First: even with a two-month hiatus from adding content to this page, daily traffic has been higher than I expected.
Second: Richardson's State of the State address is finished. His speech has two distinct peaks: one in his 'Year of the Child' section, and another in 'Moving New Mexico Forward'. On the whole, though, Richardson appears to have given a generally communitarian address. Details to follow soon.
I also decided to test (Current Governor of Texas) Rick Perry's last State of the State address. I was surprised: Perry's speech yields an egalitarian with a strong libertarian leanings in three out of four measures. The fourth measure shows a libertarian with some egalitarian tendencies.
Third: I have two preliminary scores for President Bush's Fifth State of the Union address. One is a handcount gathered as he delivered the address, while the other is an automated frequency measurement of specific indicators predefined in four distinct test tables.
They conflict.
This difficulty reveals a likely limitation in the dictionary file. The handcount yielded a moderate conservative with strong libertarian leanings, which is what I expected. However, the transcript of the address found a completely different score. Three out of four measures gave me an egalitarian score, while the fourth gave me a left libertarian score.
The disadvantages of a dictionary file or automated test table are clear: dictionary files are blind to context as well as to grammatical transformations not included in the list. Without a systematic means of electing terms to the dictionary file, many items are therefore left out.
Manual counts, however, have other disadvantages. Although context can be captured with a manual count, the process is far more cumbersome--one might even brand such demands a great deal more time and effort, and the observer's personal bias will taint the sample.
Here is a possible means to correct it: Constructing any test table for the LEO test should begin with a frequency distribution of substantive words and phrases from known ideologues. Those words and phrases with the highest frequency become test indicators of that particular ideological preference. The terms I am currently using in the dictionary files are based on their proximity to particular ideas advanced by known ideologues. I believe several terms are missing from the test tables I am currently using.
That will take quite some time.
Comments