RobustNMT-ISDST

Code for "Towards Robust Neural Machine Translation with Iterative Scheduled Data-Switch Training" (COLING 2022, Oral)

Requirements

• Python 3.7.0
• CUDA 11.6
• Pytorch 1.12.0
• Fairseq 0.12.2

Quickstart

Here, we take IWSLT14 De-En translation task as an example.

Step1: Preprocess

# Download and clean the raw data
bash examples/translation/prepare-iwslt14.sh

# Preprocess/binarize the data
mkdir -p data-bin
fairseq-preprocess --source-lang de --target-lang en \
    --trainpref /path/to/iwslt14_deen_data/train \
    --validpref /path/to/iwslt14_deen_data/valid \
    --testpref /path/to/iwslt14_deen_data/test \
    --destdir data-bin/iwslt14.tokenized.de-en \
    --workers 20

Step2: Training

Our training processes are conducted in an iterative manner. The iterative number K is set as 5 for IWSLT14 De-En translation task.

bash train_iwslt14-de-en.sh

Step3: Evaluation

# generate translations
fairseq-generate data-bin/iwslt14.tokenized.de-en \
--path /path/to/stage5/checkpoint_best.pt \
--beam 5 --remove-bpe > res.out
# calculate BLEU score
bash scripts/compound_split_bleu.sh res.out

Citation

@inproceedings{miao2022towards,
  title={Towards Robust Neural Machine Translation with Iterative Scheduled Data-Switch Training},
  author={Miao, Zhongjian and Li, Xiang and Kang, Liyan and Zhang, Wen and Zhou, Chulun and Chen, Yidong and Wang, Bin and Zhang, Min and Su, Jinsong},
  booktitle={Proceedings of the 29th International Conference on Computational Linguistics},
  pages={5266--5277},
  year={2022}
}