The second employs an unsupervised encoder-decoder, and the third combines supervised and unsupervised mechanisms using a parallel corpus of Egyptian Arabic to standard Arabic.