The Textual Work of August Walla, mentally disabled artist

[Alin_J]

Yes, there were some mistakes in my code... I had missed to correctly capture the = signs at line-breaks. It was corrected and I rerun the program. The data now changed more, but I don't think the changes to the graph is that significant, but here it is anyway.

I would also suggest not to make the search case sensitive. Maybe this would decrease the unique word percentage to a more reasonable number.

Yes, I had planned this as well, it will make a separate analysis. But since the idea is to compare it with an unknown writing system which might also have separate incidents of characters with identical meaning, I think it could tell something as well.

The new basic data looks as following:
Average word length (counting each unique word only once): 8.53786 characters, standard deviation: 5.3916
Average word length (text aggregate): 7.52895 characters, standard deviation: 4.99348
Number of unique words: 1149
Total number of words (tokens): 1658

So, the type-token ratio is now 69% (not a big change from 72%).

Changes to the number of repeated phrases:
Number of pairs: 59
# Triplets: 31
# Quadruplets: 22
# Quintuples: 16
# Sixtuples: 13

Here is the updated graph: 
   

[Alin_J]

One thing the Walla text has in common with the Voynich manuscript is the high number of consecutive repetitions (re-duplication) of the same word two- or three times in a row. Here is the data on that.

Total number of pair repetitions: 67 (Expected in randomly word-shuffled text: 2.8068)

Total number of triplet repetitions: 4 (Expected in randomly word-shuffled text: 0.014482)

All duplications/triplications are mostly on different words each time.

[Ben Trovato]

That was actually the first thing that drew my attention! Can you compare the repetitive patterns between VMS and Walla?

[Alin_J]

That was actually the first thing that drew my attention! Can you compare the repetitive patterns between VMS and Walla?

Yes, I think so. Can you be more specific?

[Alin_J]

Converting all upper-case characters to lower-case, there was a reduction in the number of unique words to 1019, making the type-token ratio about 63%, still very high. The number of repeated phrases increased to the following (old values in parentheses):

Pairs: 75 (59)
Ttriplets: 38 (31)
Quadruplets: 26 (22)
Quintuples: 16 (no change)
Sixtuples: 13 (no change)

There was also an increase in the amount of consecutive repetitions of the same word:
Pairs: 74 (67)
Triplets: 5 (4)
Quadruplets: 1 (0)

[Ben Trovato]

Yes, I think so. Can you be more specific?

Well that's hard to tell. It's just a (very subjective) impression that the "typical" Voynich repetitions like this one, taken randomly from p.42

shol.chol.shoky.okol.sho.chol.chhol.chal.shol.chol.chol.shol

have kind of a same "feeling" like Walla's

donatus donatus don augustana don hein hein donatus donatushein

Totally subjective, I know, but I would be interested if one can reproduce this feeling by statistical analysis...

[Ben Trovato]

Yeah, I think it would be a good idea to also examine the writings in German too

So here is some running text from Walla. It covers three pages (ca. 9.000 characters) he wrote on a typewriter (page breaks are indicated by "[@]"). The text contains prayers, charms, a list of "russian" words and a section that mixes up geography, theology and a private diary. And obviously he's having fun playing around with special characters.

*** Edit: I just realized the .txt file does not display special characters (umlauts) in a proper way. I uploaded the file as a word document. ***

[Alin_J]

Yes, I think so. Can you be more specific?

Well that's hard to tell. It's just a (very subjective) impression that the "typical" Voynich repetitions like this one, taken randomly from p.42

shol.chol.shoky.okol.sho.chol.chhol.chal.shol.chol.chol.shol

have kind of a same "feeling" like Walla's

donatus donatus don augustana don hein hein donatus donatushein

Totally subjective, I know, but I would be interested if one can reproduce this feeling by statistical analysis...

Yes, the VM has a lot of quasi-repetition (similar words that are close to each other). There was a very comprehensive analysis of this in the VM in another thread, I think by Marco P, but don't remember which thread. Marco, It would be interesting if you could run the same analysis on Walla's text. I will also start thinking about how to analyse this.

[MarcoP]

Hi Jonas and Benjamin,
thank you both for the interesting thread!

The attached graph is based on the second graph in You are not allowed to view links. Register or Login to view.. As always, I may have messed up something, so take everything with a grain of salt.
You can refer to that post for further details. 
The X axis represents the % of reduplicating words with respect to the total number of consecutive word-couples. The Y axis shows the corresponding measure for quasi-reduplication (more details below).

The reddish circles on the right are about Wallas' corpus:

• cc and cl correspond to Cyrillic and Latin
  
• misc is the rest of the first file (i.e. excluding CC and CL)
  
• walla_1 is the whole first file (CC+CL+MISC)
  
• walla_2 is the second file
  

The blue cloud at the top corresponds to various Voynich (EVA) samples. TTAS is the output of Timm and Schinner's software. All the other samples (close to the origin) are various language texts (from Brian Cham's and Koen's corpora).

Walla consistently exhibits a reduplication rate much higher than the VMS (more than 4%, vs about 1% in the VMS).
The VMS has a higher quasi-reduplication frequency (~2% vs ~1%), according to the specific definition of quasi-reduplication used for these experiments: two consecutive words at least 4 characters long and having edit distance = 1 (i.e. differing by a single character).

These are quasi-reduplication examples in Walla's text:
CL: corporor corboror
CC: foao moao
Walla_2: hokus bokus

Other interesting patterns in Walla's text are not detected by my script. E.g.:
ohrohb feuer satanohrohb satanfeuer

[Ben Trovato]

Thank you very much, Marco!

Now the question to me seems to be:
Does this observation make Walla and the VMS similar or different?

Who wants to take shot?