Automatic Persian-Arabic Phonetic mapping

Abstract

Language is the most essential means of human communication and comes in several forms, the most significant being sound. Studying the phonetic relationship between different languages helps in building models that process and understand these interlingual connections. Phonetic mapping refers to converting the phonetic of words from one language to another. The main objective of this research is to develop a framework for phonetic mapping from Persian to Arabic. We created a bilingual Persian-Arabic phonetic dataset and applied a statistical model to identify shared phonetic elements. Additionally, we used a Hidden Markov Model (HMM) and developed a rule-based approach to refine the dataset and derive Arabic phonetic representations from Persian. The proposed model was evaluated based on accuracy, phonetic error rate (PER), and word error rate (WER). Using the rule-based approach, the accuracy of phonetic mapping from Persian to Arabic reached 85.6%.