For this exercise I counted each noun as it occurred, i.e. as a token. If I counted correctly, there are 25 nouns in the machine translation, 21 in the version of Human Translator A and 15 in that of Human Translator B. Remember that Human Translator B’s version was shorter than that of A.
The finding suggests that Human Translator B uses not only a wider vocabulary but also a greater range of grammatical constructions. Whereas the machine translation and Translator A keep using nouns, Translator uses different classes of words – and this means fewer nouns than the other two translations. Hence the increased variation and readability.