Real-world example: OCR tended to turn "war" into "wan", which the spellchecker has no problem with.

I used Tesseract OCR, which other reports here have suggested has the best quality among the three choices ogmrip can use.