Neural machine translation (NMT) uses deep neural networks to perform machine translation from the source language to the target language. The neural translation machine takes in text in the source language as a sequence of inputs and encodes these to a hidden representation, which is then decoded back to produce the translated text sequence in the target language. One of the key advantages of this NMT system is that the whole machine translation system can be trained from end-to-end together, unlike the rule-based and statistical machine translation systems. Generally, RNN architectures such as LSTMs (long short term memory) and/or gated recurrent units (GRUs) are used in the neural translation machine architecture.
A few of the advantages of NMTs over other traditional methods are as follows:
- All of the parameters of an NMT model are trained end...