An Essay on Entropy: what is it, and why is it so important?

[kckluge]

I was naive enough to believe that statistics in general and entropy in particular were means, among others, to achieve a purpose, which would be common to all Voynich fans: understanding the text.

It's not an issue of ends/purpose, it's an issue of methodology. Quoting a comment I made recently on Voynich Revisionist (and making clear here, as I did there, that I am not singling out the theory in question to pick on it, I am merely using it as an example):

"Darius claims that the underlying plaintext is in a 12th cent. CE dialect of Persian (You are not allowed to view links. Register or Login to view.) which has been converted from the normal script used for Persian to Voynich glyphs in accordance with the table at You are not allowed to view links. Register or Login to view. and then obfuscated with nulls that can be stripped out using the rules at You are not allowed to view links. Register or Login to view..

"So…if you take some set of 12th cent. Persian texts and map them to Voynich glyphs using his theory, how do the entropy stats of those texts compare to the entropy stats of the Voynich text after you strip out the nulls? How similar are the glyph frequency distributions of the de-nulled Voynich sample and the converted-to-Voynich-glyphs Persian text samples? For that matter, where do the h1 and h2 values of the de-nulled Voynich text using Darius’ transcription scheme fall relative to known natural languages period? Those are questions that have nothing whatsoever to do with my (or Rene’s or Glen’s or anyone else’s) interpretation of the glyphs or any existing transcription of the mss, They are purely questions about what happens when you apply an empirical test _to his theory using his assumptions_."

Someone might object, "Why bother to do that? Why not just proceed with the work of extracting Persian plaintext from the manuscript?" One answer is that if the theory can't pass that test, then the history of proposed decipherments/translations of MS 408 suggests very strongly that the probability is close to zero that the project of extracting substantial amounts of continuous, coherent Persian plaintext will succeed. Another answer is that if you do that, analyzing the discrepancies might productively help identify problems with the theory and suggest potential fixes to those problems.

We're talking about entropy here because the thread is specifically talking about entropy, but there are a variety of statistical properties and structural characteristics of the MS 408 text that ought to arise naturally from any given theory for it to be credible -- and if they don't, the odds that the deciphered/translated text produced by that theory will turn out to be plausible from a linguistic perspective are fairly slender.