Home

Awesome

<!-- * @Company: Sogou MT * @FilePath: /Cross-lingual-Summarization/README.md * @LastEditors: Mingming-Yin * @LastEditTime: 2020-10-22 23:23:40 -->

Cross-Lingual Abstractive Sentence Summarization(CL-ASSUM)

Introduction

We implemented CL-ASSUM on fairseq. In this repo, it contains of four parts.

Teacher models

Before staring the experiment, you should first use Transformer to train the teacher model of NMT model and momolingual summarization model.

Please refer to Transformer for more deatils.

Teaching-Generation

Teaching-Attention

Teaching-Generation-Attention

Evaluation sets of CL-ASSUM

The test-data file contains evaluation sets of CL-ASSUM, which is built by manual translation.

Requirements and Installation

Cross-Lingual Test Set

In our experiments, we manually translate the English sentences into the Chinese sentences for the validation and evaluation sets of Gigaword and DUC2004.

License

Reference

If you find CL-ASSUM useful in your work, you can cite this paper as below:

@inproceedings{duan-etal-2019-zero,
    title = "Zero-Shot Cross-Lingual Abstractive Sentence Summarization through Teaching Generation and Attention",
    author = "Duan, Xiangyu  and Yin, Mingming  and Zhang, Min  and Chen, Boxing  and Luo, Weihua",
    booktitle = "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
    month = jul,
    year = "2019",
    address = "Florence, Italy",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/P19-1305",
    doi = "10.18653/v1/P19-1305",
    pages = "3162--3172",
   }