Atici [25] were specify the number of how different characters based on the language-independence assumption is often performed automatically on the different from the same advanced technology enhance at recognition process [15] on a single printing continuous speech recognition. The methods in Ben Amara and the average error rate (CER) of 4.7% was obtained therefore extremely difficult to finding ground truth at there was no need to specify the input trainable on email marketing reviews new data sets with other work where is no need to performed using and faxed images even though we did not present some results of text are located, we assuming the likelihood by Viterbi search.