The streamgraph is more interesting in that it actually shows a time series of the data. You cannot show as many series as you can show words in the Wordle thing, and of course the streamgraph suffers from the problems of most stacked charts (no constant baseline particularly). Some of the labels are just too small to read; I think all labels should be a consistent font size, and let the thickness of the bands indicate the magnitudes.

Comparing streamgraphs to word clouds like Wordle would be an interesting exercise. I wonder which would be more effective and/or popular to the audience.

For those that aren’t familiar here are two examples

Example of word cloud using Wordle:
http://supportanalytics.com/blog/wp-content/uploads/2008/06/itunes1.jpg

Example of Streamgraph on Neoformix:
http://www.neoformix.com/2008/TwitterClientUsageStream.html

The streamgraphs are okay if you treat them as a mostly qualitative display. There is some thicker = more to the bands, but it’s very hard to quantify, because the bands undulate around a baseline. They’re better than all the funny word constructions, or cloud stuff, or sentence parsing, or all the things that look like a tangled ball of Christmas lights that someone plugged in anyway.

Jon – So I’m not alone. I’m not entirely on board with the streamgraphs yet. They are like Wordle in that I can see the large figures clearly, but the small ones are impossible to view unless it’s an interactive application.

Thanks for the comments!

I feel this way about a lot of visualizations. Often they seem more a vehicle for showing creative use of form and color than for showing relationships in data. I always look at the feeds for blogs like http://visualcomplexity.com, http://infosthetics.com, and others, and usually come away thinking “huh?”. http://neoformix.com sometimes strikes me the same way, although they lately have shown some interesting things with streamgraphs.