine appears intricate and complex, and mastering wine appears an ever daunting endeavor with the fast changing landscape of the worldwide wine industry, that encompasses a wide range of subjects such as geology, geography, viticulture, viniculture, chemistry, law, marketing, operations.
From the meticulous handling by experienced sommeliers of delicate aged bottles that have been perhaps scrutinized for provenance at the dinner table, to the selection of distribution and marketing channels possibly subject to the arguably unnecessarily complex three-tier system, from different regimes at bottling that might cause or prevent wine faults in years to come, to numerous experiments done at different stages of levage in the winery to finetune the final product, from intimate decisions on soil treatment, irrigation, vine training and trellising based on vintners experience, ideals, and terrior, to the processes and experiments revolving around scions and rootstocks in the nursery, it might strike as without doubt that there is perhaps little space for articial intelligence (AI) at its current state to take hold in the wine industry in the near future.
After all, the thought of ordering a bottle of wine with personal recommendations through a robot at a dinner table, or conversing with Alexa or Siri about the intricacies that make Musigny more different than Bonnes Mares than Les Amoureuses, devoid of any human touch of hospitality, would easily make one squirm.
On the other hand, articial intelligence has made breathtaking breakthroughs in multiple fields in the past decade, not only solving some of the worlds most pressing and challenging puzzles, but also penetrating various aspects of our daily lives. AI is making it easier for people to do things every day, whether its searching for photos of loved ones with a simple query, breaking down language barriers with smart online translators, typing emails with automatic completion, or getting things done with the Google Assistant. AI also provides new ways of looking at existing problems, from rethinking healthcare to advancing scientific discovery. One particularly relevant research theme that is quickly emerging is AI for Social Good, which uses and advances articial intelligence to address societal issues and improve the well-being of the world. In an excellent review article by the AI and Social Good Lab at Carnegie Mellon University summarized over one thousand relevant academic articles published in top computer science conferences in the following plots by application areas over time: With the steady (even exponential) growth of AI technology in various public domains, there is no reason why the wine industry, that overlaps with so many other industries wouldnt benet from AI technology. It is my strong conviction that various AI technologies can already resolve many issues of the wine industry in a surprisingly efcient manner, and I am going to show you how in this book in a fun and non-technical way that your parents would understand and hopefully agree.

1.1 Objectives of This Book

How could AI assist wine professionals in various aspects of the wine world, perhaps change the wine industry for the better, and ultimately enrich consumers experience? I try to illustrate by

examining the essential qualities and responsibilities of wine professionals or sommeliers through the lens of AI, and detailing how to use AI to help with relevant tasks;

identifying components of the wine industry where AI could potentially improve upon wine professionals, or make their job easier and more efcient at the very least;

solving challenging problems that have shaken the sommelier circles in recent years;
laying out future plans to building an ultimate sommelier-in-the-loop AI system for the wine industry.
1.2 The Structure of This Book

Chapter by chapter, I will discuss wine-related topics and the associated challenges, followed by to what extent could AI be of help, and in what ways, with introductions to the relevant topics in AI. I try to include visualizations and demos when possible to make it less dry. More interactive visualizations, technical versions of the AI parts, and other topics are available online at http://ai-for-wine.com .
This book can be read in at least three distinctive ways. First, each chapter provides visualizations (can be accessed online) that aggregate certain wine knowledge in a (hopefully) creative way. Second, every chapter addresses one or more of the wine-related challenges commonly faced by wine professionals and/or enthusiasts with a set of AI solutions, introduced in a relatively nontechnical way alongside visualizations, sample results, and demos, whenever applicable. Third, every chapter furthers the discussion, in subsections, of relevant AI methods with a self-contained review of method development and evolution over the past decade in the AI community where citations are included for scientic accuracy. Therefore, each chapter assumes in parallel two themes, one relevant to the wine industry, the other the AI industry. Yet both parts would be self-contained thus no previous knowledge is required to grasp the texts, except a curious mind and a playful heart. In addition, each chapter is in itself self-contained and can be read separately with pointers to other sections throughout the book when necessary, therefore readers are welcome to read in whatever order. Hopefully, this book makes a unique and novel addition to the wine literature and the AI literature, while being broad enough in scope to be of use across the wine profession, and perhaps inspire AI applications in other fields as well. Because of the fast-evolving nature of the AI technology (and the wine industry too), I hope to continuously update the chapters with the newest methods and introduce new topics, possibly in a second edition, a third edition, and so on.

1.3 A Preview of Chapters

Chapter discusses in-depth about what wine professionals and enthusiasts love (and hate): blind tasting. It has been an essential part of training for wine professionals. However, it does appears that everyone has his or her own unique marker or method, on top of the generally accepted socalled deductive tasting. I detail some of the many schools of thought about how to conduct deductive tasting, highlighting their major flaws and inconsistencies, while illustrating how this exact problem corresponds to some of the most classic machine learning methods, which in turn could be used to prevent pitfalls and identify the optimal strategy of deduction. Chapter gets into the weeds of the vast body of wine knowledge touching on various distinct yet intertwined subjects such as geology, geography, chemistry, viticulture, vinication, economics, etc. A solid grasp of a large body of wine knowledge is fundamental to being a qualified wine professional, just as how knowledge graphs are fundamental to various AI models and their generalizability and flexibility. We recount the important roles knowledge graphs have been playing in modern AI ecosystems, and illustrate with examples how knowledge graphs could be integrated to build question-answering systems like chatbox applications tailored to the wine industry.