ZeroAE: Pre-trained Language Model based Autoencoder for Transductive Zero-shot Text Classification

This is the Pytorch implementation of ZeroAE in the paper: [ZeroAE: Pre-trained Language Model based Autoencoder for Transductive Zero-shot Text Classification]

Figure 1. The network architecture of ZeroAE.

Requirements

• Ubuntu OS
• Python 3.9
• pytorch 1.8.0
• CUDA 11.1

Dependencies can be installed by:

pip install -r requirements.txt

Data preparetion

The first three datasets (Situation, Topic, and Emotion) used in this paper can be downloaded from the following links:

• https://github.com/CogComp/Benchmarking-Zero-shot-Text-Classification

The downloaded datasets can be put in the 'data' directory.

To preprocess the dataset, running:

source init.sh && python ./train/download_topic.py --emb_type bert --init False

Training

To train the model on the topic dataset, run:

python -m torch.distributed.launch --nproc_per_node 4 ./train/train_topic_entail.py

The meaning of each command line argument is explained in train_topic_entail.py, train_situation_entail.py, train_emotion_entail and train_kesu_entail, respectively.

Evaluate

TODO

Citation

@inproceedings{guo-etal-2023-zeroae,
    title = "{Z}ero{AE}: Pre-trained Language Model based Autoencoder for Transductive Zero-shot Text Classification",
    author = "Guo, Kaihao and Yu, Hang and Liao, Cong and Li, Jianguo and Zhang, Haipeng",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
    year = "2023",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.findings-acl.200",
    pages = "3202--3219",
}

Contact

For any questions w.r.t. ZeroAE, please submit them to Github Issues.