Neural Machine Translation from English to Belarusian using Seq2Seq Learning The actual data contains 3883 phrase pairs of English and Belarusian. The data was pre-processed and split into the train and test sets. The model used is an encoder-decoder LSTM model. The architecture of the model is as follows: