I think some day in the future we might get to a point where political speech writing, or maybe just the writing industry in general, develops a dichotomy a bit like hacking where you have white hats on the one side using speech analysis tools and generators to create works and test them against mock audiences, and black hats on the other side using similar tools to figure out the audiences they were targeting and the parameters involved.
Take this to the extreme and you get neural nets analysing your live speech, finishing your sentences with viable suggestions, each suggestion optimised for sentiment and meaning towards personal goals and relationships with your audience. A little guy in your ear to tell you how you're statistically doing in your job interview or date and suggest some things to say to steer you toward your goal. Of course the person you're taking to also has the same little guy and they are both trying to outplay each other by predicting the other's next move.
Also, this is cool, nice work! Feature request: hover over the data points to see the year and name of the president.
Take this to the extreme and you get neural nets analysing your live speech, finishing your sentences with viable suggestions, each suggestion optimised for sentiment and meaning towards personal goals and relationships with your audience. A little guy in your ear to tell you how you're statistically doing in your job interview or date and suggest some things to say to steer you toward your goal. Of course the person you're taking to also has the same little guy and they are both trying to outplay each other by predicting the other's next move.
Also, this is cool, nice work! Feature request: hover over the data points to see the year and name of the president.