The Encoder-Decoder Seq2Seq model is composed of two recurrent neural networks (RNNs): an encoder and a decoder. The encoder RNN takes in a sequence of inputs, such as words or characters, and ...

In this paper, we propose a new method for optimizing the encoder-decoder model based on a sequence-level evaluation metric. Since the WER is not directly differentiable, we adopt a policy gradient ...

Natural Language Processing has many interesting applications and Sequence to Sequence modelling is one of those interesting applications. It has major applications in question-answering systems and ...

sequence training of encoder-decoder model using policy gradient for end- to-end speech recognition ...