Discussion of "A possible generating algorithm of the Voynich manuscript"

[ReneZ]

Some references to Timm's text generation came up in a recent thread related to the text on f85.

The following quote from JKP was not referring directly to it, but includes a few useful key words:

My personal opinion (based on studying it) is that there may be a set of rules governing it, but there are a few places where it is extremely difficult to discern why a particular character was inserted in a particular place. I've wrestled with this for years.

Is it mostly rule-based with some room for exceptions? Or is it mostly rule-based with exceptions that encode meaning? If so, the proportion of meaningful text is small.

That there are rules governing word forms, and that these word forms create a pattern of very similar words is one thing. There should be no doubt that this is true.

That the Voynich MS was written by taking previous words and randomly changing it a bit and writing down the result is the proposal of Timm, but this is quite a different thing.

There is nothing arbitrary about the word patterns. Stolfi's core-mantle-crust paradigm explains something like 95% of all words in the MS. Therefore, there are quite strict rules which changes would have been allowed, from the beginning of the MS text till the very end, some 36,000 words in total. In that sense, JKP's question can be answered, I think, that it is more rule-based than exception based.

One could do a numerical experiment as well, but I don't have the right tool to do. However, I can describe it.

Take the common word chedy for example.


There are three ways to change it with a Levenshtein distance of 1.

1) Delete a character.
If we assume ch to be a unit, this gives four possibilities: edy , chdy , chey , ched
I don't believe that all four occur, but only the middle two are reasonsbly frequent. So, deleting one character would not not have been done arbitrarily.

2) Swap two characters.
This gives three possibilities: echdy , chdey , cheyd
These are all rare or non-existent, so swapping would not have been used a lot. However, some cases exist.

3) Insert a character.
Now there are five possible insertion positions, and at each point in principle over 20 different characters could be inserted. This gives well over 100 different possibilities. I dare say (but I can't check) that only a fraction of them actually exist. So here were see the influence of the 'rules'

Effectively, it means that inside the well-known network of words that has been shown also in this forum, there is a much, much denser network of words that could have been created with the Timm method, but don't exist. This process is strongly governed by rules.

The conclusion that the Voynich MS text is based on a random generation is not supported by this evidence.

Note that I am not saying the opposite. The above argument does not 'prove' in any way that the text is meaningful. I am just saying that we can't tell at the present time.