I dig seeing the different visualizations that come out following big national speeches.
Here's a really cool network analysis of last night's State of the Union address (via many Twitter links). Built by Nodus Labs (who did similar work on past inaugural speeches), it's a graphical network display of the speech. Larger nodes represent words that appeared most in the text and nodes grouped by color were most often clustered together (in a sentence/phrase).
Moving the mouse around a little, it's not hard to see that most words connect back to "jobs" and "education". Also interesting is that the "gun control" theme got a lot of talk in the press today but words related to it are really hard to find. Not surprising, I guess, that with a hot-button topic even a few words make a big impact.
In past years, these were often done as tag clouds but the interactive approach of this network is much cooler. It encourages more exploration and paints a better picture of the speech's "themes." Hovering in areas of concentrated color shows that, for example, the purple (right-ish) is focused on deficit reduction, while the magenta (center-ish) was all about creating and protecting jobs in the US.
Visualization as Writing Tool?
This has me wondering... most visualization focuses on analyzing something that's already done. Today's prices or volume, operational data, or - in this case - spoken words. But is it (or will it soon be) common to use tools like this in advance?
If you're a speechwriter, do you already have a stable of tools like this to craft a speech before it's given? To make (mathematically) sure the words that are spoken most are "on message"? Is a computer scientist in the room now, working alongside the editor, policy wonk, and fact-checker?