App should offer to redo the OCR if the result was very, suggesting a different language

[renard314]

If it was possible for the algorithm to give a "quality" of the matched text, and then offer to change the tesseract language parameter and redo the OCR. The quality could also be calculated afterwards by checking the word distribution against measures for standard text (% of "real" words in the chosen language, using simple 1k-2k word base samples)