Contextual Encoding for Translation Quality Estimation
Implemented by Junjie Hu
Contact: junjieh@cs.cmu.edu
If you use the codes in this repo, please cite our WMT18 paper.
@InProceedings{hu-EtAl:2018:WMT,
author = {Hu, Junjie and Chang, Wei-Cheng and Wu, Yuexin and Neubig, Graham},
title = {Contextual Encoding for Translation Quality Estimation},
booktitle = {Proceedings of the Third Conference on Machine Translation: Shared Task Papers},
month = {October},
year = {2018},
address = {Belgium, Brussels},
publisher = {Association for Computational Linguistics},
pages = {788--793},
url = {http://www.aclweb.org/anthology/W18-6462}
}
Install
sklearn
pytorch >= 0.4.0
scipy
scipy.stats
conda env create --file wmt-qe-py3.5-pt4.0.txt
Download Data
cd data/WMT2018QE/
bash download_data.sh
Train Model
bash scripts/WMT2018QE/de-en/train-word-ClBasic-final.sh [GPU ID]