The LSTM-based encoder extracts a dense representation of the content in the source language, and the LSTM-based decoder generates the output sequence in the target language. And, of course, the ...