Word clouds done right
Jacob Harris is absolutely right to hate word clouds. You take a long and complex text, and then you boil it down to a group of individual words, with the most-used words being the biggest? That’s just silly. “Reporters sidestepping their limited knowledge of the subject material by peering for patterns in a word cloud,” he says, is “like reading tea leaves at the bottom of a cup”. Word clouds are crude, inaccurate, misapplied, and place the onus of understanding onto the reader.
But there’s one place where word clouds are I think both useful and accurate — and that’s when a pollster has asked a group of people to say the one word they would use to describe X. Here, for instance, is the word cloud generated when a Reuters/Ipsos poll asked Republican voters for the first word that came to mind after watching the weird Herman Cain “smoking ad”:
And here’s the word cloud from the latest Kauffman poll of econobloggers:
Here the size of the words is interesting, but more germane is the overwhelming negativity of the vast majority of words used. There’s a couple of tiny good ones in there — “rebounding” us up by the Canadian border, and “bounceback” is in the Bay Area somewhere — but they’re in a distinct minority.
Incidentally, that Kauffman poll has some fascinating responses elsewhere, too. Check out the sudden enormous popularity of NGDP targeting:
There’s also a very high degree of skepticism when it comes to how good colleges are at teaching kids useful stuff.
The bar charts here are again an effective way of communicating information. Things like chart types and word clouds are tools, and you have to know which tools are best used in various different circumstances. And while word clouds are usually stupid, sometimes they can be exactly right.