NLP analysis done on a dataset of about 8,000 transcripts of Dumpster (as in “Trump’s a Dumpster-fire”) back to 2007. Unfortunately there are no trends that obviously jump out. He has probably been keeping to book more closely than generally expected, at least in these prepared interviews.
Here are three early transcripts: “xx00133” from Showbiz Tonight (CNN in 2006), “xx00598” from Your World with Neil Cavuto (Fox in 2009), and “xx00911” from Nightline (ABC in 2009):
How has the implied grade-level and complexity of Dumpster’s speaking changed over time?
How has his information content changed over time (empirical bag-of-words entropy)?
If I assume the distant past is the benchmark for Dumpster authorship, does the recent speaker seems like the same person? (This is function word distribution.)
