Skip to content

A machine translation reading list maintained by Tsinghua Natural Language Processing Group

License

Notifications You must be signed in to change notification settings

DengNingyuan/MT-Reading-List

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 

Repository files navigation

Machine Translation Reading List

This is a machine translation reading list maintained by the Tsinghua Natural Language Processing Group.

The past three decades have witnessed the rapid development of machine translation, especially for data-driven approaches such as statistical machine translation (SMT) and neural machine translation (NMT). Due to the dominance of NMT at the present time, priority is given to collecting important, up-to-date NMT papers; the Edinburgh/JHU MT research survey wiki has good coverage of older papers and a brief description for each sub-topic of MT. Our list is still incomplete and the categorization might be inappropriate. We will keep adding papers and improving the list. Any suggestions are welcome!

10 Must Reads

Tutorials and Surveys

Statistical Machine Translation

Word-based Models

Phrase-based Models

Syntax-based Models

Discriminative Training

System Combination

Human-centered SMT

Interactive SMT

Adaptation

Evaluation

Neural Machine Translation

Model Architecture

Attention Mechanism

Open Vocabulary

Training Objectives and Frameworks

Decoding

Low-resource Language Translation

Semi-supervised Learning

Unsupervised Learning

Pivot-based Methods

Data Augmentation Methods

Data Selection Methods

Transfer Learning

Meta Learning

Multilingual Machine Translation

Prior Knowledge Integration

Word/Phrase Constraints

Syntactic/Semantic Constraints

Coverage Constraints

Document-level Translation

Robustness

Interpretability

Linguistic Interpretation

Fairness and Diversity

Efficiency

Pre-Training

Non-Autoregressive Translation

Speech Translation and Simultaneous Translation

Multi-modality

Ensemble and Reranking

Domain Adaptation

Quality Estimation

Human-centered NMT

Interactive NMT

Automatic Post-Editing

Poetry Translation

Eco-friendly

Compositional Generalization

Endangered Language Revitalization

Word Translation

WMT Winners

WMT is the most important annual international competition on machine translation. We collect the competition results on the news translation task since WMT 2016 (the First Conference of Machine Translation) and summarize the techniques used in the systems with the top performance. Currently, we focus on four directions: ZH-EN, EN-ZH, DE-EN, and EN-DE. The summarized algorithms might be incomplete; your suggestions are welcome!

WMT 2019

WMT 2018

  • The winner of ZH-EN: Tencent

    • System report: Mingxuan Wang, Li Gong, Wenhuan Zhu, Jun Xie, and Chao Bian. 2018. Tencent Neural Machine Translation Systems for WMT18. In Proceedings of the Third Conference on Machine Translation: Shared Task Papers.
    • Techniques: RNMT + Transformer + BPE + Rerank ensemble outputs with 48 features (including t2t R2l, t2t L2R, rnn L2R, rnn R2L etc.) + Back Translation + Joint Train with English to Chinese systems + Fine-tuning with selected data + Knowledge distillation
  • The winner of EN-ZH: GTCOM

    • System report: Chao Bei, Hao Zong, Yiming Wang, Baoyong Fan, Shiqi Li, and Conghu Yuan. 2018. An Empirical Study of Machine Translation for the Shared Task of WMT18. In Proceedings of the Third Conference on Machine Translation: Shared Task Papers.
    • Techniques: Transformer + Back-Translation + Data Filtering by rules, language models and translation models + BPE + Greedy Ensemble Decoding + Fine-Tuning with newstest2017 back translation
  • The winner of DE-EN: RWTH Aachen University

    • System report: Julian Schamper, Jan Rosendahl, Parnia Bahar, Yunsu Kim, Arne Nix, and Hermann Ney. 2018. The RWTH Aachen University Supervised Machine Translation Systems for WMT 2018. In Proceedings of the Third Conference on Machine Translation: Shared Task Papers.
    • Techniques: Ensemble of 3-strongest Transformer models + Data Selection + BPE + Fine-Tuning + Important Hyperparameters (batch size and model dimension)
  • The winner of EN-DE: Microsoft

WMT 2017

  • The winner of ZH-EN: Sogou

    • System report: Yuguang Wang, Shanbo Cheng, Liyang Jiang, Jiajun Yang, Wei Chen, Muze Li, Lin Shi, Yanfeng Wang, and Hongtao Yang. 2017. Sogou Neural Machine Translation Systems for WMT17. In Proceedings of the Second Conference on Machine Translation: Shared Task Papers.
    • Techniques: Encoder-Decoder with Attention + BPE + Reranking (R2L, T2S, N-gram language models) + Tagging Model + Name Entity Translation + Ensemble
  • The winner of EN-ZH, DE-EN and EN-DE: University of Edinburgh

    • System report: Rico Sennrich, Alexandra Birch, Anna Currey, Ulrich Germann, Barry Haddow, Kenneth Heafield, Antonio Valerio Miceli Barone, and Philip Williams. 2017. The University of Edinburgh’s Neural MT Systems for WMT17. In Proceedings of the Second Conference on Machine Translation: Shared Task Papers.
    • Techniques: Encoder-Decoder with Attention + Deep Model + Layer Normalization + Weight Tying + Back-Translation + BPE + Reranking(L2R, R2L) + Ensemble

WMT 2016

  • The winner of DE-EN: University of Regensburg

    • System report: Failed to find it
    • Techniques: Failed to find it
  • The winner of EN-DE: University of Edinburgh

About

A machine translation reading list maintained by Tsinghua Natural Language Processing Group

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • TeX 100.0%