摘要
相较于早期的生成式摘要方法,基于序列到序列模型的文本摘要方法更接近人工摘要的生成过程,生成摘要的质量也有明显提高,越来越受到学界的关注。本文梳理了近年来基于序列到序列模型的生成式文本摘要的相关研究,根据模型的结构,分别综述了编码、解码、训练等方面的研究工作,并对这些工作进行了比较和讨论,在此基础上总结出该领域未来研究的若干技术路线和发展方向。
Compared with the early abstractive summarization method,the text summarization method based on Sequence to Sequence models is much closer to the process of human-written summaries,and the quality of the generated summary has also been significantly improved,which has attracted increasing attention from the academic community.This paper reviews the research related to abstractive summarization based on Sequence to Sequence models in recent years.According to the structure of the model,this paper summarizes the research on the model in terms of encoding,decoding,training,and so on,and it compares and discusses these works.On this basis,some technical routes and development directions for future research in this field are put forward.
作者
石磊
阮选敏
魏瑞斌
成颖
Shi Lei;Ruan Xuanmin;Wei Ruibin;Cheng Ying(School of Management Science and Engineering,Anhui University of Finance and Economics,Bengbu 233030;School of Information Management,Nanjing University,Nanjing 210023;School of Chinese Language and Literature,Shandong Normal University,Jinan 250014)
出处
《情报学报》
CSSCI
CSCD
北大核心
2019年第10期1102-1116,共15页
Journal of the China Society for Scientific and Technical Information
基金
国家社会科学基金重大项目“中国近现代文学期刊全文数据库建设与研究（1872—1949）”(17ZDA276）
