Deep Learning for Unsupervised Neural Machine Translation
Kuai Yu · 2021
With the breakthrough of deep learning techniques, many Natural Language Processing tasks have exploited the techniques to enhance performance. Neural Machine Translation, as a sub-field of NLP, also leverages deep learning methods to enhance performance. There are two main categories of NMT, the first one is supervised, and the other one is unsupervised. Supervised NMT uses labelled data and large parallel corpus for training, while unsupervised NMT only uses independent monolingual corpora for training the model. The latter one gives many conveniences in data collection but poses a great difficulty in engineering the architecture. Thus, in this paper, we give a comprehensive review of deep learning techniques for unsupervised neural machine translation.as illustrated by the portions given in this document.