A key to understand the VMS

[Torsten]

I think that the images do present some hints of clustering, but the clusters tend to appear as vertical regions in the diagrams, while Torsten's colors tend to be distributed horizontally.
The orange 'i' region seems to be somehow separated from the others.
The region I numbered 3 seems to correspond to sh- and ch- words.

Hello MarcoP, 

the gap between 2 and 3 in your picture only exists since EVA is using two signs for writing "ch". If "ch" would be transcribed with "2" and "sh" with "3" such a gap would not exist.

The word types similar to "daiin" are indeed separated.

The colors are chosen for the suffixes of the word types. If you would use the prefixes "d", "ch", "qo" for coloring the graphs would be distributed vertically.  
In the following graph word types starting with "c" or "sh" words starting with "q" or "o" are purple, words starting with "d" are orange and words starting with "s" are red: