Transliteration and interpretation issues

[ReneZ]

Isn't there a problem with glyph-level (as opposed to "word" level) attacks that simply use EVA as input? For example those that detect consonant-vowel alternation patterns?

Yes, there is, but the same problem exists with all other transliterations.

This is also why lately I have used several alphabets in parallel, as this allows to show the impact of this.
The best way to handle this problem in general is to realise (and state) that results are based on certain assumptions or conditions.

A typical case is the word length distribution. The result depends on the alphabet used, so it should be stated.

Word-level statistics (e.g. repeating strings) also depend on the transliteration used. In v101 you will find fewer than in extended Eva, while in Currier or basic Eva you will find more.