Assuming the VM originally existed as a stack of loose bifolia, has it ever been tested if there are textual similarities within a bifolio (4 pages on the same vellum sheet)? I'm aware the sample size is probably too small for proper statistics but it would be interesting to see if there are patterns that link the text on a bifolio compared to single pages.
Nick wrote about potential BAAFU patterns You are not allowed to view links.
Register or
Login to view..
That's exactly what Colin and I have done with Latent Semantic Analysis. More here, and article coming soon:
You are not allowed to view links.
Register or
Login to view.
The evidence from Colin's analytics shows exactly what you're asking - a very strong textual correlation across conjoint bifolia in both the balneology and stars sections. We did NOT find that correlation across conjoint bifolia in the herbal section, which suggests that, as long suspected, each herbal page is it's own semantic unit.
In other words, 104v and 115r (conjoint) are more closely related than, say, 104v and 105r (consecutive).
There is plenty of evidence that suggests this. I have posted something on this before.
You are not allowed to view links.
Register or
Login to view.
(13-12-2025, 02:22 PM)LisaFaginDavis Wrote: You are not allowed to view links. Register or Login to view.In other words, 104v and 115r (conjoint) are more closely related than, say, 104v and 105r (consecutive).
Yes, but there are 4 exceptions in Q20 (4/10 versos are not more closely related to the conjoint recto than to the consecutive recto), not a "very" strong textual correlation across conjoint bifolia. Probably there is a better metric than the one I'm using. And, of course, the result would be different if the bifolios were reordered, so there may be a way to make consecutive pages more closely related.
[
attachment=12963]
Comparison of the vocabulary (set of words) of Q20 pages using the RF1b-er transliteration (without '?' words, with all spaces kept, i.e. all '.' and all ','). The % is the You are not allowed to view links.
Register or
Login to view. coefficient.
You should watch the video of my Toronto lecture for an explanation of the LSA methodology and our results:
You are not allowed to view links.
Register or
Login to view.
It seems clear to me, based on the physical and data evidence, that the manuscript was intended to be a stack of unbound singulions rather than nested quires. Colin and I are now working on how we can hypothesize the order of singulions by maximizing the LSA cosine results across sections. Our article will include a link to a GitHub respository with all of the code and data.
I understand. It doesn't seem to matter for the plants and recipes section. But what about the leaflets? They are significantly larger.
Why are they 2x3 in the astrology section and not 3x2 like the others? Then they would all be the same. How do you explain that in theory?
And what about Q13?
Next question: stacked open or folded.
Folded, almost certainly.
What do you mean by "leaflets"? The foldouts? And I'm not sure what you mean by "2x3" and "3x2".
(15-12-2025, 01:06 AM)LisaFaginDavis Wrote: You are not allowed to view links. Register or Login to view.Folded, almost certainly.
Does the LSA suggest they were meant to be read folded or unfolded, independently of how they were stored (i.e., a reading order of (say) 115v -> 104r -> 104v -> 115r vs. 104r -> 104v -> 115r -> 115v)?
(13-12-2025, 02:22 PM)LisaFaginDavis Wrote: You are not allowed to view links. Register or Login to view.That's exactly what Colin and I have done with Latent Semantic Analysis. More here, and article coming soon:
You are not allowed to view links. Register or Login to view.
The evidence from Colin's analytics shows exactly what you're asking - a very strong textual correlation across conjoint bifolia in both the balneology and stars sections. We did NOT find that correlation across conjoint bifolia in the herbal section, which suggests that, as long suspected, each herbal page is it's own semantic unit.
In other words, 104v and 115r (conjoint) are more closely related than, say, 104v and 105r (consecutive).
It seems to me that this (the difference in LSA behavior between the balneology & stars section bifolia on the one hand and the herbal bifolia on the other) speaks strongly against theories involving glossolalia/automatic writing, Ruggian grids, self-citation methods, etc. because there is no obvious reason for such text generation mechanisms to produce bifolia with different LSA behavior in different (types of) sections. Or am I misunderstanding what the LSA analysis is saying about the herbal bifolia vs the balneology & stars bifolia?