> There is no phonetic overlap between the words "racist" and "Trump" This is true in a room full of people who aren't linguists just like it's true that the g

> There is no phonetic overlap between the words "racist" and "Trump" This is true in a room full of people who aren't linguists just like it's true that the government doesn't use SQL in a room full of Elon orbiters. I don't even like Apple. /ˈtɹəmp/ /ˈɹeɪsɪst/ There is a plausible overlap right at the beginning. You just made incorrect assumptions. You seem to assume that English has a universal phonology, that speakers all pronounce things the same way every time (i.e. you failed to account for alternation[0] on an individual speaker level, let alone the variance between dialects), and the fact that small tongue movements have an outsized effect on the produced vowel sound. [ᵗ°ɹ] and /ɹ/ sound the same under the right environmental conditions. If the /t/ is quiet enough, then the /ɹ/ can easily "outcompete" the /t/. /ə/ and /e/ are close neighbors. With natural variance, they can sound similar under the right environmental conditions. And finally, many people who say /ɹ/ also introduce a very short, almost unvoiced /ə/ before transitioning to /eɪ/, since, again, /ə/ and /e/ are close neighbors. So yes, they do have phonetic overlap, maybe not for you, but for other people it's plausible... and probable. If you watch the original video, they're recording a very short clip which gives the STT engine very little context (and they had to do it twice to reproduce the behavior because one of the attempts did not replace "racist" with "Trump"). [0]: https://en.wikipedia.org/wiki/Alternation_(linguistics) https://en.wikipedia.org/wiki/Alternation_(linguistics)