02-05-2021, 08:22 PM
Many thanks Marco.
I really appreciate your efforts in trying this out.
To respond to the comment about the word length distribution, there are two "problems". The first is that, in order to obtain the Voynich binomial word length distribution, which also includes shorter words, one should have three empty entries in the left column, three in the centre column, and three length-1 items in the right column, corresponding with one of the earlier tables in the paper (Table 4?).
The next problem is, that this works if all combinations are equally likely to occur, but this will not be the case in this particular application of the method. The number of different trigrams in English is far less than 24 to the power 3. It is also (almost certainly) less than the number of word types in the MS.
The single character distribution frequency is very uneven, so the final WLD can be skewed either way easily, by making different mappings in the table.
However, I understand that you approached this in an unbiased and statistically reasonable way.
Most of these problems disappear, when the method is not used to map trigrams, but in a 'nomenclator' style. However, that is not so easy to experiment with.
I really appreciate your efforts in trying this out.
To respond to the comment about the word length distribution, there are two "problems". The first is that, in order to obtain the Voynich binomial word length distribution, which also includes shorter words, one should have three empty entries in the left column, three in the centre column, and three length-1 items in the right column, corresponding with one of the earlier tables in the paper (Table 4?).
The next problem is, that this works if all combinations are equally likely to occur, but this will not be the case in this particular application of the method. The number of different trigrams in English is far less than 24 to the power 3. It is also (almost certainly) less than the number of word types in the MS.
The single character distribution frequency is very uneven, so the final WLD can be skewed either way easily, by making different mappings in the table.
However, I understand that you approached this in an unbiased and statistically reasonable way.
Most of these problems disappear, when the method is not used to map trigrams, but in a 'nomenclator' style. However, that is not so easy to experiment with.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)