TF-IDF Cosine matching algorithm using ZL.2a Transcription visualized by Gephi v0.92

11 images in 1 GIF file
colors indicate the Scribe
all points are folios
curved lines are edges

img1 uses only the strongest edge weights;  adding weaker edges until  img11  which shows all the edges.

vocab is missing 200 words...grrr..rare chars and single chars is probably the problem.

Key:
Scribe No.  Color        Total pages 
Scribe1      Green          113pages       
Scribe2      Blue              46pages     
Scribe3      Red              33pages   
Scribe4      Pink              27pages
Scribe5      Yellow           07pages

Outliers at top:
f96v( green ), f68r2( pink ), f68r1( pink ), f65r( red )
Outlier at bottom:
v67v2( pink )

Gif by ezgif

Nice ! Can you list the pages for scribe 1, 2, 3? I don't have the same numbers.

The beauty of this visualisation alone is worth appreciating. There seems to be a bit of red vs blue clustering in the right hemisphere?

(02-05-2022, 06:19 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.There seems to be a bit of red vs blue clustering in the right hemisphere?

Yes, it shows just how tighly knitted together Scribe_2 and Scribe_3's words are.

Also you can see, even by img11 that if a Scribe_1(green) folio was taken at random, its nearest neighbor would most likely also be green and it would be within a mesh of green edges(words)
thus visually demonstrating, in a limited way, what Kevin Farrugia found,
You are not allowed to view links. Register or Login to view.
that separating Scribe_1, from Scribes 2 and 3 can be done most of the time, whilst successfully separating Scribe_2 from Scribe_3 is much harder.

Interestingly ( if i interpet these images correctly    ) then Scribe_4 (pink) should also be able to be separated out from the other scribes more often than not.

Hi nablator, sorry for the 3 year delay    ... woops 

Scribe1
Nbr. folios 113

Scribe2
Nbr. folios 46

Scribe3
Nbr. folios 33

Comment:
# pharma, f101r,  {f102r empty} ,  f101v.
# <f101r2>      {$I=P $Q=S $P=F} -- ZL
# All text on this page in normal paragraphs covered by f101r1 --ZL
# f101r2 is empty

(03-05-2022, 02:02 PM)RobGea Wrote: You are not allowed to view links. Register or Login to view.separating Scribe_1, from Scribes 2 and 3 can be done most of the time, whilst successfully separating Scribe_2 from Scribe_3 is much harder

While much harder to quantify, this is also true for VM plants.
You are not allowed to view links. Register or Login to view.

A plants on pages written by by Scribe 1 can be distinguished from B plants on pages by Scribes 2,3,5 with relative ease. But distingushing plants from Scribes 2 and 3 is nearly impossible. Arguably the sample size is much smaller but still they have a very similar 'vibe'