Edit: As of 8 January, 2021, @realdonaldtrump is no longer a Twitter user, but he was at the time of this post.

Version 2.0.1 of my iOS app NastyWriter has 184 different insults (plus two extra special secret non-insults that appear rarely for people who’ve paid to remove ads 🤫) which it can automatically add before nouns in the text you enter. “But Angela,” I hear you not asking, “you’re so incredibly nice! How could you possibly come up with 184 distinct insults?” and I have to admit, while I’ve been known to rap on occasion, I have not in fact been studying the Art of the Diss — I have a secret source. (This is a bonus joke for people with non-rhotic accents.)

My secret source is the Trump Twitter Archive. Since NastyWriter is all about adding gratuitous insults immediately before nouns, which Twitter user @realdonaldtrump is such a dab hand at, I got almost all of the insults from there. But I couldn’t stand to read it all myself, so I wrote a Mac app to go through all of the tweets and find every word that seemed to be an adjective immediately before a noun. I used NSLinguisticTagger, because the new Natural Language framework did not exist when I first wrote it.

Natural language processing is not 100% accurate, because language is complicated — indeed, the app thought ‘RT’, ‘bit.ly’, and a lot of twitter @usernames (most commonly @ApprenticeNBC) and hashtags were adjectives, and the usernames and hashtags were indeed used as adjectives (usually noun adjuncts) e.g. in ‘@USDOT funding’. One surprising supposed adjective was ‘gsfsgh2kpc’, which was in a shortened URL mentioned 16 times, to a site which Amazon CloudFront blocks access to from my country.

For each purported adjective the app found, I had a look at how it was used before adding it to NastyWriter’s insult collection. Was it really an adjective used before a noun? Was it used as an insult? Was it gratuitous? Were there any other words it was commonly paired with, making a more complex insult such as ‘totally conflicted and discredited’, or ‘frumpy and very dumb’? Was it often in allcaps or otherwise capitalised in a specific way?

But let’s say we don’t care too much about that and just want to know roughly which adjectives he used the most. Can you guess which is the most common adjective found before a noun? I’ll give you a hint: he uses it a lot in other parts of sentences too. Here are the top 35 as of 6 November 2020:

1. ‘great’ appears 4402 times
2. ‘big’ appears 1351 times
3. ‘good’ appears 1105 times
4. ‘new’ appears 1034 times
5. ‘many’ appears 980 times
6. ‘last’ appears 809 times
7. ‘best’ appears 724 times
8. ‘other’ appears 719 times
9. ‘fake’ appears 686 times
10. ‘American’ appears 592 times
11. ‘real’ appears 510 times
12. ‘total’ appears 509 times
13. ‘bad’ appears 466 times
14. ‘first’ appears 438 times
15. ‘next’ appears 407 times
16. ‘wonderful’ appears 375 times
17. ‘amazing’ appears 354 times
18. ‘only’ appears 325 times
19. ‘political’ appears 310 times
20. ‘beautiful’ appears 298 times
21. ‘fantastic’ appears 279 times
22. ‘tremendous’ appears 270 times
23. ‘massive’ appears 268 times
24. ‘illegal’ appears 254 times
25. ‘incredible’ appears 254 times
26. ‘nice’ appears 251 times
27. ‘strong’ appears 250 times
28. ‘greatest’ appears 248 times
29. ‘true’ appears 247 times
30. ‘major’ appears 243 times
31. ‘same’ appears 236 times
32. ‘terrible’ appears 231 times
33. ‘presidential’ appears 221 times
34. ‘much’ appears 217 times
35. ‘long’ appears 215 times

So as you can see, he doesn’t only insult. The first negative word, ‘fake’, is only the ninth most common, though more common than its antonyms ‘real’ and ‘true’, if they’re taken separately (‘false’ is in 72nd position, with 102 uses before nouns, while ‘genuine’ has only four uses.) And ‘illegal’ only slightly outdoes ‘nice’.

He also talks about American things a lot, which is not surprising given his location. ‘Russian’ comes in 111st place, with 62 uses, so about a tenth as many as ‘American’. As far as country adjectives go, ‘Iranian’ is next with 40 uses before nouns, then ‘Mexican’ with 39, and ‘Chinese’ with 37. ‘Islamic’ has 33. ‘Jewish’ and ‘White’ each have 27 uses as adjectives before nouns, though the latter is almost always describing a house rather than people. The next unequivocally racial (i.e. referring to a group of people rather than a specific region) adjective is ‘Hispanic’, with 25. I’m not an expert on what’s unequivocally racial, but I can tell you that ‘racial’ itself has nine adjectival uses before nouns, and ‘racist’ has three.

“But Angela,” I hear you not asking, “why are you showing us a list of words and numbers? Didn’t you just make an audiovisual word cloud generator a few months ago?” and the answer is, yes, indeed, I did make a word cloud generator that makes visual and audio word clouds, So here is an audiovisual word cloud of all the adjectives found at least twice before nouns in tweets by @realdonaldtrump in The Trump Twitter Archive, with Twitter usernames filtered out even if they are used as adjectives. More common words are larger and louder. Words are panned left or right so they can be more easily distinguished, so this is best heard in stereo.

There are some nouns in there, but they are only counted when used as attributive nouns to modify other nouns, e.g. ‘NATO countries’, or ‘ObamaCare website’.