Debate bingo: 2016 presidential edition
Bingo cards with frequently repeated phrases from 50 recent Hillary Clinton and Donald Trump speeches.
Print this bingo card
Generate another card
September 5, 2016
The presidential election debates between Hillary Clinton and Donald Trump are coming up in a few weeks, and they promise to be some of the finest reality TV programming of the year. Debate bingo is the perfect way to take it to the next level!
Only nearly unpredictable
Although it feels like nearly anything could happen during the debate, some of what the candidates say can be predicted with high confidence. Regardless of the questions or the topic of the debate, is there any chance that Donald Trump won't talk about making America great again? Or that Hillary Clinton won't talk about making the economy work for everyone?
This is their "tried-and-true" messaging. They know the issues that their voters care about, and through their months of campaigning have worked out the best way to discuss those issues. In the heat of the debate, there's a pretty good chance that they'll fall back on some of this honed messaging. If this hypothesis is true, then the candidates' recent speeches should give some indication of the specific things they will say at the debate, despite the different format.
Constructing debate bingo cards
My intention here is to make bingo cards for actual use that are actually fun. It is critically important that the phrases have a high probability of use, otherwise you won't get to mark off many squares and it will be totally boring. To get a collection of likely phrases, I extracted frequently repeated phrases from their recent speeches.
I started by gathering transcripts of 25 speeches from each candidate, from the last 9 months. This involved searching around various news sites and the candidates' sites. I then extracted all sequences of between 2 and 20 words (each sequence here called a "phrase") and filtered these to phrases that appeared in at least 5 speeches. I did a very aggressive stop-word filtering, by requiring every phrase to have at least 2 non-stopwords. This certainly eliminates some single words that are nearly certain to come up (like "emails"), but also focuses the phrases on slightly more complex ideas or catch-phrases, which I think makes for more interesting cards. Finally, I removed any phrases that were substrings of other phrases. The result of this extraction was 355 frequently used phrases, which I manually pruned down to 153, removing any that weren't specific enough to be interesting.
There are several issues with this approach, the biggest of which is probably that it doesn't account for the nonstationarity in their messaging. There are things that the candidates talked about frequently earlier in the campaign that they talk about less now, and so will be less likely to come up in the debate. Similarly, any issues that were recently added to their stump speeches haven't had time to make it past the 5 speech filter, and so won't show up. That said, I'm pretty hopeful that the phrase collection used to generate the cards is good enough to be fun.
The bingo card above is generated by randomly sampling from the 153 extracted phrases. This is a large enough number that we are all but guaranteed every card has a unique collection of phrases, not to mention the variability in the positions. Even the overlap in phrases between any two cards will be quite low, which should make for great bingo! Good luck!
Photos of Hillary Clinton and Donald Trump by Gage Skidmore.
Comments? Email me at email@example.com.