English

Memory-enhanced Decoder for Neural Machine Translation

Computation and Language 2016-06-08 v1

Abstract

We propose to enhance the RNN decoder in a neural machine translator (NMT) with external memory, as a natural but powerful extension to the state in the decoding RNN. This memory-enhanced RNN decoder is called \textsc{MemDec}. At each time during decoding, \textsc{MemDec} will read from this memory and write to this memory once, both with content-based addressing. Unlike the unbounded memory in previous work\cite{RNNsearch} to store the representation of source sentence, the memory in \textsc{MemDec} is a matrix with pre-determined size designed to better capture the information important for the decoding process at each time step. Our empirical study on Chinese-English translation shows that it can improve by 4.84.8 BLEU upon Groundhog and 5.35.3 BLEU upon on Moses, yielding the best performance achieved with the same training set.

Keywords

Cite

@article{arxiv.1606.02003,
  title  = {Memory-enhanced Decoder for Neural Machine Translation},
  author = {Mingxuan Wang and Zhengdong Lu and Hang Li and Qun Liu},
  journal= {arXiv preprint arXiv:1606.02003},
  year   = {2016}
}

Comments

11 pages

R2 v1 2026-06-22T14:19:13.710Z