LING001 -- Homework 6

(Due 11/05/2007)

Apply any two of the five types of analysis listed below to the debate performances of at least two of the candidates in the primaries for the 2008 presidential election.

Whatever analyses you do, try to use them to draw some conclusions about the candidates and their personalities, ideas, and rhetorical self-presentation.

You can find links to transcripts and video or audio recordings in the Wikipedia pages for the "Republican presidential debates, 2008" and the "Democratic presidential debates, 2008"; or you can search for them elsewhere on the web, to find e.g. the New York Times transcripts of the Oct. 9 Republican debate in Dearborn MI, or the Sept. 26 Democratic debate in Hanover NH.

For each candidate that you choose, analyze their entire contribution to at least one debate -- depending on the kind of analysis you're doing, and the depth of your interest, you may choose to analyze several debates.

1. Use of metaphorical language. Try to relate a number of particular examples to a larger metaphorical scheme, perhaps in the general style discussed here and here.

2. Sentence length and depth of syntactic embedding. See here and here for some sample analyses in this style.

3. Use of disfluencies (filled pauses such as um and uh, or notable silent pauses; repetitions such as "I see no reason to believe we're headed for -- (pause) -- for economic downturn"; self-corrections such as "I mean, the free market is our -- one of our greatest assets"). At what rate do disfluencies occur? What is the effect on the candidate's message?

4. The dimensions of James Pennebaker's "LIWC" analysis. (Since this has not been covered in class, if you choose this option, you'll need to read at least the basic online description, and perhaps some of the papers written about it. Note that a free online version is available -- we are not asking you to buy the program.)

5. Some word- or phrase-frequency calculations of your own, whose results you think are interesting. Feel free to write your own software for this, if you know how to hack perl or NLTK or some other suitable language or package -- otherwise, you can use free pre-packaged software like TextSTAT. (As one trivial example, comparing the Oct. 9 Republican debate to the Sept. 26 Democratic debate, the Republicans used the word "no" at an overall rate of about 2.25 per thousand words, whereas the Democratic "no" rate was about 3.63 per thousand words.)

[Note that the on-line transcripts generally do not reproduce disfluencies accurately, so if you choose to do analysis type 3, you should correct the transcripts by reference to the recordings.]

Feel free to improvise on these general ideas, as long as the results are a) reasonably objective, factually correct, and well documented, and b) interesting.