I created this Jupyter Notebook a couple weeks before the 2016 election. I took the transcriptions from the three presidential debates, as well as their tweets over the preceding months, and used Watson's Tone Analyzer to predict the emotional and tonal attributes of each candidate's tweets/debates.
You can click into the .ipynb file to see the data and results.
After graphing the results with matplotlib, I found that, perhaps unsurprisingly, the primary tone of each was anger. The level of anger for each remained approximately constant over the time period I analyzed.
In the future, I'd like to see how the winner's tone changes in the Inauguration, State of the Union, etc.