Is it somehow possible to train the character recognition (Vista) to
recognize specific character sequences? E.g. it is regularly a pain to
correct an ellipsis ("..." / i.e. three consecutive dots). It regularly
delivers ". . ." for these and I need to remove the interspersed blanks..

Or to train specific character shapes? E.g. is grew accustomed to write 0
(digit 0) with a slash through it. the reco then always yields 0/ for it.

Michael