Note that Dasher is currently unable to correctly train itself with 
non-accented characters! We are working on this bug.

Two things are required for Dasher to operate in a non-English language or 
character set:

1) An additional alphabet or set of alphabets. This should be generated from 
the alphabet.xml file. This may just be a matter of adding extra characters
to the end of the letter groupings.

Things to consider: How should characters be grouped? Should accented 
characters be part of the standard character grouping, and if so should they
be segregated at the end or mixed in with other characters?

Note that the alphabet file is UTF8 encoded - non ASCII characters (that is,
pretty much any character except letters without accents, numbers and some 
punctuation) should either be inserted directly in UTF8 (you'll need a special
editor for this) or as an XML entity - an example of this is é for an e
acute character.

2) A new training text. This should be a Unicode encoded file - if you don't
have tools for converting your local character encoding to UTF8, we can do 
this.