Modeling Vocal Knowledge for Speech Synthesis of Standard Arabic