The Challenge of Analyzing a Dynamic Text

[Torsten]

There's not much information in the paper about how the model recreates conditional dependencies (e.g., word-to-word transition statistics, positional entropy) do you have anything to elaborate on that?

Also, have you ever done a systematic correlation between particular glyph clusters and illustration types?
Because that would be hard to reconcile with pure pseudo-text.

The study thats linked about people defaulting to copying takes place in a modern context though and I don't know how much it would relate to a medieval scribe, but is interesting nonetheless.

Thank you for your excellent questions.

To question 1) Positional entropy patterns:
The positional entropy patterns emerge because of two reasons. One is the permanent copying process leading to a feedback loop. "The rules to modify a source word normally don't affect the order of the glyphs. This is one reason for the observation that the words in the VMS share the same rigid word structure" (Timm & Schinner 2020, p. 10). And "As for the actual selection process of source words, it is clear from the results of section 2 (as well as simply suggested by the scribe's convenience) that they are to be chosen at least from the same page. Since it is handy to copy a word from the same position some lines above, our implementation of the algorithm includes a mechanism that selects (with a given probability) even tokens from the previous line at the same writing position." You are not allowed to view links. Register or Login to view..
The second one is that "for a text created by self-citation it appears as a logical assumption that the scribe also used aesthetically motivated design rules for glyph selection, in order to harmonize the overall appearance of the text. These rules specify when two glyphs may follow one another" (Timm & Schinner 2020, p. 10). Schwerdtfeger (You are not allowed to view links. Register or Login to view.) described in 2008 the following four design rules for the VMS: (1) line-glyphs can follow line-glyphs or <a>; (2) curve-glyphs and <a> can follow curve-glyphs; (3) the <l>-glyph can be used as a curve-glyph or as a line-glyph; and (4) gallows glyphs count as curve glyphs." Timm & Schinner 2020, p 10.

To question 2) Correlation between text and illustrations:
The shift from Currier A to B suggests that the order of sections was: Herbal in Currier A, Pharma in Currier A, Astro, Cosmo, Herbal in Currier B, Stars in Currier B, Biological in Currier B (see section "Shift from <chol>/<chor> to <chedy>/<shedy>" in You are not allowed to view links. Register or Login to view.). This means with the exception of the pages sharing Herbal illustrations, the scribe wrote pages sharing the same illustration together.
In short, the two Herbal sections are devastating to the semantic correlation hypothesis. If vocabulary tracked illustration meaning, both Herbal sections should use similar words. Instead, they differ because they were written at different evolutionary stages. Topic modeling detects vocabulary differences between sections and misinterprets them as "topics" when they're actually just temporal snapshots of an evolving pseudo-text generation process.

To question 3) Modern vs. medieval context:
The Gaskell & Bowern experiment validates a universal cognitive constraint, not a culturally-specific behavior. If so, medieval scribes would be more likely to use copying strategies, since their primary professional activity was copying texts. The medieval context therefore strengthens rather than weakens the self-citation argument.