计算机应用 ›› 2021, Vol. 41 ›› Issue (10): 2864-2870.DOI: 10.11772/j.issn.1001-9081.2020121945

所属专题: 人工智能

• 人工智能 • 上一篇    下一篇

基于宏观篇章结构的科技论文摘要模型

付颖, 王红玲, 王中卿   

  1. 苏州大学 计算机科学与技术学院, 江苏 苏州 215006
  • 收稿日期:2020-12-11 修回日期:2021-03-04 出版日期:2021-10-10 发布日期:2021-07-16
  • 通讯作者: 王红玲
  • 作者简介:付颖(1994-),女,河南安阳人,硕士研究生,CCF会员,主要研究方向:自然语言处理;王红玲(1975-),女,江苏苏州人,副教授,博士,CCF会员,主要研究方向:自然语言处理;王中卿(1987-),男,江苏苏州人,副教授,博士,CCF会员,主要研究方向:自然语言处理。
  • 基金资助:
    国家自然科学基金面上项目(61976146)。

Scientific paper summarization model using macro discourse structure

FU Ying, WANG Hongling, WANG Zhongqing   

  1. School of Computer Science and Technology, Soochow University, Suzhou Jiangsu 215006, China
  • Received:2020-12-11 Revised:2021-03-04 Online:2021-10-10 Published:2021-07-16
  • Supported by:
    This work is partially supported by the Surface Program of National Natural Science Foundation of China (61976146).

摘要: 针对传统的神经网络模型不能较好地反映科技论文内不同章节之间的宏观篇章结构信息,从而容易导致生成的科技论文摘要结构不完整、内容不连贯的问题,提出了一种基于宏观篇章结构的科技论文摘要模型。首先,搭建了一种基于宏观篇章结构的层级编码器,并利用图卷积神经网络对章节间的宏观篇章结构信息进行编码,从而构建章节层级语义表示;然后,提出了一个信息融合模块,旨在将宏观篇章结构信息和单词层级信息进行有效融合,从而辅助解码器生成摘要;最后,利用注意力机制优化单元对上下文向量进行更新优化操作。实验结果表明,所提出的模型比基准模型分别在ROUGE-1、ROUGE-2以及ROUGE-L上分别高出3.53个百分点、1.15个百分点和4.29个百分点,并且通过对生成的摘要内容进行分析对比,可进一步证明该模型可有效提高生成摘要的质量。

关键词: 神经网络, 宏观篇章结构, 科技论文摘要, 图卷积神经网络, 生成式摘要

Abstract: The traditional neural network model cannot reflect the macro discourse structure information between different sections in scientific paper, which leads to the incomplete structure and incoherent content of the generated scientific paper summarization. In order to solve the problem, a scientific paper summarization model using macro discourse structure was proposed. Firstly, a hierarchical encoder based on macro discourse structure was built. Graph convolution neural network was used to encode the macro discourse structure information between sections, so as to construct the hierarchical semantic representation of sections. Then, an information fusion module was proposed to effectively fuse macro discourse structure information and word-level information, so as to assist the decoder to generate the summarization. Finally, the attention mechanism optimization unit was used to update and optimize the context vector. Experimental results show that the proposed model is 3.53, 1.15 and 4.29 percetage points higher than the baseline model in ROUGE (Recall-Oriented Understudy for Gisting Evaluation)-1, ROUGE-2 and ROUGE-L respectively. Through the analysis and comparison of the generated summarization content, it can be further proved that the proposed model can effectively improve the quality of the generated summarization.

Key words: neural network, macro discourse structure, scientific paper summarization, graph convolution neural network, abstractive summarization

中图分类号: