Bigrams across uncertain spaces

[pfeaster]

This appears to be related with what Patrick Feaster discusses in his You are not allowed to view links. Register or Login to view. § 4 Word Breaks, Line Breaks, Paragraph Breaks, Labels. I read this post a while ago and at the moment I can't tell how close the measures are.

I limited my analysis to "paragraphic" text, while I suspect RobGea included labels, circles, radii, etc.  For that reason, my counts are likely to be a bit lower in each case.  But in general the ratios seem comparable.  Here are just a couple examples:

For r_a, I counted 631 apart (r.a) and 248 ambiguous (r,a), plus 573 together (ra).
RobGea has 730 apart (r.a) and 285 ambiguous (r,a).

For l_o, I counted 847 apart (l.o) and 72 ambiguous (l,o), plus 499 together (lo).
RobGea has 996 apart (l.o) and 78 ambiguous (l,o).

Beyond these raw counts, I think we're trying to measure different things, so the rest of the figures may not be comparable.

If I'm reading RobGea's post correctly, the idea there is to measure differences between the proportion of total "dot breaks" and the proportion of total "comma breaks" corresponding to each bigram.  And there do seem to be some striking differences.

My figures came out of an attempt to study the predictability of spacing -- i.e., given a string of Voynichese with all spaces removed, how accurately could the spaces be reinserted according to consistent rules about whether each "adjacency" should have a space inserted or not.  Spacing in this sense turned out to be around 95% predictable.  That remaining 5% consisted of cases where an adjacency that's usually spaced wasn't, or vice versa.  So far this analysis ignored comma breaks.  But the interesting detail, I thought, was that the same adjacencies that were most inconsistent when there was no question about there being a space or not -- for example, lots of unambiguous cases of both "r.a" and "ra" -- also made up the greatest proportion of the comma breaks ("r,a").  That is, the adjacencies that were being written inconsistently also seemed to be the ones that were being written ambiguously enough to leave doubt and trigger a comma.

One explanation would be that "comma spaces" aren't just an artifact of transcription, but reflect a real ambiguity in the script that might shed light on how it's structured.  But I'll admit that isn't the only possible explanation. 

Do you recall the order in which the pages were transcribed?

I ask because I'm interested in how strong the uncertain spaces were affected by subjectivity. I wonder if uncertain spaces became more or less common during the course of the transcription. That is, either you or Gabriel became more aware or word patterns and they influence judgements.

So I guess we could imagine a learning process in which as the transcribers gained more experience with which glyph pairs are typically spaced or unspaced, they would become more confident in their judgment calls and more likely to opt for a dot than for a comma.  That said, I have a feeling that Rene and Gabriel were already immersed in the patterns of Voynichese before they started their transcription, and when I've checked comma spaces against the scans, they seem for the most part to be legitimately ambiguous.  And sometimes ambiguous spacing seems even more curiously patterned: