The Textual Work of August Walla, mentally disabled artist

[Ben Trovato]

So I was thinking about how to prepare the text sample. To demonstrate the scheme, I take the "SAÄRILL"-picture posted above. The text file would look like this:

[1-Np98#163-????-G0]
Walla August.!
[@]BIN DER HEILIGE
GOTT SAÄRILL.!
[@]GOTT,
SAÄRILL.!
[@]I 1[@]II 2[@]I 1[@]III 3
[@]Sonne hinterm
Berg Bisasmberg
in den Bezirk
Korneuburg an
der Donau?
[@]KPÖ
[@]GOTT
ALLAH.
[@] SARARILL
UND, ALLAH.!

Line 1: squared brackets are not used in Walla's work, so I can take them to indicate metadata: here, a consecutive number, source (book/page/figure number), the year of origin, ???? unknown in this case, and the "genre". I would propose to have the "genres" of "understandable" text (like this one), fake languages (like "Turkish" and "latin" in the samples in my initial post, samples #7 and #8) and plain gibberish (sample #6). My goal would be to collect text from the last two genres.

The transcription moves from top-left to bottom-right. Line breaks are line breaks. [@] indicates that the text is interrupted by parts of the drawings, while [@] plus a line break means that the text is from a different area in the picture. (@ is just another symbol Walla did not happen to know...)

Then I would simply type this into a text file, one after another...

@Alin_J, does that sound reasonable for you, or should things be done differently? (And what about the other mighty wizards of computational science?) 

I hope I get some books next week, so I could start working on this...