The vowel exchange "a" to "o"

[Jorge_Stolfi]

It seems that the a and o can be exchanged flexibly without a ‘rule’

Beware that transcribers often have to guess whether a glyph is an a or an o (or a y that lost its tail).  Thus some of that "flexibility" may be just transcription errors.

And the (ahem! ahem!) scribes who restored and retouched faded text faced the same problem.  I think I can see many cases where they could not make out the original and just guessed o or a.  Or intentionally drew am intermediate glyph shape that could be read as either of them...

Thus, in statistical analyses, it may be safer to map "confusible" letters to the same letter: {a,o,y} -> o, {r,s} -> r, {ch,ih} -> ee, etc.  The analysis may then miss some structure, but it would be less affected by such restoration and transcription noise.

All the best, --stolfi