Text Analysis with Voyant

I initially thought that I’d compare the inaugural speeches of Presidents Obama and Trump (which I did), and then thought I’d look at the first and second inaugural speeches of President Bush (43) and Obama.  While I found these comparisons somewhat interesting, I didn’t think that they were as enlightening as I’d hoped.  I was actually surprised with the way that Trump’s speech appeared, in that it did not portray the somewhat bleak picture of the nation that I thought his speech conveyed.

(somehow I was not able to display these slides here, so I added the links…..)


I thought that comparing speeches of Presidents was an obvious choice, so then I decided to compare the opening statements of Brett Kavanaugh and Dr. Blasey Ford’s during the recent Senate Hearing.  Here again, while interesting, I was somewhat disappointed by what the software displayed.

Kavanaugh’s Opening Statement:


If one were to rely on the links and word trend screens, it is accurate in that he was speaking of his high school years, and the frequency of words paint the picture of an adolescent’s focus on friends, parties, beer, and boys and girls.  Sounds innocent enough, right?  But what did not get conveyed was his adamant denial of the allegations, the outright partisan statements that he made, and the overall impression that he left on many that questioned his judicial temperament, and his ability to be impartial and balanced.

Dr Ford’s opening statement:


The word links and word trends window for Dr. Ford’s statement appear to be a more accurate representation of her testimony.  One could infer that it is more specific and focused on the event.  But again, what is missing here is the general impression that the speaker imparts to the world, which is conveyed in tone of voice, cadence, emphasis, and overall demeanor.

I want to also temper my impressions here, in that this is the first time that I’m using this tool, and I could be missing something, or not parsing the data, deleting certain words which could skew the results, or any number of techniques that would make the results more meaningful.  Which is to say that with any tool, understanding HOW to use it correctly, and WHEN are critical.

I then looked at some tutorials on YouTube to learn more.  The question that I find interesting is what are the best uses of this tool, in order to provide data and insights that are not evident from reading specific texts?  I stumbled upon a presentation given by by Stéfan Sinclair (McGill), one of the developers of Voyant.  There was one slide that I found very interesting:


At 13.00 minutes : This was a comparison of the texts from advertising for toys, for boys, and for girls.  These two images are striking in that they really say it all, and there is no other explanation needed to illustrate the gendered stereotypes that are still reinforced by advertising to children.

If I were to make a conclusion, perhaps this tool is more instructive when looking at a body of work, as opposed to evaluating a single text.  Then patterns that cannot easily be gleaned from reading individual documents or transcripts might be teased out of the analysis that this type of a tool could provide.

One thought on “Text Analysis with Voyant

  1. Dax Oliver

    I think the Voyant similarity between the inaugural speeches of Trump and Obama is itself an interesting pattern. Especially since most voters probably felt they were very different when listening to them. I believe this can be explained by the fact that most Republicans and Democrats are human beings living in the United States, so they’re equally attracted to rhetoric about “America” and “new” and “dreams.” The difference between the speeches is more in their singular choices of phrases, such as Trump’s line about “radical Islamic terrorism” and Obama’s line about how “we will harness the sun and the winds.” These differences can only be seen through close reading.

Comments are closed.