Hi geujfys, and welcome to the forum!
The method that you use makes sense conceptually. We've been doing something like that with RobGea in a yet unfinished experiment, the results of which are not published yet. The major problem with this methodology, however, is to correctly choose the rules (both association rules and filtering rules) when you are trying to establish the context. Basically, it's a blind shot most of the time in the hope that it might be on target, but it might not, as well. For example, you expect that the word "blue" would appear on pages where blue flowers are depicted. That might not be the case, for many possible reasons. Some of them:
a) The flowers are not actually blue, but purple. But due to the purple paint being rare and expensive, the artist used blue paint instead. However, the word used in the text would be correct - "purple".
b) The flowers are painted erroneously by a later owner, and we don't know their real color. What's worse, they may have been painted inconsistently, e.g. a pink and a yellow flower were both painted blue.
c) The text just does not mention flower colors. That would be perfectly possible for a pharmacopaea focusing on the plants' use, not on their botanical description.
As to the Pareto principle. Yes, it is applicable to the VMS. For example, consider the "Voynich stars" (f68r1 and r2). 17% of "dayside" (68r1) stars provide 75% of their occurrences in the text, 21% of nightside (68r2) stars provide 82% of their occurrences in the text, and 17% of all stars provide 75% of total star occurrences in the text.
The method that you use makes sense conceptually. We've been doing something like that with RobGea in a yet unfinished experiment, the results of which are not published yet. The major problem with this methodology, however, is to correctly choose the rules (both association rules and filtering rules) when you are trying to establish the context. Basically, it's a blind shot most of the time in the hope that it might be on target, but it might not, as well. For example, you expect that the word "blue" would appear on pages where blue flowers are depicted. That might not be the case, for many possible reasons. Some of them:
a) The flowers are not actually blue, but purple. But due to the purple paint being rare and expensive, the artist used blue paint instead. However, the word used in the text would be correct - "purple".
b) The flowers are painted erroneously by a later owner, and we don't know their real color. What's worse, they may have been painted inconsistently, e.g. a pink and a yellow flower were both painted blue.
c) The text just does not mention flower colors. That would be perfectly possible for a pharmacopaea focusing on the plants' use, not on their botanical description.
As to the Pareto principle. Yes, it is applicable to the VMS. For example, consider the "Voynich stars" (f68r1 and r2). 17% of "dayside" (68r1) stars provide 75% of their occurrences in the text, 21% of nightside (68r2) stars provide 82% of their occurrences in the text, and 17% of all stars provide 75% of total star occurrences in the text.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)