智能回复系统研究综述

岳世峰; 林政; 王伟平; 孟丹

引用本文：

岳世峰,林政,王伟平,孟丹.智能回复系统研究综述[J].信息安全学报,2020,5(1):20-34 [点击复制]
YUE Shifeng,LIN Zheng,WANG Weiping,MENG Dan.Research on Intelligent Reply System: A Survey[J].Journal of Cyber Security,2020,5(1):20-34 [点击复制]

本文已被：浏览 5066次下载 4196次	码上扫一扫！
智能回复系统研究综述
岳世峰^1,2, 林政¹, 王伟平¹, 孟丹¹
0 字体:加大+\|默认\|缩小-
(1.中国科学院信息工程研究所信息内容安全技术国家工程实验室北京中国 100093);2.中国科学院大学网络空间安全学院北京中国 100049)

摘要:

网络舆情形成快、影响大，如何对其进行智能导控一直是网络安全中的难点和重点。本文提出使用智能回复系统对网络舆情进行自动引导的观点，然后对智能回复系统研究进行了综述。本文首先介绍了当前智能回复系统的主流研究方向，如视觉问答、基于知识图谱的问答和推理问答等不同类型的智能回复系统；接着根据应用场景的不同分别介绍了垂直领域和开放领域的智能回复系统，然后从技术手段上对实现智能回复系统的各种主流方法进行了详细的介绍和探讨。最后本文总结归纳了当前智能回复系统的自动评价方法以及当前智能回复系统存在的主要问题及未来可能的研究方向。

关键词: 网络舆情深度学习问答系统生成模型检索模型

DOI：10.19363/J.cnki.cn10-1380/tn.2020.01.03

投稿时间：2018-03-27修订日期：2018-05-17

基金项目:国家自然科学基金（No.61502478，No.61602467）；国家高技术研究发展计划（No.“863”计划）基金资助项目（No.2013AA013204）。

Research on Intelligent Reply System: A Survey

YUE Shifeng^1,2, LIN Zheng¹, WANG Weiping¹, MENG Dan¹

(1.National Engineering Laboratory for Information Security Technologies, Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China);2.School of Cyber Security, University of Chinese Academy of Sciences, Beijing 100049, China)

Abstract:

Online public opinion develops rapidly and has a great impact on society. It has always been difficult and important to lead it in network security. In this paper, we propose the idea of using intelligent reply system to lead the online public opinion, and then we provide a general overview of the current intelligent reply system. First, we introduce the current mainstream research methods for the intelligent reply system, including the visual question answering, knowledge-based question answering and inference question answering; second, we demonstrate the task-oriented and untask-oriented intelligent reply syst em in terms of different application scenarios. After that we discussed the mainstream methods of building reply system. Finally we summarize the automatic evaluation methods of the current reply system and the main problems as well as future research directions in this field.

Key words: online public opinion deep learning question-answer system generation model retrieval mode