What’s In Your Cup Of Coffee?

AI meets tasting notes

Quilt.AI
Quilt.AI
3 min readMar 1, 2019

--

Like many of you, I love coffee. A million years ago, I worked on a coffee plantation for a brief period of time. Fast forward, two decades, I find myself at Ronin — this epic, dark, way-too-cool-for-me craft coffee shop down the street from our office. “Espresso Macchiato back to back please with 5 minutes between them” is my order of choice but sometimes it’s a piccolo (or a gibraltar or a cortado by any other name). And then nothing (almost nothing) can go wrong with the day.

Craft coffee of the third coffee wave over the last few years has taken the second wave (driven by Starbucks) and made it about style and story. Above all, today, we seek intentionality of experiences — rich experiences that have deep emotional resonance. Heightened sensorials derived from different coffee tasting notes enable vivid mental transportation of the drinker to a different landscape. Chermelle Edwards, who runs the webzine The Coffeetographer, perfectly captures this in her memory of a good cup of coffee:

I’ve tasted a purely brewed pour-over that made me feel like I was in the American midwest, in a rocking chair, with notes of blueberries and glazed cantaloupe.

Can AI Track Sensorials?

To explore how AI can quantify and bring meaning to sensorial products, building on the base of such a sensory lexicon, we decided to experiment with digital coffee data.

Coffee taste scores were gathered using the cupping scores from the Q tasting coffee database, where they are ranked using the overall cupping score of regional coffees.

Focusing on the best coffees of the major coffee producing areas, we chose to deep dive into Brazil, Ethiopia, Columbia, and Indonesia. Regionally specific coffee reviews were then found on social media using hashtags, and the taste mentions were extracted and analyzed by an AI NLP process that weighed keywords that were used. This quantification of taste data was then overlaid on a coffee tasting wheel and quantified.

These consumer-generated taste perceptions were then mapped on a taste wheel, provided by the Specialty Coffee Associations.

Lastly, just for fun, the taste parameters of a consistent 3-in-1 Instant Coffee were taken from reviews, and mapped against the user-generated parameters as well as professional-generated parameters. Here are how professionals and laymen (that’s you and me) judge coffee.

And I will leave you with three interesting facts:

  1. There is no difference in how consumer reviews rate coffees vs how professionals review coffees.
  2. 3-in-one coffee tastes like “cardboard” but that is what constitutes “coffee-ness” in that price band.
  3. Indonesian coffee and I get along the best — they’re nutty and best for piccolos!

Read more stories on Quilt.AI.

--

--

Quilt.AI
Quilt.AI

We are a culturally rooted, AI powered insights firm that converts millions of data signals into human understanding. Visit us: https://quilt.ai