Active-to-Passive Arabic Word Conversion and MSD Identification using RNN

Identifiying part of speech of word is critical for Arabic language morphology aspects. Existing approaches either 1) predict morphological description from active voice Arabic words with neural based; or 2) predict morphological description from active and passive voice Arabic words with rule based. Both kinds of approaches have shortcomings. Therefore, we propose on adding some other Arabic type of word, which is passive voice word. Specifically, we convert the active voice to passive voice Arabic with computation and morphological description identification from that result. Experiments show that our system sucessfully to change active to passive voice automatically and achieves good performance on morphological description identification using neural based method.