(26-08-2024, 06:19 PM)Torsten wrote: You are not allowed to view links. Register or Login to view.I did not argue with a cumulative effect of a tendency to switch towards [ed] more frequently than away from it. This might suggest that the text was generated by a stateless machine, such as a mechanical device.
What I had in mind here wasn't a strictly algorithmic process such as a machine might carry out, but a scribe's evolving and subjective preferences about which parts of words to change, and to what, and which parts of words not to change.
"It is possible to distinguish Currier A and B based on frequency counts of tokens containing the sequence <ed>. The summary in Table 2 shows, e.g., that if <chedy> is used more frequently, this also increases the frequency of similar words, like <shedy> or <qokeedy> ...."
There seem to be two paths by which a word containing [ed] could come to be used under your hypothesis: (1) the scribe has changed a source word such as [chey] to a "related" word such as [chedy]; or (2) the scribe has copied a preexisting source word containing [ed] and kept the [ed].
Since Currier A doesn't contain source words with [ed], path (2) requires that path (1) must have happened at some earlier point to produce an appropriate source word. So the process apparently needs to start, either directly or indirectly, with path (1). And for path (1) to furnish an increasing proportion of source words for path (2), the scribe would need to be changing words without [ed] to words with [ed] more often than the other way around, perhaps not according to any strict algorithm, but nevertheless consistently over the long term.
"Now, reordering the sections with respect to the frequency of token <chedy> replaces the seemingly irregular mixture of two separate languages by the gradual evolution of a single system from 'state A' to 'state B'. Since words typical for Currier A also exist in Currier B, but not the other way round, it is reasonable to assume that the order shown in Table 2 indeed represents the original sequence in which the sections of the VMS had been created."
If path (1) were to be followed consistently, the proportion of previously-modified source words containing [ed] would steadily increase, and path (2) would then also be more likely to occur because the probability of selecting a source word that already contains [ed] would go up, leading to a kind of "snowballing" effect. This is the (subjective, human-powered) mechanism I assumed you were describing here, but maybe I read more into your explanation than you intended. Did you have a different mechanism in mind for the gradual evolution from Currier A to Currier B?
(28-08-2024, 09:24 AM)obelus wrote: You are not allowed to view links. Register or Login to view.A trend does seem to emerge in the current-page plot on the left, with later lines again evolving away from the first. On the right, with edit distance keyed to the following page, no trend is visible through the noise. Unlike generated_text.txt, Q20 shows no evidence here of page-to-page continuity (in the canonical page order).
These are some very interesting plots. If we disregard the single value for line 31 on the left as an outlier, the trend is less pronounced, but would still slope upward a bit.
(28-08-2024, 09:24 AM)obelus wrote: You are not allowed to view links. Register or Login to view.Curiously, the vms trend may be driven by the first line in particular. When we use the second line on the page as a reference, subsequent Q20 lines do not evolve away from it as convincingly as they did from the first:
We know (e.g., from tavie's work) that there are systematic differences between the composition of first lines of paragraphs and the lines that follow -- differences that go well beyond the conspicuously disproportionate frequencies of [p] and [f] in first lines. I wonder what would happen if you were to try to control for that somehow, maybe by disregarding a few select edit types ([p] to [t], [f] to [k], [Sh] to [ch], etc.). Or if you were to run a similar study at the paragraph level rather than the page level -- or else at the page level, but omitting the first paragraph and always starting with the first line of the second paragraph. Or reversing the process and gauging edit distance from the last full line, working "backwards" up the page. Or.....?