Awesome

Mitigating Gender Bias by Counterfactual Data Augmentation

Citation

This code is for the paper Counterfactual data augmentation for mitigating gender stereotypes in languages with rich morphology featured in ACL 2019. The paper can be found here. Please cite as:

@inproceedings{zmigrod-etal-2019-counterfactual,
    title = "Counterfactual Data Augmentation for Mitigating Gender Stereotypes in Languages with Rich Morphology",
    author = "Zmigrod, Ran  and
      Mielke, Sabrina J.  and
      Wallach, Hanna  and
      Cotterell, Ryan",
    booktitle = "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
    month = jul,
    year = "2019",
    address = "Florence, Italy",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/P19-1161",
    doi = "10.18653/v1/P19-1161",
    pages = "1651--1661"
}

Requirements

• Python version >= 3.6
• Pytorch

Running and Training

We provide pre-trained models for French and Spanish in the models folder. All input files should be in conllu format.

In order to run a pretrained model, use the command.

python src/main.py --in_files [input conllu files] --psi [path to psi .pt file] --reinflect [path to reinflectino model] --animate_list [path to animacy list] --inc_input --get_ids  --out_file [path to output_file] --part 100

In order to train the model, use the following command

python src/neural-mrf.py --data [path to training data] --out_dir [path to output directory]--log_alpha 1 --lr 0.005 --wd 0.0001

You can train the reinflection using reinflection_train.py. This has been lightly modified by the Sigmorphon cross-lingual-baseline. If you use this code please cite the shared task appropriately.