DELE_pytorch

This is the official implementation of the TNNLS2022 work "Description-Enhanced Label Embedding Contrastive Learning for Text Classification"

Requirement

1. pytorch >= 1.7.1
2. photinia
3. Transformer

Data Preparation

1. SNLI [https://nlp.stanford.edu/projects/snli/]
2. SICK [https://huggingface.co/datasets/sick]
3. SciTail [https://allenai.org/data/scitail]
4. Quora Question Pair [https://huggingface.co/datasets/quora]
5. MSRP [https://www.microsoft.com/en-us/download/details.aspx?id=52398]
6. Yahoo Answer [https://huggingface.co/datasets/yahoo_answers_topics]
7. SST [https://nlp.stanford.edu/sentiment/treebank.html]

Note that most of the datasets can be obtained from the huggingface dataset

Bibtex

If you use this dataset or code in your work, please cite our AAAI2021 paper and our_TNNLS2022_paper:

@inproceedings{zhang2021making,
  title={Making the relation matters: Relation of relation learning network for sentence semantic matching},
  author={Zhang, Kun and Wu, Le and Lv, Guangyi and Wang, Meng and Chen, Enhong and Ruan, Shulan},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={35},
  number={16},
  pages={14411--14419},
  year={2021}
}

@article{zhang2022descrip,
  author={Zhang, Kun and Wu, Le and Lv, Guangyi and Chen, Enhong and Ruan, Shulan and Liu, Jing and Zhang, Zhiqiang and Zhou, Jun and Wang, Meng},
  journal={IEEE Transactions on Neural Networks and Learning Systems}, 
  title={Description-Enhanced Label Embedding Contrastive Learning for Text Classification}, 
  year={2023},
  pages={1-14},
  doi={10.1109/TNNLS.2023.3282020}
}