Posted by John Sides on January 20, 2009 01:59 PM|Permalink
Comments
John,
You’ve put these up before, and I admit they can be kind of cool to look at. But … what do they mean?
I ask in all seriousness, because textual analysis is taking off, and it should. An analysis of word frequency might be interesting, but I think font size is a clumsy way to represent it. Isn’t there a better way to represent these data?
That’s a very good question. I think Wordles (and their equivalents in the NY Times and elsewhere) are good mainly for very crude comparisons — e.g., Obama used such-and-such word more than Bush. (And crude comparisons are the stuff of blog posts.) Like you, I have doubts about font size. I don’t know that people process font size in a way that doesn’t distort the underlying data. Other kinds of descriptive representations would probably be easier to comprehend accurately. Another problem, of course, is that individual words only tell us so much about meaning. Better textual analysis would consider the context in which words appear.
Comments
John,
You’ve put these up before, and I admit they can be kind of cool to look at. But … what do they mean?
I ask in all seriousness, because textual analysis is taking off, and it should. An analysis of word frequency might be interesting, but I think font size is a clumsy way to represent it. Isn’t there a better way to represent these data?
Posted by: Hans Noel | January 20, 2009 08:28 PM
Hans,
That’s a very good question. I think Wordles (and their equivalents in the NY Times and elsewhere) are good mainly for very crude comparisons — e.g., Obama used such-and-such word more than Bush. (And crude comparisons are the stuff of blog posts.) Like you, I have doubts about font size. I don’t know that people process font size in a way that doesn’t distort the underlying data. Other kinds of descriptive representations would probably be easier to comprehend accurately. Another problem, of course, is that individual words only tell us so much about meaning. Better textual analysis would consider the context in which words appear.
Posted by: John Sides | January 20, 2009 09:18 PM
A follow-up: Junk Charts, a pretty discerning blog about the visual presentation of data, seems fond of tag clouds, which are similar to Wordles:
http://junkcharts.typepad.com/junk_charts/2005/10/tag_clouds_are_.html
Posted by: John Sides | January 20, 2009 09:24 PM
FYI, there are Wordles for that awful poem and Rev. Lowrey’s (IMO terrific) benediction up there as well.
Posted by: C. Zorn | January 21, 2009 10:45 AM
great application. wanted to try it out so I wordled your convo…
http://www.wordle.net/gallery/wrdl/471980/conversation
Posted by: some chick | January 25, 2009 10:42 AM