Experimental replica of VMS properties with a given corpus

[MarcoP]

Hi @Marco, would you estimate that the sum of the abbreviations is more or less 7.5% of the total text ?

Hi Matthias, there are different ways to answer your question, I hope one of these is what you mean:

Code:

.                            abbreviated  non-abbreviated  % abbreviated words in ms           1344          790      63.0% (of total) abbreviation characters in ms     1661         9571      14.8% (of total) ms length vs printed length      11232        13048      86.1%

14.8% of the characters correspond to abbreviation marks, rather than symbols of the Latin alphabet.
The abbreviated text is 13.9% shorter than the printed text. So it seems that averagely a single abbreviation character stands for two ordinary characters. Also, much of abbreviating actually is about saving horizontal space by representing characters as superscripts: if one writes ī ('i' with a macron, i_ in my notation) for "in", two symbols are still used, but now they appear one above the other, instead than one after the other. In other words, I expect saved horizontal space  to be considerably more than 14%.

When I saw the shifted graph, my reaction was the same as your remark that the correction is likely excessive. The transformation is from raw transcription towards normalized text, but we don't quite know what an EVA text would look like if it were normalized. Or to what extent we are normalizing it already in the first place?

Hi Koen,
I would add that VMS values should probably be shifted by a smaller amount because they are smaller than Bonaventura's values: e.g. I could have computed the shift as a percentage, rather than an absolute value. The shifted graph is just a way to visually present what the problem is and of course must not be taken too seriously.
As you say, we don't know much about how normalized Voynichese could look like. This subject is somehow related to the recent ch/sh discussion started by Davidsch. Also, I wonder if quasi-reduplication was really meant to be the same as reduplication (of course assuming that there is some meaning somewhere). I doubt we can say much about these problems without fully understanding the text.
I am more optimistic about the fact that EVA does not add much spurious normalization (i.e. TTR reduction) to the original text; conversely it's almost sure that all Voynich transcriptions contain many hundreds of errors. The number of uncertain spaces in the Zandbergen-Landini transcription is a good indication of how serious the problem is.

________________

About the effort for this transcription, I should probably refrain from complaining: the number of characters in these ten pages is not much more than 5% of the VMS. Others have done much more work than I did! Since I typically work on texts that I am trying to understand, it could be that knowing the meaning beforehand made things more tedious than usual, taking away the pleasure of "discovery". Another side of the problem is that I defined the encoding system as I went, changing my mind a few times about some details. Having a clear transcription system in mind before one starts would certainly be more efficient.