Arabic words (non-ascii characters) training data

[yoosif0]

Currently the training data looks like this

nawar-spec-00001.npy|nawar-mel-00001.npy|1223| وَرَجَّحَ التَّقْرِيرُ الَّذِي أَعَدَّهُ مَعْهَدُ أَبْحَاثِ هَضَبَةِ التِّبِتِ فِي الْأَكَادِيمِيَّةِ الصِّينِيَّةِ لِلْعُلُومِ - أَنْ تَسْتَمِرَّ دَرَجَاتُ الْحَرَارَةِ وَمُسْتَوَيَاتُ الرُّطُوبَةِ فِي الْإِرْتِفَاعِ طَوَالَ هَذَا الْقَرْنْ

Do you think that using phonetised words instead of arabic words would make training easier?

An example of training data of phonetised words is shown below which is similar to cumdict phonetised words

nawar-spec-00001.npy|nawar-mel-00001.npy|1223| W R A0 J A E A L T AE Q R E2 ..........

Thank you

[keithito]

I'm not sure. It would depend on how closely the Arabic alphabet maps to phonemes.