Skip to content

Large and Distilled Models for Quality Estimation of Machine Translation

License

Notifications You must be signed in to change notification settings

sheffieldnlp/deepQuest-py

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

91 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

deepQuest-py

deepQuest-py is a framework for training and evaluation of models for Quality Estimation of Machine Translation. This is a new version of deepQuest - the first framework for neural Quality Estimation.

deepQuest-py provides:

  • high performing sentence-level and word-level models based on finetuning pre-trained Transformers;
  • light-weight and efficient sentence-level models implemented via knowledge distillation.

deepQuest-py includes implementations of several approaches for Quality Estimation proposed in recent research:

See our examples for instructions on how to train and test specific models.

Online Demo

Check out our web tool to try out most of our trained models on your own data!

Installation

deepQuest-py requires Python 3.6 or later.

git clone https://github.com/sheffieldnlp/deepQuest-py.git
cd deepQuest-py
pip install -e .

Licence

deepQuest-py is licenced under a CC BY-NC-SA licence.

Citation

If you use deepQuest-py in your research, please cite our EMNLP 2021 Demo paper:

@inproceedings{alva-manchego-etal-2021-deepquest,
    title = "deep{Q}uest-py: {L}arge and Distilled Models for Quality Estimation",
    author = "Alva-Manchego, Fernando  and
      Obamuyide, Abiola  and
      Gajbhiye, Amit  and
      Blain, Fr{\'e}d{\'e}ric  and
      Fomicheva, Marina  and
      Specia, Lucia",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
    month = nov,
    year = "2021",
    address = "Online and Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.emnlp-demo.42",
    pages = "382--389",
}

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •