R²F: A General Retrieval, Reading and Fusion Framework for Document-level Natural Language Inference

This is the repository for our EMNLP 2022 paper: R²F: A General Retrieval, Reading and Fusion Framework for Document-level Natural Language Inference

Prepare Dataset

Please download DOCNLI dataset.

Our complementary sentence-level annotation file is at here.

Run Model

To reproduce our results, please set appropriate file path parameters, and set do_train, do_eval, or do_predict as True for model training, evaluation, or prediction. Then for rouge retrieval (similar for other retrieval methods), please run

python rouge_retrieval_base.py

To conduct sentence-level evalaution, please set appropriate file path parameters. Then for rouge retrieval (similar for other retrieval methods), please run

python rouge_retrieval_base_sentence_evaluation.py

Checkpoint Files

Our checkpoint files for base encoder and large encoder are also released.

Contact

If you have any question about our work, please feel free to contact us at hao.wang@nudt.edu.cn.

Citation

Please cite our work as {
title={R2F: A General Retrieval, Reading and Fusion Framework for Document-level Natural Language Inference},
author={Hao Wang, Yixin Cao, Yangguang Li, Zhen Huang, Kun Wang, Jing Shao},
booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
year={2022}
}