The Naibbe cipher

[Rafal]

According to Reddy and Knight (2011), the VMS is ~38,000 tokens long and contains ~8100 word types.

I thought myself a little bit about unique words count and ratio in VM and my observation is that in this aspect VM is actually very similar to natural languages. The diagram above shows it too.

It is also more similar to languages with high inflection (like Latin) than to languages with low inflection (like English).

I made a few tests myself and may show them at some moment. Generally the same text in Latin and English will have more unique words in Latin. You may argue that they aren't "true" unique words because many of them will be grammatical forms of the same word (like flos, floris, flore, florem...) while in English it will be always "flower". Yet most computer algorithms will treat these grammatical forms as distinct unique words.

So the ratio of unique words in VM is more similar to Latin than to English, yet we must remember that a lot depends on the topic of the text.