Writing a fake Voynich by rules: copies, habits and the link between words
#81
I found something that my fake Voynich generator (the skribe, no meaning inside) doesnt do, and the real text does.

Every hand has its usual words. On some pages it uses the most common ones a lot, on other pages it goes more for the less common ones. Same vocabulary, just a different mix. I call it the “temperature” of the page.

The funny thing is that it goes by bifolio. The four pages of a bifolio have the same temperature, even when the two leaves end up far apart in the book. Two bifolios of the same quire are not similiar at all. Pages of the same bifolio correlate around 0.4, in the generator around 0.05.

Its not the transcription, I get the same with Takahashi, Glen Claston and Zandbergen/Fagin Davis. Pages where the transcribers were unsure are not hotter or colder, so its not misreadings. And it doesnt slowly change from the start of the book to the end, it just jumps from one bifolio to the next.

If I give the generator one small random push per bifolio (a bit more common words, or a bit more rare ones) it does the same thing, and still passes the other tests.

So it looks like each bifolio was written in one go, before binding, with its own pace or mood. Davis & Layfield saw something similar with LSA. It could also be one topic per bifolio, I cant tell the two apart yet.

A word about the skribe, becuase people sometimes think it learns the language. It doesnt. It has no meaning, no sentences and no codebook. It only copies writing habits from the real text: which glyph tends to follow which (counted per hand and section), some rules for line starts, line ends, first lines and drawings, and now a random "mood" per page and per bifolio. Then it writes glyph by glyph. So when it matches something, that something can be explained by habits alone.

This is where it is now, all 88 tests I use, real against the scribe (average of 5 texts). A test passes if the scribe's number is less than 25% away from the real one, so for example if the real text gives 4.0 the scribe has to land between 3.0 and 5.0 (for the tiny ones near zero, like 10 and 11, it just has to be within 0.01). It passes 85, before the bifolio mood it was 84. Line 89 is the new one.

#testrealscribepass?
1Adjacent identical words (%)0.900.94ok
2Adjacent identical words / chance3.093.30ok
3Consecutive words at edit distance 1 / shuffled2.211.76ok
4Same word at distance 2 / shuffled3.273.74ok
5Same word at distance 3 / shuffled3.453.52ok
6Same word at distance 5 / shuffled2.853.43ok
7Frequent words repeated >3x chance (%)51.842.2ok
8Frequent words repeated <0.3x chance (%)16.125.5FAIL
9Boundary MI, distance 1 (last glyph to next first glyph)0.150.15ok
10Boundary MI, distance 20.0090.004ok
11Boundary MI, distance 30.0030.003ok
12Lift -y | q-1.691.66ok
13Lift -r | a-3.123.14ok
14Lift -y | a-0.190.19ok
15Lift -n | q-0.470.53ok
16First word of line starts ch/sh (%)8.158.29ok
17First word of line starts y (%)14.315.0ok
18Last word of line ends -m (%)15.919.0ok
19Line-end -m: lines of 10+ words / 1-4 words3.402.49FAIL
20Same first glyph as line n-1 / null0.580.59ok
21Same first glyph as line n-2 / null1.301.29ok
22Same first glyph as line n-3 / null1.161.12ok
23Distinct initial glyphs per paragraph (4+ lines)4.364.51ok
24Adjacent repeats: first line of paragraph / rest0.380.41ok
25Spearman correlation of neighbour word lengths0.160.20ok
26Word types7,9157,606ok
27Hapax (% of types)71.365.7ok
28Words fitting the slot grammar (%)97.799.9ok
29Words with 2+ gallows (%)1.801.89ok
30Mean word length (glyphs)4.804.79ok
31SD of word length1.711.76ok
32e and i in the same word / expected0.240.24ok
33H(prefix), marginal1.541.52ok
34H(prefix | previous glyph, core)0.920.93ok
35q- before o- after -y (%)57.358.7ok
36q- on first word of paragraph (%)4.195.30FAIL
37q- on first word of line (%)14.916.2ok
38q- inside the line (%)17.919.0ok
39a- at line start (%)0.350.32ok
40d-/s-/y- at line start (%)48.046.9ok
41Bare gallows at line start (%)8.979.65ok
42Closed last word, non-final lines (%)27.129.1ok
43Closed forms inside the line (%)2.342.33ok
44Closed last word of paragraph (%)17.317.6ok
45Length of 2nd word / 1st word0.880.92ok
46Gallows on first word of paragraph (%)93.292.5ok
47p/f among gallows, first line of paragraph (%)31.029.8ok
48p/f among gallows, other lines (%)2.732.82ok
49Plain p/f followed by e (%)0.310.33ok
50Plain p/f followed by ch/sh (%)59.257.5ok
51Word length, first line / rest1.081.01ok
52Closed word before a drawing gap (%)10.210.1ok
53d-/s-/y- after a drawing gap (%)41.341.5ok
54q- after a drawing gap (%)3.574.08ok
55Same first glyph after a gallows-initial line / null0.610.68ok
56Words with an edit-distance-1 twin earlier on the page (%)64.565.7ok
57Lines back to the nearest twin (median)2.002.00ok
58Page overdispersion of q- (after -y)6.326.31ok
59Vocabulary cosine: same sheet / neighbour pages, other sheets1.301.48ok
60Strange words (rare trigram in the real text) (%)4.593.70ok
61Strange words per line0.360.29ok
62Page overdispersion of e-words24.624.9ok
63e-words: biological / herbal2.041.90ok
64-edy words in biological (%)27.529.4ok
65Share of the page's most frequent word (median, %)4.714.61ok
66Words of 8+ glyphs (%)5.256.13ok
67Character entropy h13.913.90ok
68Conditional entropy h22.142.10ok
69Conditional entropy h31.861.83ok
70Zipf slope (first 1,000 types)1.000.98ok
71Words of 2-3 glyphs (%)21.622.7ok
72Words of 5-6 glyphs (%)41.039.1ok
73Shared vocabulary, neighbouring pages (cosine)0.330.30ok
74Shared vocabulary, 10 pages apart0.250.22ok
75Shared vocabulary, 100 pages apart0.150.15ok
76Neighbouring / 100 pages apart2.131.95ok
77Bursts of the 50 most frequent words (CV of gaps)1.751.69ok
78Page overdispersion of the 50 most frequent words2.061.97ok
79Words starting with a gallows, not line-initial (%)5.055.01ok
80Gallows-initial word after another / chance (in line)1.451.44ok
81sh share (sh/ch), first line / other lines1.341.31ok
82t share (t/k), first line / other lines1.231.27ok
83frequent words, first line / other lines0.660.67ok
84words with ee, first line / other lines0.870.86ok
85Line-final -m written -am, not -im (%)94.596.0ok
86Plain p among plain p+f, first line (%)75.478.3ok
87q- after -y before o, last line / other lines0.810.80ok
88ch- words, last line / other lines1.251.23ok
89Same bifolio, page to page correlation of temperature0.390.36ok

The three that fail: frequent words that avoid each other (the real text has fewer of them), -m at the end of long lines against short ones (the real effect is stronger), and q- on the first word of a paragraph (a bit too many in the scribe). Everything else is inside 25%.
[+] 4 users say Thank You to quimqu for this post
Reply


Messages In This Thread
RE: Writing a fake Voynich by rules: copies, habits and the link between words - by quimqu - 11-11-2026, 12:20 AM

Possibly Related Threads…
Thread Author Replies Views Last Post
  Evidence for 2 rules linking words in Herbal A->B npcompl33t 30 7,146 08-10-2026, 05:37 AM
Last Post: npcompl33t
  Could synthetic “fake Voynich” tests tell us what we’re dealing with? Phoenixjn 4 4,852 10-09-2025, 09:19 AM
Last Post: dexdex

Forum Jump:


Users browsing this thread: 9 Guest(s)