Menu
Home
People
Places
Arts
History
Plants & Animals
Science
Life & Culture
Technology
Reference.org
Google Neural Machine Translation
open-in-new
<p>Google Neural Machine Translation (GNMT) was a <a href="/facts/Neural_machine_translation/BsjdmTSt">neural machine translation</a> (NMT) system developed by Google and introduced in November 2016 that used an <a href="/facts/Artificial_neural_network/6V1jMlkx">artificial neural network</a> to increase fluency and accuracy in <a href="/facts/Google_Translate/AR0cRb9n">Google Translate</a>. The neural network consisted of two main blocks, an encoder and a decoder, both of <a href="/facts/LSTM/NBBVEoft">LSTM</a> architecture with 8 1024-wide layers each and a simple 1-layer 1024-wide <a href="/facts/Feedforward_neural_network/CP0pPGDF">feedforward</a> <a href="/facts/Attention_(machine_learning)/AQ566C3b">attention mechanism</a> connecting them. The total number of parameters has been variously described as over 160 million, approximately 210 million, 278 million or 380 million. It used WordPiece <a href="/facts/Lexical_analysis/T1JYWpIf">tokenizer</a>, and <a href="/facts/Beam_search/EFPaqKkM">beam search</a> decoding strategy. It ran on <a href="/facts/Tensor_Processing_Unit/aVlg0LFF">Tensor Processing Units</a>. </p><p>By 2020, the system had been replaced by another deep learning system based on a Transformer encoder and an RNN decoder. </p><p>GNMT improved on the quality of translation by applying an <a href="/facts/Example-based_machine_translation/GhXgEDyU">example-based</a> (EBMT) <a href="/facts/Machine_translation/DGF3NwuI">machine translation</a> method in which the system learns from millions of examples of language translation. GNMT's proposed architecture of system learning was first tested on over a hundred languages supported by Google Translate. With the large end-to-end framework, the system learns over time to create better, more natural translations. GNMT attempts to translate whole sentences at a time, rather than just piece by piece. The GNMT network can undertake <a href="/facts/Interlingual_machine_translation/NNCdra1F">interlingual machine translation</a> by encoding the semantics of the sentence, rather than by memorizing phrase-to-phrase translations. </p>