Perhaps the following github repo can help you with some ideas:

https://github.com/zaidalyafeai/ARBML

It has some datasets as well as models that you might find useful or interesting.

https://arxiv.org/pdf/1610.05540

https://arxiv.org/pdf/1804.00344

https://www.aaai.org/ocs/index.php/AAAI/AAAI18/paper/download/17282/16060

Thank you Zohra Rezgui and Cristian Ramos-Vera ! I will look into the suggested links.

Mohammed

There is proper documentation for NMT implemented using tensorflow framework in python.

You should try seq-2-seq modelling for translation, any try various combination of RNN layers including LSTM and GRU for training the model.

Find tutorial by tensorflow on this link: https://www.tensorflow.org/tutorials/text/nmt_with_attention

How latent variable is defined for Non auto-regenerative model?

Any suggestion is appreciated.