Clouding the Debate
By Joshua Levy, 04/27/2007 - 4:52pm

Pollster.com reports that Pollster friend Janet Harris has compiled tag clouds (they're really "word frequency clouds") that visually depict the most-often used words each candidate used in last night's Democratic debate, and the results are perfect illustrations of how great these tag-cloud thingies are.

Each candidate gets a tag cloud, so it's easy to see what words they each used the most. There are some interesting surprises here, and it's funny to note how often a candidate uses words that he or she is pejoratively associated with.

For example, Joe Biden's most-used words are, "better," "able," and administration," but he also used the words "talk" and "talking" a lot -- which supports the popular conception of a chatty Senator from Delaware.

tagclouds_debate

Hillary Clinton's top words were "president," "ready," "health," and "care."

Chris Dodd: "years," administration."

John Edwards: "America," "believe," "united," "states"

Mike Gravel: "vain," "war," "Iraq."

Dennis Kucinich: "war," "president."

Barack Obama: "going," "sure," "women," "intelligence."

Bill Richardson: "deal," "health," "mental."


We made our own cloud for moderator Brian Williams:

williams_cloud

Visualizing Data

I love this and did it just a week or so ago to Bush's Iraq push back speech and Alberto Gonzales's testimony.

I tried it for the first time last year using Bush's speech commemorating the 5 year anniversary of 9-11.

I think the key is that "tag crowd" program allows you to insert hyperlinks for each tag. So if you were a really savvy campaign, you would click, let's say "terrorism" and it would take you to a page that said "my opponent says x,y,z about terrorism, but here is my position".

That would be a good policy, messaging use of a whiz bang visual technology.

What does it mean?

Normally, the value of a tag cloud is that readers choose the tags that are useful to them. Thus, the bigger ones are useful to more people.

These clouds are drvien by frequency of the candidates using words (I assume prepositions and definite articles were excluded). This approach does not allow anyone to choose tags based on utility, just frequency of use.

While it is interesting that some words were used frequently, most of the leading words offer minimal insight into the focus of the candidate.

What did jump out at me were coincidental sequences in the cloud, like Dodd's "Support terrorism today" and Gravel's "president respect (skip 1) soldiers."

Overall, this is curious with potential, but not yet analytically valuable.

Alan Rosenblatt
Executive Director, Internet Advocacy Center
AKA DrDigiPol (drdigipol.com)

not to be contrarian...

but I think that the value of the tag cloud is as a visual representation of a blogs taxonomy. It's not open to others to ad too, so much as show what the writers priorities are. I think that this could work the same way.

I would agree with you that, as currently used, it's interesting, but not valuable. That's why I think linking the tags to pages that introduce compare and contrast messaging is a way to take something that's visually engaging and making it actually informative.

The candidates don't have well established narratives yet. So it's a lot of wide ranging stuff. But if you took pre-November 2004 Bush speeches and clouded them, you would see 911, terrorism et all show up bright as day.

Or here's one for you, what if you took every administration members' speech in the run up to the Iraq war and clouded them, how many times would "smoking gun", "mushroom cloud" or "nuclear weapon" show up? They would be huge.

So when an administration tries to back track one something like that, here's a nice visual way to show folks in one fall swoop the words used to message and persuade. If people want to drill down a little further, link the words to a more built out history. It's like a language pie chart.

Not meant for hard analysis

These clouds aren't meant to be used for content analysis, but they do offer a "soft" analysis of each candidates' worldview, however hard it might be to quantify that. I certainly didn't mean to present them as evidence of the candidates' opinions or policies; they're more a fun way to draw out fuzzy impressions of the candidates, and see more clearly how each candidate presents him or herself.



© 2008 Personal Democracy Forum | All Rights Reserved |