Vord paradigm tool

[MarcoP]

The possible poetic nature of the text tends to be overlooked.

To my knowledge, most researchers regard poetry as a possibility. For instance:

• In their statistical analyses, You are not allowed to view links. Register or Login to view., You are not allowed to view links. Register or Login to view., You are not allowed to view links. Register or Login to view. included both poetry and prose.
  
• In his 2014 paper, You are not allowed to view links. Register or Login to view. mentioned that "although such repetition is an unusual feature in natural languages, it is not unknown in particular genres (e.g. poetry and incantation)".
  
• You are not allowed to view links. Register or Login to view. compared poetry and features of Voynichese.
  

I guess that the reason why the subject is not explicitly mentioned more often is that, both linguistically and statistically, poetry is very similar to prose: they both are language. Measures like character conditional entropy and MATTR do not vary considerably between poetry and prose: this tells us that Voynichese cannot be a simple substitution of prose or poetry written in any ordinary European language (e.g. Latin, Italian, German, English). It would be great if there was a longer text of any kind that exhibits the same rates of perfect reduplication and partial reduplication that we see in the VMS, but as far as I know nothing really close has been found yet. For instance, the Finnish poem Kalevala (collected by Brian Cham and pointed out by Jonas Alin) features both phenomena, but the rate is very far from Voynichese (check FIN vs EVA in You are not allowed to view links. Register or Login to view.).

Here I put together a few plots comparing verse (V) and prose (P). I included the "Veni Virgo" poem pointed out by Yulia and the Donatus text pointed out by stenog. It should be noted that "Veni Virgo" is too short for reliable statistics. Also, I could have made errors somewhere along the process. The texts are:

P-DE-weltchr Nuremberg Cronicle
V-DE-tristan Tristan
P-ENG-g.herb Grete Herball
P-ENG-kjames King James Revelations
V-ENG-shake Shakespeare's Sonnets
P-LAT-don.all The whole Donatus text
P-LAT-don.pron Donatus "Pronoun" paragraph (including the quote You are not allowed to view links. Register or Login to view.)
P-LAT-pliny Historiae
V-LAT-veni Veni Virgo
V-LAT-virgil Aeneid
P-ITA-mach Machiavelli: Il Principe
V-ITA-dante Divina Commedia
EVA transliteration of the individual sections

Character entropy and character conditional entropy
   
Character conditional entropy is possibly the best known statistics. The VMS has consistently low values, also using different transliterations. Yulia and stenog's texts appear to "move" in the right directions: this is partially due to the fact that they are short, but the repetitiveness contributes as well. The whole Donatus text is much  closer to ordinary Latin. In general, there is no apparent systematic difference between Verse and Prose.


Mattr5 and Mattr20
   
Similar plots were You are not allowed to view links. Register or Login to view.. This is one of those plots that makes Voynichese look almost normal: the difference is that it falls above the regression line because MATTR5 has lower values than normal due to the frequent identical repetition of words. Donatus, VeniVirgo and King James' Revelations are outliers with particularly low MATTR5 values. In the case of King James, this is not due to the consecutive repetition of words, but to the same word occurring at a distance of less than 5 words, eg:

and riches and wisdom and strength and honour
of god and of the testimony of jesus
had a face as a man

Again, the fact that the text is verse or prose does not appear to have a clear effect.


Perfect reduplication and partial reduplication (1 character difference)
   
The bulk of verse and prose samples fall near the origin with low values for both measures. VeniVirgo and (more clearly) Donatus come closer to Voynichese. I suspect that the spacial distribution of reduplication is different in Voynichese and Donatus. In the VMS the phenomenon is more or less uniformly distributed, in Donatus I guess it is concentrated in paragraphs where declensions are listed.


In conclusion, I find that considering both poetry and prose is the safest course of action. The assumption that the text is poetry does not explain much by itself. We don't know if the text is meaningful and, if it is, how much of the weirdness is due to the underlying text and how much is due to the writing system, so it's impossible to say anything conclusive about the nature of the underlying text.