吉林大学学报(信息科学版) ›› 2019, Vol. 37 ›› Issue (6): 623-630.

• • 上一篇    下一篇

口语对话系统中对话管理的演变

霍晓凯1a,1b,史冬辉2,石一锐1a,1b,周柚1a,1b,时小虎1a,1b,吴春国1a,1b   

  1. 1. 吉林大学a. 符号计算与知识工程教育部重点实验室; b. 计算机科学与技术学院,长春130012;2. 大连市公安局交通警察支队,辽宁大连116022
  • 出版日期:2019-11-24 发布日期:2020-01-03
  • 通讯作者: 吴春国( 1976— ) ,男,黑龙江鹤岗人,吉林大学副教授,主要从事机器学习和进化计算研究,( Tel) 86-431-85168510( E-mail) wucg@ jlu. edu. cn。 E-mail: wucg@ jlu. edu. cn
  • 作者简介:霍晓凯( 1991— ) ,男,山西大同人,吉林大学硕士研究生,主要从事自然语言理解研究,( Tel) 86-431-85168510( E-mail)tmq2010@126. com; 通讯作者: 吴春国( 1976— ) ,男,黑龙江鹤岗人,吉林大学副教授,主要从事机器学习和进化计算研究,( Tel) 86-431-85168510( E-mail) wucg@ jlu. edu. cn。
  • 基金资助:
    国家自然科学基金资助项目( 61772227) ; 吉林省科技厅基金资助项目( 20180201045GX; 20160101259JC)

Evolution of Dialogue Management in Spoken Dialogue System

HUO Xiaokai1a,1b,SHI Donghui2,SHI Yirui1a,1b,ZHOU You1a,1 b,SHI Xiaohu1a,1b,WU Chunguo1a,1b   

  1. 1a. Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education; 1b. College of Computer Science and Technology,Jilin University,Changchun 130012,China; 2. Traffic Police Detachment,Dalian Public Security Bureau,Dalian 116022,China
  • Online:2019-11-24 Published:2020-01-03

摘要: 为了对人机语言交互领域的主要工作进行归纳和总结,分析了口语对话系统的各个组成模块,包括主要研究内容和关键技术; 然后,针对其核心的对话管理技术进行了综述,通过详细分析并比较几类主要对话管理实现方法,力求在分析现有算法的同时勾勒出其技术的发展与演变过程; 最后,对对话管理领域的深入研究进行了分析与展望。

关键词: 口语对话系统, 对话管理, 自然语言处理, 深度学习, 部分可观测, 马尔可夫决策过程

Abstract: The spoken dialogue system has great practical value and broad commercial prospects,and has significant research significance for its implementation which involves many directions in the field of natural language processing. Dialogue management is the core technology of spoken dialogue system,and plays an important role. The evolution of dialogue management forms the main historical line of the development of spoken dialogue system. In recent years,the research in the field of spoken dialogue system,especially the research on dialogue management,has yielded fruitful results. In order to generalize and summarize the main work in this field,this article first introduces the various components of the spoken dialogue system,including the main research contents and key technologies. Then focuses on the core module of the system. The development and evolution of the technology are outlined. Finally,we have studied the further developments of dialogue management methods,and have forecasted the future develepment.

Key words: spoken dialog system, dialog manage, nature language process, deep learning, partially observable, markov decision process

中图分类号: 

  • TP39