Comprehensive Document Summarization with Refined Self-Matching Mechanism

Under the constraint of memory capacity of the neural network and the document length, it is difficult to generate summaries with adequate salient information. In this work, the self-matching mechanism is incorporated into the extractive summarization system at the encoder side, which allows the enc...

Full description

Bibliographic Details
Main Authors: Biqing Zeng, Ruyang Xu, Heng Yang, Zibang Gan, Wu Zhou
Format: Article
Language:English
Published: MDPI AG 2020-03-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/10/5/1864
id doaj-3ca05b7310bb4cc39184fad63c0a6fbf
record_format Article
spelling doaj-3ca05b7310bb4cc39184fad63c0a6fbf2020-11-25T03:01:45ZengMDPI AGApplied Sciences2076-34172020-03-01105186410.3390/app10051864app10051864Comprehensive Document Summarization with Refined Self-Matching MechanismBiqing Zeng0Ruyang Xu1Heng Yang2Zibang Gan3Wu Zhou4School of Software, South China Normal University, Foshan 528225, ChinaSchool of Computer, South China Normal University, Guangzhou 510631, ChinaSchool of Computer, South China Normal University, Guangzhou 510631, ChinaSchool of Software, South China Normal University, Foshan 528225, ChinaSchool of Computer, South China Normal University, Guangzhou 510631, ChinaUnder the constraint of memory capacity of the neural network and the document length, it is difficult to generate summaries with adequate salient information. In this work, the self-matching mechanism is incorporated into the extractive summarization system at the encoder side, which allows the encoder to optimize the encoding information at the global level and effectively improves the memory capacity of conventional LSTM. Inspired by human coarse-to-fine understanding mode, localness is modeled by Gaussian bias to improve contextualization for each sentence, and merged into the self-matching energy. The refined self-matching mechanism not only establishes global document attention but perceives association with neighboring signals. At the decoder side, the pointer network is utilized to perform a two-hop attention on context and extraction state. Evaluations on the CNN/Daily Mail dataset verify that the proposed model outperforms the strong baseline models and statistical significantly.https://www.mdpi.com/2076-3417/10/5/1864extractive summarizationdeep learningdocument summarization
collection DOAJ
language English
format Article
sources DOAJ
author Biqing Zeng
Ruyang Xu
Heng Yang
Zibang Gan
Wu Zhou
spellingShingle Biqing Zeng
Ruyang Xu
Heng Yang
Zibang Gan
Wu Zhou
Comprehensive Document Summarization with Refined Self-Matching Mechanism
Applied Sciences
extractive summarization
deep learning
document summarization
author_facet Biqing Zeng
Ruyang Xu
Heng Yang
Zibang Gan
Wu Zhou
author_sort Biqing Zeng
title Comprehensive Document Summarization with Refined Self-Matching Mechanism
title_short Comprehensive Document Summarization with Refined Self-Matching Mechanism
title_full Comprehensive Document Summarization with Refined Self-Matching Mechanism
title_fullStr Comprehensive Document Summarization with Refined Self-Matching Mechanism
title_full_unstemmed Comprehensive Document Summarization with Refined Self-Matching Mechanism
title_sort comprehensive document summarization with refined self-matching mechanism
publisher MDPI AG
series Applied Sciences
issn 2076-3417
publishDate 2020-03-01
description Under the constraint of memory capacity of the neural network and the document length, it is difficult to generate summaries with adequate salient information. In this work, the self-matching mechanism is incorporated into the extractive summarization system at the encoder side, which allows the encoder to optimize the encoding information at the global level and effectively improves the memory capacity of conventional LSTM. Inspired by human coarse-to-fine understanding mode, localness is modeled by Gaussian bias to improve contextualization for each sentence, and merged into the self-matching energy. The refined self-matching mechanism not only establishes global document attention but perceives association with neighboring signals. At the decoder side, the pointer network is utilized to perform a two-hop attention on context and extraction state. Evaluations on the CNN/Daily Mail dataset verify that the proposed model outperforms the strong baseline models and statistical significantly.
topic extractive summarization
deep learning
document summarization
url https://www.mdpi.com/2076-3417/10/5/1864
work_keys_str_mv AT biqingzeng comprehensivedocumentsummarizationwithrefinedselfmatchingmechanism
AT ruyangxu comprehensivedocumentsummarizationwithrefinedselfmatchingmechanism
AT hengyang comprehensivedocumentsummarizationwithrefinedselfmatchingmechanism
AT zibanggan comprehensivedocumentsummarizationwithrefinedselfmatchingmechanism
AT wuzhou comprehensivedocumentsummarizationwithrefinedselfmatchingmechanism
_version_ 1724692284935503872