Refer to:
Graham Taylor's implementation:
https://github.com/gwtaylor/theano-rnn
Mohammad Pezeshki's implementation:
https://github.com/mohammadpz/Recurrent-Neural-Networks
Kyunghyun Cho's implementation:
https://github.com/kyunghyuncho/dl4mt-material
Paper:
Neural Machine Translation by Jointly Learning to Align and Translate