On Data Mining and Barack Obama
President Barack Obama is a compelling speaker. A speech can convey information or trigger feelings. Speeches also generates data. We can learn something about the speakers intentions by data mining his text. For example, below are three tables of the most frequently occurring words from Barack Obama's State of The Union Address from 2009 (when he was first elected), 2012(when he was re-elected) and 2016 (when he was leaving office). in 2016 the word "American" fell off of the top 10 while "the world" was added to the list. This is not to say that Obama was a globalist, whatever that means. The data gives us tools to better understand what we are being told. Comparing a speech holistically from one time period to another yields little in the way of quantifiable information. By breaking the speech into frequency and proportion of word usage relative to the speech and relative to the past, we can plot a time series. We can see how ideas rise and fall in importance. We can see how priorities shift from when he first got started to when he was leaving office.
For example, the following words received no mention in 2016, but were prominent in the earlier years. we will our economy responsibility debt success These had no mention in 2012 health care that is address provide necessary planet love gun And these received no mention in 2009 want to better love gun
The perceived financial concerns of 2009 faded away in 2016.
Meanwhile new concerns came to the fore. Where "Will" was prominent early in the"presidency, "Want". Security was an issue as the Islamic State and widespread war in the middle east presented Americans with renewed anxieties about terrorism. The strident rhetoric of the presidential campaign brought "politics" to the fore as a concept, once again resulting in anxiety among the American people. Obama hoped to leave Americans with some sense of purpose, highlighting Work, Love and Need To as motivational messages. Finally Obama mentioned the word Gun, recognizing the rising prominence of gun violence in American communities.
On its own, data mining speeches will leave us with a bare impression of the thinking of the speaker. Just as archaeology provides less information than written history which in turn provides us less information than lived experience. Data on its own proves one-dimensional. Discovering that dimension is exciting, even if only vaguely directional. It gives one the feeling of insight beyond the intended communication strategy of the speaker.
At a minimum it provides a map of concepts. Given a time series that map is even more interesting.
After seeing all this, one wonders what kind of comparisons can be drawn between Obama and Trump using this same method.
Even beyond politics, can this method be applied to better understand the intentions of business leaders, football coaches and others who might inadvertently telegraph their intentions by way of their word choices?
