I’m not talking about paleography here, but I wanted to share some results from my own You are not allowed to view links.
Register or
Login to view. (DMR with K=6).
I don’t know if the topics I found are really semantic topics or if they are capturing something more like writing styles. But what struck me is that these clusters line up quite clearly with three different subdivisions: the bifolia, the quires, the writing hands, and the language A/B division.
So if these topics are in fact reflecting style (even though I try to avoid that bias), it could be another argument in favor of the manuscript being created by multiple scribes, each with their own writing style.
Here are the three stacked-bar plots showing the distributions (each topic is a color, measured by paragraph, not by folio, that's why a folio can have multiple topics):
By bifolio (divided by bifolio and by order of appearance within the MS) (look for example how quire E is divided clearly in two topics, two bifolia red, two bifolia brown)
![[Image: dhUy0gC.png]](https://i.imgur.com/dhUy0gC.png)
By quire (divided by quire and by order of appearance within the MS)
![[Image: o5tUvCI.png]](https://i.imgur.com/o5tUvCI.png)
By writing hand (divided by hand and by order of appearance within the MS)
![[Image: AGQZo9t.png]](https://i.imgur.com/AGQZo9t.png)
By language (A/B) (divided by language and by order of appearance within the MS)