Real Genius
Thursday, October 14, 2004
 
A Little Word Counting
Here's what I did...

First, I divided the answers into three types for each of the candidates:


-- The "answer" that was 120 seconds
-- The first rebuttal that was 90 seconds
-- The second rebuttal when there was one, that was 30 seconds


Then, I counted the number of words in each section (actually, Microsoft Word did).

We're talking about a matrix with 21 rows here --one for each question and the closing-- and 12 columns, that's not sparse, but has more empty cells than not. Empty cells exist for the obvious reason. A total of 400 words in two minutes is a good-enough ceiling value, so this is the normalization I made:


-- For two-minute answers, I simply divided the word count by 400
-- For 90-second rebuttals, I divided the word count by 3/4 first, then by 400
-- For 30-second rebuttals, I divided the word count by 100


Keep in mind that since this is simply normalizing the number of words spoken in a given amount of time, in essence it's comparing speed.

Then, I plotted the x-y pairings, for the abscissa spanning 1 through 21 and the ordinate containing values that vary between 0.1 and 1.1, roughly. And finally, I fitted a polynomial curve to each of the data sets, which, although not very reliable, is quite good in telling the untrained eye of trends.


Comparing Word Counts Posted by Hello

The normalized values that you see in the legend are all called mu, the letter "k" is for Senator Kerry and "b" is for President Bush. Finally, the numbers in the subscript denote the number of seconds of speech (interval) the data point represents.

As I saw, and so should you, Senator Kerry had a very solid, stable and low-variance profile in all three types of answers and managed to put in more words per minute than his opponent did at every answer except one. His profile shows a warm-up towards a modest peak, stays there and then gradually declines to its starting point. And notice how close the fitted curves also are.

Now look at President Bush's curves. Notice anything special? Aside from the fact that his values are rather erratic, higher-variance than his opponent's and lower in general, it's also quite visible that his main answers are in a steady decline in density whereas his first rebuttals are increasing.

What does that tell you about a man who has not much to say about his own record but sings like a bird when it comes to trashing his opponent?

What does it tell you about his national campaign?

Comments: Post a Comment

Powered by Blogger Weblog Commenting and Trackback by HaloScan.com