Refining Arabic Language OCR For Easing Of Documents Summarization

Optical character recognition (OCR) is populated after the wide enchantment on the deep learning and image processing. Depending on the amount of train data, deep learning based OCR can yield the expected reliability. This paper focuses on step ahead after applying the OCR into Arabic hand written dataset. It dose keen on summarization of the OCR results through the use of association rues technology. Three algorithms were deployed to recognize the Arabic text namely: Convolutional neural network (CNN), Recurrent with long short memory neural network (RNN-LSTM). Hidden makov Model (HMM) is also used as third algorithm for Arabic text recognition. Results shown that RNN-LSTM is outperformed over the other algorithms by producing accuracy of recognition of 92%.