A Deep Belief Network Classification Approach for Automatic Diacritization of Arabic Text

Deep learning has emerged as a new area of machine research. It is an approach that can learn features and hierarchical representation purely from data been successfully applied to several fields such images, sounds, text motion. The techniques developed deep research have already impacting the on Natural Language Processing (NLP). Arabic diacritics are vital components remove ambiguity words reinforce meaning text. In this paper, Belief Network (DBN) used diacritizer for DBN algorithm among recently proved be very effective variety problems. We evaluate use DBNs classifiers in automatic diacritization. was trained individually classify each input letter with corresponding diacritized version. Experiments were conducted using two benchmark datasets, LDC ATB3 Tashkeela. Our best settings achieve DER WER 2.21% 6.73%, receptively, improvement 26% over published results. On Tashkeela benchmark, our system continues high accuracy 1.79% 14% improvement.