Back to Nlp Progress

Machine translation

english/machine_translation.md

0.34.4 KB
Original Source

Machine translation

Machine translation is the task of translating a sentence in a source language to a different target language.

Results with a * indicate that the mean test score over the the best window based on average dev-set BLEU score over 21 consecutive evaluations is reported as in Chen et al. (2018).

WMT 2014 EN-DE

Models are evaluated on the English-German dataset of the Ninth Workshop on Statistical Machine Translation (WMT 2014) based on BLEU.

ModelBLEUPaper / Source
Transformer Big + BT (Edunov et al., 2018)35.0Understanding Back-Translation at Scale
DeepL33.3DeepL Press release
Admin (Liu et al., 2020)30.1Very Deep Transformers for Neural Machine Translation
MUSE (Zhao et al., 2019)29.9MUSE: Parallel Multi-Scale Attention for Sequence to Sequence Learning
DynamicConv (Wu et al., 2019)29.7Pay Less Attention With Lightweight and Dynamic Convolutions
TaLK Convolutions (Lioutas et al., 2020)29.6Time-aware Large Kernel Convolutions
AdvSoft + Transformer Big (Wang et al., 2019)29.52Improving Neural Language Modeling via Adversarial Training
Transformer Big (Ott et al., 2018)29.3Scaling Neural Machine Translation
RNMT+ (Chen et al., 2018)28.5*The Best of Both Worlds: Combining Recent Advances in Neural Machine Translation
Transformer Big (Vaswani et al., 2017)28.4Attention Is All You Need
Transformer Base (Vaswani et al., 2017)27.3Attention Is All You Need
MoE (Shazeer et al., 2017)26.03Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer
ConvS2S (Gehring et al., 2017)25.16Convolutional Sequence to Sequence Learning

WMT 2014 EN-FR

Similarly, models are evaluated on the English-French dataset of the Ninth Workshop on Statistical Machine Translation (WMT 2014) based on BLEU.

ModelBLEUPaper / Source
DeepL45.9DeepL Press release
Transformer Big + BT (Edunov et al., 2018)45.6Understanding Back-Translation at Scale
Admin (Liu et al., 2020)43.8Understand the Difficulty of Training Transformers
MUSE (Zhao et al., 2019)43.5MUSE: Parallel Multi-Scale Attention for Sequence to Sequence Learning
TaLK Convolutions (Lioutas et al., 2020)43.2Time-aware Large Kernel Convolutions
DynamicConv (Wu et al., 2019)43.2Pay Less Attention With Lightweight and Dynamic Convolutions
Transformer Big (Ott et al., 2018)43.2Scaling Neural Machine Translation
RNMT+ (Chen et al., 2018)41.0*The Best of Both Worlds: Combining Recent Advances in Neural Machine Translation
Transformer Big (Vaswani et al., 2017)41.0Attention Is All You Need
MoE (Shazeer et al., 2017)40.56Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer
ConvS2S (Gehring et al., 2017)40.46Convolutional Sequence to Sequence Learning
Transformer Base (Vaswani et al., 2017)38.1Attention Is All You Need

WMT22 shared task on Large-Scale Machine Translation Evaluation for African Languages

ModelBLEUPaper / Source
vanilla MNMT models17.95Tencent’s Multilingual Machine Translation System for WMT22 Large-Scale African Languages

Go back to the README