All posts filed under: Machine Learning

Brexit in a glass

One week on… what happened? Through an exercise in good taste, wine may help us to understand Brexit related attitudes: An Isomap reduction of Brexit related wine-tasting vocabulary shows power orbited by seductiveness, pit, cut, hole and loose. This display is framed laterally by a taste for what is expensive and complex. Away from power it gets sickly, disjointed, sour. Note in passing that the right looks exciting and energetic to some. Complexity settles and sinks, the top is overwhelmed. (If you’re new to this series, the introduction may be useful.) When viewed through a PCA lens, power lies closer to austerity. In both pictures, there’s some bitterness over access. Whatever the reasons May be, there appears to be a seductive connection between  intrusive  and  borderline. Flavours of coffee and alcohol conjoin to cover an unpleasant situation. We form natural associations between words and these associations crop up in wine-tasting notes and in politics. Good taste is a transferable skill. I suspect these word-maps become increasingly easy to understand as you work your way through a borderline English white or an edgy continental red. Cheers!   MK@WineQuant   Advertisements

Fruity

And the orange squeezed into the water seemed to yield to me, as I drank, the secret life of its ripening growth, its beneficent action upon certain states of that human body which belongs to so different a kingdom, its powerlessness to make that body live, but on the other hand the process of irrigation by which it was able to benefit it, a hundred mysteries concealed by the fruit from my senses, but not from my intellect. Marcel Proust, Sodome et Gomorrhe, trans. Moncrieff.   Before there is wine, there is fruit* and juice. Before we learn to appreciate wine and cider, we instinctively enjoy the taste of grapes and apple juice, just like we enjoy beer before we move on to whiskey.   In this picture of clustered fruit descriptors, we see three interesting clusters. On the bottom left we have the citrus fruit moving up into riper fruit. Some of the less frequent descriptors stand out on the side, like mirabelle, lichee and banana and fig. Then we have a berry cluster which moves up …

Chocolate and Fennel?

This is Models vs. Reality, Part 2. I looked at what our tool has to say about chocolate as a wine descriptor and I was a bit surprised by the result. Top 10 chocolate related descriptors in critic wine notes cocoa coffee pepper tar mocha licorice truffle fennel vanilla espresso I would never have made an association between chocolate and fennel. It turns out that only a few writers make this connection. You can verify this yourself. For example on jancisrobinson.com you find that there are only six tasting notes including fennel and chocolate. Out of these, five are written by Tamlyn Currin and one is by Richard Hemming. None of these notes are in our dataset, so it’s a good test to look at them in a bit more detail. Chocolate and fennel, really? Richard Hemming’s description of an Alvaro Palacios Priorat (65% Carignan) features “leather and fennel aromas” with “chocolate-covered cherry”. Tamlyn Currin describes a “delicious mouthful of sweet red pepper and dried herbs, liquorice and chocolate” followed by a long finish with “a thrill of fennel and aniseed” in a …

Models vs. Reality, Part 1

Wine appreciation requires language. But the way you use language depends on what you consider to be a “good tasting note.” What is good? What’s the norm? …writing is a learned activity, no different in that regard from hitting a golf ball or playing the piano. Yes, some people naturally do it better than others. But apart from a few atypical autodidacts (who exist in all disciplines), there’s no practical way to learn to write, hit a golf ball, or play the piano without guidance on many points, large and small. And everyone, even the autodidact, requires considerable effort and practice in learning the norms. The norms are important even to those who ultimately break them to good effect. Bryan A. Garner, Garner’s Modern American Usage (2009, p. 104) Famous critics and formal tasting systems provide models/norms/reference points. But how good are those norms? What does “green apple, citrus peel, medium+ acidity” mean, exactly? Models are useful, but only if we don’t lose touch with what is actually going on. So let’s calibrate our models to reality. …

Tasting Algos, Spring I.

You shall know a word by the company it keeps (Firth, J. R.) It’s reasonable to believe that wine words which occur in similar contexts have related meanings for the tasters who wrote them down. Join us on our machine assisted journey to explore the unplumbed depths of wine writing with natural language processing (NLP) techniques. As machine learning meets human taste, let’s strive to make the symbiosis fruitful: with NLP algorithms and computing power at our fingertips, here’s an opportunity to discover more about how we record sensory impressions with words. We’ve trained neural network models on a large corpus of wine notes to recognize words by the company they keep. The results are fascinating. It’s spring-time, so let’s start by investigating a couple of fresh, uplifting, seasonal descriptors that come to mind naturally. Each word comes with its top 10 contextual associates picked out by our machine assistant.  The score (cosine similarity) indicates how close the association is. Blossoms ‘peaches’, 0.64 ‘wax’, 0.64 ‘poached’, 0.62, ‘bursts’, 0.61 ‘hazelnuts’, 0.61 ‘apples’, 0.61 ‘buttered’, 0.60 ‘apricots’, 0.60 ‘candle’, …