Code for our ACL 2021 paper - Modeling Discriminative Representations for Out-of-Domain Detection with Supervised Contrastive Learning

Requirements

python 3.6.2
torch 1.4.0
tensorflow 1.14.0

For detailed dependencies, please refer to requirements.txt

Get Started

prepare Glove or BERT pretrained embeddings from https://github.com/stanfordnlp/GloVe and https://github.com/google-research/bert

put the embedding file glove.6B.300d.txt into ./glove_embeddings

modify the script to train or test model in different modes

bash run.sh

Citation

@inproceedings{Zeng2021ModelingDR,
  title={Modeling Discriminative Representations for Out-of-Domain Detection with Supervised Contrastive Learning},
  author={Zhiyuan Zeng and Keqing He and Yuanmeng Yan and Zijun Liu and Yanan Wu and Hong Xu and Huixing Jiang and Weiran Xu},
  booktitle={ACL/IJCNLP},
  year={2021}
}