Sharing Pre-trained BERT Decoder for a Hybrid Summarization

Ran Wei, Heyan Huang*, Yang Gao

*此作品的通讯作者

科研成果: 书/报告/会议事项章节会议稿件同行评审

7 引用 (Scopus)

摘要

Sentence selection and summary generation are two main steps to generate informative and readable summaries. However, most previous works treat them as two separated subtasks. In this paper, we propose a novel extractive-and-abstractive hybrid framework for single document summarization task by jointly learning to select sentence and rewrite summary. It first selects sentences by an extractive decoder and then generate summary according to each selected sentence by an abstractive decoder. Moreover, we apply the BERT pre-trained model as document encoder, sharing the context representations to both decoders. Experiments on the CNN/DailyMail dataset show that the proposed framework outperforms both state-of-the-art extractive and abstractive models.

源语言英语
主期刊名Chinese Computational Linguistics - 18th China National Conference, CCL 2019, Proceedings
编辑Maosong Sun, Yang Liu, Zhiyuan Liu, Xuanjing Huang, Heng Ji
出版商Springer
169-180
页数12
ISBN(印刷版)9783030323806
DOI
出版状态已出版 - 2019
活动18th China National Conference on Computational Linguistics, CCL 2019 - Kunming, 中国
期限: 18 10月 201920 10月 2019

出版系列

姓名Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
11856 LNAI
ISSN(印刷版)0302-9743
ISSN(电子版)1611-3349

会议

会议18th China National Conference on Computational Linguistics, CCL 2019
国家/地区中国
Kunming
时期18/10/1920/10/19

指纹

探究 'Sharing Pre-trained BERT Decoder for a Hybrid Summarization' 的科研主题。它们共同构成独一无二的指纹。

引用此

Wei, R., Huang, H., & Gao, Y. (2019). Sharing Pre-trained BERT Decoder for a Hybrid Summarization. 在 M. Sun, Y. Liu, Z. Liu, X. Huang, & H. Ji (编辑), Chinese Computational Linguistics - 18th China National Conference, CCL 2019, Proceedings (页码 169-180). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); 卷 11856 LNAI). Springer. https://doi.org/10.1007/978-3-030-32381-3_14