An Approach for Arabic Diacritization

摘要

Modem Standard Arabic (MSA) contains optional diacritical marks (diacritics, in Arabic harakat), which became less used in Arabic books, newspapers and other written media. Diacritics are very important for readability and understandability of texts. Their absence causes critical problems that add to the lexical, morphological and semantic ambiguities. In this paper, we present an automatic diacritization system of the Arabic language, using Hidden Markov Models with the Viterbi's algorithm, based on probabilities based on learning on diacritized Arabic texts. The corpus used was mostly composed of religious texts. Our results were satisfactory, achieving a precision of up to 80% at the word level.