Janet Harris, a friend and Pollster reader, sends an
interesting bit of analysis she did on last night's MSNBC Democratic debate. Using the free site TagCrowd, she created a
set of "tag clouds" that
provides a visual depiction of the words used most often last night by each of
the candidates.
For those not familiar with the term, you have probably seen
tag clouds appearing on many web sites (and hopefully, very soon here on Pollster). They were apparently first implemented on
the photo sharing site Flickr, and typically provide a visual representation
of the most popular "tags" assigned to web pages. The type size of each word varies according to its frequency of usage. The larger the type size, the more often each candidate used that word.
Here are the clouds Janet created:
She also created a PDF version suitable for printing.
Now of course, this is a quick blog post, which probably
raises as many questions for me as it answers. Each of the clouds consists of
the 50 words used most often, omitting common words like "and," "of," "the,"
etc. I am not sure if the scale of the words is comparable across clouds -I suspect
that Professor Franklin will feel strongly that they should be. Finally, for what it's worth, Janet also
sends along this total word count for each candidate:
- 1,872- Senator Obama
- 1,766- Senator Clinton
- 1,518- Senator Edwards
- 1,281- Governor Richardson
- 1,180- Representative Kucinich
- 961 - SenatorBiden
- 912 - SenatorDodd
- 753 - SenatorGravel
A few quick observations, with an assist from Janet (who is
the president of the media analysis firm, Upstream
Analysis):
- Noticethe more frequent use of wonkier language by Chris Dodd, particularly theuse of "administration," "multinational," "stateless," etc.
- Nowcontrast that to John Edwards, whose answers tend to use everyday languageand deliver a message loud and clear message: "America," "believe," "united."
- Theone-issue emphasis of Kucinich and Gravel - "war" -- is obvious.
Obviously, this feature is a bit off-topic for a site
devoted to polling methodology, but it does deal with the graphic analysis of
political data. I can certainly see potential applications of this sort of
graphic for those that conduct and transcribe focus groups and other "qualitative"
analysis.
But enough wonkiness. Readers, what do you see in these
clouds? Our comment section is wide open...