See this post of mine, for example.
And @chrillek, I’m not sure keeping Peruvians in the ABBYY OCR library is acceptable; it’s certainly something DEVONtech should be actively committed to stopping, rather than hoping for the next version to rumble around trying to solve it itself.