#5 docs: include advice for non-English languages

[Scott James Cederberg]

From Beate:

Another thing: We may want to let people know that for
preprocessing we
only consider characters A-Z and a-z, and that if
someone was to build a
model from a German or French corpus, they'd have to
change the my_isalpha
function to include the ASCII code of German Umlauts or
e accent aigu,
grave, etc. for French.