I was trolling 538 earlier, and came across this little tidbit:
It uses a measure called tf-idf (term frequency-inverse document frequency) which, not to get too technical, sorts out the words or phrases that are used most often in a given set of documents that aren’t the standard, highly common words such as ‘a’ and ‘the.’ This is a form of a strategy that search engines use to determine which pages are most relevant to the search terms entered.
So the author applied it to the Dems’ campaign speeches, and that page is what he came up with. (Note that the scores listed are only relative, and in no way proportional; in other words, a phrase with twice the score of another phrase doesn’t necessarily occur twice as often.)
I’m almost afraid to see what Trump’s listing looks like.
Naah, screw that. The prurient train-crash side of me wants to see.
EDIT: Found it.