Wordle pops up alot in visualizations on the internet these days, partly because it’s simple to do and partly because it’s pretty. The bright colors, good choice of fonts, and effective display of word counts make it an interesting way to view large blocks of text. However, it really does nothing from a “visualization” perspective, as there’s no real information to gleam from it. Seth Grimes makes a good argument that word-counts alone are a pretty useless metric.
Wordle images are pretty but they may give a deceptive picture of word significance. The occurrence-counting approach is simply too simplistic. Author Jonathan Feinberg does writes in the Wordle FAQ that Wordle doesn’t stem words, that doing so would be beyond his goals, and there are other linguistic deficiencies. Wordle is nice effort despite the short-comings, but as it is, it pretty much implements a very shallow approach as presented by Luhn in his 1958 paper:
As an example of his argument, I fed the entire text of his article into Wordle to generate the image above. From that image, you can’t really gleam anything about the meaning of his article, other than it’s about Wordle and words, with something about physicians and chemistry in Chicago thrown in there. Very pretty? Yes. Very Useful? No.
via Limits of Visualization: Wordle Misses Meaning | The Intelligent Enterprise Blog.
