Neural Machine Translation: A Review and Survey
This is an incremental review for researchers in machine translation, summarizing existing knowledge without new results.
The paper reviews the shift from statistical to neural machine translation, tracing the development of NMT architectures and surveying recent trends in the field.
The field of machine translation (MT), the automatic translation of written text from one natural language into another, has experienced a major paradigm shift in recent years. Statistical MT, which mainly relies on various count-based models and which used to dominate MT research for decades, has largely been superseded by neural machine translation (NMT), which tackles translation with a single neural network. In this work we will trace back the origins of modern NMT architectures to word and sentence embeddings and earlier examples of the encoder-decoder network family. We will conclude with a survey of recent trends in the field.