Ah, a good idea. But unfortunatly the strings can be more complicated. For example a string could look like

"ABCÅDEFÄGHI"

or even tricky combinations like

"ABCÅÄ"

where the analyzer can not be sure if it should be interpreted as

{'Å'=>'ë', 'Ä'=>'ë'} or 
{'Å'=>'', 'Ä'=>'ëë'} or 
{'Å'=>'ëë', 'Ä'=>''}

If at all possible I think this kind of ambiguity has to be resolved when encountering similar strings with just one "interpretation"

/L