1. Original Entry + Comments2. Write a Comment3. Preview Comment


January 25, 2019  |  Friday words #157  |  2755 hit(s)

I run across a lot of new-to-me technical terms at work, but they're usually too technical to be generally interesting. However, I recently ran across the term gray sheep, which I thought others might find as amusing as I do.

There is some background, but bear with me. When you go to Netflix or Amazon, they have many suggestions about what you should watch or buy next. One way to create recommendation systems, as they're called, is to use collaborative filtering. Suppose you've watched the movies Clueless, Dirty Harry, and Frozen. Another user who also watched those movies then watches The Matrix. Collaborative filtering leads the system to suggest The Matrix to you, too. (Needless to say, the system is a bit more complex than this.)

Basically, collaborative filtering systems assume that people who have shown similar tastes in the past will show similar tastes in the future. But then there are the gray sheep. You loved Clueless, Dirty Harry, and Frozen, but you hate The Matrix. One of the articles (downloadable .gz file) on collaborative filtering describes gray sheep as users who are problematic "because their opinions do not consistently agree or disagree with any group of people."

In this context, black sheep are people whose tastes can't be predicted at all. (They're a failure condition for recommendation systems.) I like how they borrowed the idea of a black sheep as an outlier, and then modified it as gray sheep for users who are kinda-sorta the same as other people, but not quite. It feels like a term that could have wider application than in this narrow realm of predictive systems. In some senses, we're all gray sheep, eh? (One maybe doesn't want to think too hard about the idea that per this terminology, all users are sheep.)

For surprising origins, today we have anthrax. (I forget where I found this. Twitter, probably.) Today anthrax is the name of a disease, but in its earliest use in English it referred to "carbuncle, pustule"—in other words, to the outward signs of a disease. Latin had the word anthrax in this second sense also; they got it from Greek.

Here's the interesting part: in Greek, the word was used for carbuncles also, but that meaning apparently was a simile; anthrax in Greek referred to a dark precious stone and to charcoal. As the OED says, the ancient Greek anthrax "denoted burning charcoal, hence the use of the word for things red in colour (e.g. precious stone, carbuncle) as well as for things black in colour (e.g. coal)." Do you see coal there? One type of coals is anthracite, same root. So a disease of animal husbandry is improbably connected to burning coals. You never know how words will wander.

Like this? Read all the Friday words.