It's not, in this case, a translation, it's speech to text by machine. I have the original audio. I can confirm there are no cats, dogs, socialists, shells, hurricanes, high spots, and the only correct name is John. Normally a human types up the recording and another human (like me) turns the typed speech into something readable (cos people are vague and windbaggy). With machine stuff it's like doing both jobs in one.