This repository contains the code and data for our ACL 2021 paper:

Meta-Learning Adversarial Domain Adaptation Network for Few-Shot Text Classification

If you find this work useful and use it on your own research, please cite our paper.

@inproceedings{MLADA:conf/acl/HanFZQGZ21,
  author    = {Chengcheng Han and
               Zeqiu Fan and
               Dongxiang Zhang and
               Minghui Qiu and
               Ming Gao and
               Aoying Zhou},
  title     = {Meta-Learning Adversarial Domain Adaptation Network for Few-Shot Text
               Classification},
  booktitle = {Findings of the Association for Computational Linguistics: {ACL/IJCNLP}
               2021, Online Event, August 1-6, 2021},
  series    = {Findings of {ACL}},
  volume    = {{ACL/IJCNLP} 2021},
  pages     = {1664--1673},
  publisher = {Association for Computational Linguistics},
  year      = {2021},
}

We propose an adversarial domain adaptation network to enhance meta-learning framework, with the objective of improving the model’s adaptive ability for new tasks in new domains.

We first utilize two neural networks competing against each other, separately playing the roles of a domain discriminator and a meta-knowledge generator. The adversarial network is able to strengthen the adaptability of the meta-learning architecture.

Moreover, we aggregate transferable features generated by the meta-knowledge generator with sentence-specific features to produce high-quality sentence embeddings.

Finally, we utilize a ridge regression classifier to obtain final classification results.

The figure gives an overview of our model.

We ran experiments on a total of 4 datasets. You may unzip our processed data file data.zip and put the data files under data/ folder.

Please download pretrained word embedding file wiki.en.vec from here and put it under pretrain_wordvec/ folder.

After you have finished configuring the data/ folder and the pretrain_wordvec/ folder, you can run our model with the following commands.

cd bin
sh mlada.sh

You can also adjust the model by modifying the parameters in the malada.sh file.

• Python 3.7
• PyTorch 1.6.0
• numpy 1.18.5
• torchtext 0.7.0
• termcolor 1.1.0
• tqdm 4.46.0
• CUDA 10.2