Hi Davidd, I am happy to see new research on this subject. I believe there are patterns that can be discovered with the POS-approach (though I haven't been very successful).
I have a few suggestions and a question:
I have a few suggestions and a question:
- submit your research to the Voynich Day presentation on August 4; it doesn't have to be perfect, you could just describe in more detail what you have done so far
- add a few more words as labels of word classes, e.g. I believe that "shedy" includes "chedy" and vice-versa - it would be nice if this was clear from the graphs
- try your method at least on a simple English text with the same word count as the Voynich samples you are processing; this will tell you what are the results you can expect in a best-case scenario. A more complex English text (e.g. Shakespeare) would also be interesting, for comparison.
- Question: why the double arcs between "aiin" and "otaiin"? The "aiin" loop is also doubled, but in that case the % is the same, so it's less of a problem [EDIT: I guess it could be because you are plotting an arc for "aiin" as a class following "otaiin" and another arc for "otaiin" as a class preceding "aiin"; a possible solution could be only plotting "following" or "preceding" arcs. IIRC my approach was computing % as the rate or total word transitions on the whole text, rather than the rate of "aiin" following "otaiin" on the total of "aiin.X" couples]