Machine Translation Using Deep Learning: A Survey

Machine Translation (MT) using Deep Learning is a newly proposed method for machine translation. The term MT is employed within the sense of translation of one language to a different, with no human improvement. It can also be referred to as machine-controlled translation. Unlike the standard statistical machine translation, the neural MT aims at building one neural network that may be collectively tuned to maximize the interpretation performance. This survey reveals the data about Deep Neural Network (DNN) and the concept of deep learning in the field of natural language process i.e. MT. It's higher to use recurrent Neural Network (RNN) in MT. This article studies numerous techniques wont to train RNN for various languages corpses. RNN structure is extremely sophisticated and to train a large corpus is also a time-consuming task. Hence, strong hardware support (Graphics Processing Unit) is needed. GPU improves the system performance by decreasing training period of time. Integration of deep learning in machine learning increase from re-modelling existing features into a statistical system to the development of a new model. Among the different neural network research work use feed forward neural network (FCNN), recurrent neural network (R-CNN), and the encoder- decoder schema.