User:Phe/Test6

Lower the rate, worst the ocr. Ocr rate is the percentage of the cumulated word length for all word of length >= 3 by the cumulated word length for these word which are recognized by a standard dictionary. Roughly < 0.70, large portion of ocr are unreadable. 0.70-0.80 many small portion of text unreadable. 0.80-0.90 many errors but not much clustered, most text readable with some efforts. > 0.90 good to excellent ocr. Variation between 0.92 to 0.88 imply much larger number of error, no idea exactly, but roughly a 0.88 will contain 5 to 10 times more error than a 0.92. These qualification of ocr rate is correct only for EB1911.