View source for Talk:Linguistic Analysis of Candidates

This is an interesting concept.  I see a potential difficulty, though.  There are certain concepts which I really don't think you can omit entirely when evaluating conservatism (the right to life, for example, and religious freedom.)  However, the trick is in determining which key phrases are most likely to be employed, and thus turn up in searches.  After all, a mention of "abortion" in a speech isn't a reliable indicator of conservatism--a rabidly pro-abortion candidate might use the word, as well.

How best to evaluate conservative leanings on these very important issues?  --[[User:Benp|Benp]] 21:22, 12 January 2009 (EST)

: I don't fully follow your criticism.  "Pro-life" is part of the test, for example.  Perhaps you'd like to look at the first analysis and reconsider or rephrase your criticism.  Thanks.--[[User:Aschlafly|Andy Schlafly]] 21:48, 12 January 2009 (EST)

::Not so much a criticism as musing on the most efficient way to go about it, Andy.  I don't think anything you've already included needs to change; I just wonder if other terms are also warranted.  Pro-life is there (and I did, indeed, overlook that the first time; thank you for pointing it out!) Pro-life is certainly a good term, and one that's only likely to be used by conservatives, but is it sufficient?  Would it be worthwhile to include terms like "right to life" as well?  What about terms that accurately describe the abortion industry, but which liberals avoid like the plague: "pro-abortion," for instance, or "abortion on demand?"

::I suppose what I'm trying to figure out here is where the optimal balance falls between being thorough and getting bogged down in too many redundant terms.  Rather than add terms wholesale, I thought I'd seek guidance here first.  --[[User:Benp|Benp]] 22:20, 12 January 2009 (EST)

::: If time were unlimited, the list could grow in an unlimited manner.  But then priority is lost.  Twenty (20) seems to balance breadth with priority, and I'd be reluctant to expand the list.  But feel free to make specific suggestions here.  So far, I don't see how to improve on the term "pro-life".  The terms "pro-abortion" and "abortion on demand" seem redundant at best.  Overlap between terms is to be avoided.--[[User:Aschlafly|Andy Schlafly]] 22:55, 12 January 2009 (EST)

::::I do see your point.  I guess the only question I still have, then, is about religious freedom and school prayer; I don't see any terms on the list that directly touch on these issues, which I would certainly consider important to most conservatives.  --[[User:Benp|Benp]] 16:35, 13 January 2009 (EST)

Shouldn't this be an essay? [[User:Commodore Guff|KevinS]] 21:50, 12 January 2009 (EST)

: No, I think the concept is encyclopedic and susceptible to objective analysis.  Encyclopedias include analysis of novels, so why not linguistic analysis of candidates?--[[User:Aschlafly|Andy Schlafly]] 22:55, 12 January 2009 (EST)
:: As I understand it, though, such analysis comes from other sources.  It's not a huge issue, anyways. [[User:KevinS|KevinS]] 16:45, 13 January 2009 (EST)

To test the effectiveness of this approach we should compare the results for known conservatives such as Palin and McCain with known Liberals such as Obama. If this test is effective it should give very different results. If I have time at the weekend I'll put something together. [[User:AndyJM|AndyJM]] 11:17, 16 January 2009 (EST)

: Please do, but note that no one ever felt McCain was a "known conservative," and Palin, though to the right of McCain, remains largely unknown in her political views.--[[User:Aschlafly|Andy Schlafly]] 11:26, 16 January 2009 (EST)

:: Excellent. Do you have any people I could test this on or should I go ahead with McCain, Palin and Obama? Anyone whose speeches are easily available on the internet would do. [[User:AndyJM|AndyJM]] 11:29, 16 January 2009 (EST)

::: Jeb Bush is considered more conservative than McCain, and many congressmen (e.g., Steve King) are more conservative than both.--[[User:Aschlafly|Andy Schlafly]] 11:30, 16 January 2009 (EST)

I'd like to suggest a supplement to this collection that might also produce interesting results. There exist uncomplicated techniques in the field of document analysis for determining similarity ratings between two bodies of text. Some work much like the work being done here, but more comprehensively, and controlling for the total number of words a speaker uses. (So a speaker that says twice as many words won't look twice as conservative.) This is very close to my personal area of expertise, so it would be very quick and easy for me to perform this sort of analysis. All it would take is full transcripts of the speeches to be analyzed, and some sort of "ideal" conservative speech, which could be as simple as a list of key words, which can be any length without increasing my workload, and can even be tailored (for example, there could be a "strong on defense" list, a "social conservative" list, and an "economic conservative" list. If someone wants to prepare such lists of key words as well as links to relevant speeches or groups of speeches to be compared, I would be happy to help. (Note that I'm basically obligated to say that the numbers I spit out are not true measures of conservatism, just a measure the extent to which the speeches analyzed use similar words to an ideal.) (Also, "phonics"?) - [[User:DaveB7|DaveB7]] 17:01, 4 July 2009 (EDT)

:::Dave, I'd love to work with you on this.  This entry has some test words, but they are based on much earlier and more primitive form of [[Essay:Best New Conservatives Words]].  I can come up with a stronger list now.  How many key words would you like, and in what form, and how many speeches?--[[User:Aschlafly|Andy Schlafly]] 18:15, 4 July 2009 (EDT)

:::: This is going to be longish. No way around it. My apologies.
:::: What I need as input depends on what we want as output. Comparing a corpus of speeches produced to a list of conservative "hot words" will produce a measure, for each speaker in question, of the rate at which they use those words, which might be indicative of their priorities. (Though I feel that it is no substitute for examining a candidate's record; actions speak louder than words, and speeches are often carefully tailored.)
:::: Making several lists of conservative "hot words", such as ones for defense, social issues, and economic issues, would allow for a comparison of how much time different candidates devote to different topics. (This would likely produce rather unsurprising results in some cases; a national politician is likely to talk about defense much more than a governor.)
:::: Another option would be to compare speeches made by candidates to existing great speeches, such as those made by Reagan. This would largely measure which candidates share Reagan's rhetorical style. (It would not be surprising, for instance, for Jeb Bush to speak similarly to either of the presidents Bush.)
:::: In all cases, the more speeches that can be included, the better, and the speeches should not be screened. (For example, a speech should not be excluded because it makes some candidate sound awfully liberal.)
:::: Any list of "hot words" should ideally be prepared by someone who isn't too familiar with any one candidates rhetorical style, to avoid (possibly unconscious) biasing of the list in favor of or against that candidate's "favorite words". I would encourage more general words, rather than words that are not especially likely to appear in any speeches at all. ("Pro-family" is good; idiosyncratic terms, unusual terms, and terms that would be political suicide to use are less helpful, but don't really hurt.)
:::: There's no reason we have to limit things to 2012 hopefuls. We can look at other recent conservatives (probably Reagan-forward, since conservatism and the issues have evolved over time.) 
:::: Ultimately, the reason that I can't offer a test that produces a "conservatism rating" is that no analysis of raw text data can really do that. Especially in politics, words are cheap, and context, personal behavior, and voting record speak volumes more than any word count can. In essence, what I'm proposing really is a "linguistic analysis of candidates" - a direct analysis of their word choices. I feel professionally obligated to say that nothing produced should be taken as direct evidence that any speaker is truly more conservative than any other. --[[User:DaveB7|DaveB7]] 17:14, 5 July 2009 (EDT)