3 ms·

I wonder if it is even possible to determine whether or not the writing is nonsense or an actual language.

• schoen 10y ago
  There's a lot of research on exactly that question using various statistical tests, but some of the tests seem to have pointed in different directions. To give an example, for a written natural language, https://en.wikipedia.org/wiki/Zipf%27s_law https://en.wikipedia.org/wiki/Zipf%27s_law should generally apply to letters, words, and, I think, to n-grams as well. This is not true for encrypted or compressed data, or for data produced by a random process like throwing dice, so you can distinguish the two pretty well with histograms. The simplest tests like this generally went in favor of the Voynich manuscript having real linguistic content, but there are other statistical tests too and I'm not familiar with what they've shown. :-)
  • DonaldFisk 10y ago
    I've been studying the Voynich Manuscript. I'm using it primarily as a way of improving my statistical libraries, rather than in the expectation of being able to read it. No-one else has got very far. It has many properties of natural language, and there are two variants - Currier A and Currier B. I'm working on the assumption that it's a natural language with fairly light encryption, similar to that described here: http://ciphermysteries.com/2016/07/06/fifteenth-century-cryptography http://ciphermysteries.com/2016/07/06/fifteenth-century-cryp... My initial aim is to identify the language, and to do that I need to map the glyphs onto characters. My suspicion based on the limited analysis I've done (running a PCA on the glyphs) is that it's in a dialect of German.