22-04-2017, 10:31 PM
(22-04-2017, 07:14 PM)Torsten wrote: You are not allowed to view links. Register or Login to view.Words in the VMS are not related with the previous words. But this did not mean that words are random or as you say very random. That I give examples for some observations did not mean that my statements are based on some details:
90 % of the paragraphs start with a gallow glyph (see You are not allowed to view links. Register or Login to view.). There are also glyphs typical for line initial position (see You are not allowed to view links. Register or Login to view.).
600 out of 1100 words ending with [m] are used as last word of a line (see You are not allowed to view links. Register or Login to view.).
See You are not allowed to view links. Register or Login to view., You are not allowed to view links. Register or Login to view. or You are not allowed to view links. Register or Login to view. for the observation that words in the VMS co-occur with similar ones.
See You are not allowed to view links. Register or Login to view. for the statistically-significant effect of word-final glyphs on the initial symbol of a following word.
See You are not allowed to view links. Register or Login to view. for some observations about the line as a functional unit.
The words in the VMS are not ordered as expected for a natural language but they are still ordered in some way.
So, when a word starting with a gallow glyph is generated, a new paragraph is sometimes started. When the last word generated ended in m, a new line is started, etc. That doesn't necessarily signify meaning.
I found the same relationships between page types, and between certain words, found previously by Montemurro. It's because different parts of the manuscript are generated slightly differently. -edy words occur so frequently in the Currier B (red) herbal pages that they appear to be related.
I'm still unconvinced. Anyone's free to attempt to reproduce my results, or find a way to infer meaning consistent with them.
However, things still aren't quite correct. I neglected the labels, as that involves a fair bit of preprocessing. It's known they have different properties to the other text. I also need to get in place a new (probably Poisson) random function and a means of choosing the next glyph which has the property that variance = mean^2.