计算机科学与探索 ›› 2024, Vol. 18 ›› Issue (1): 58-74.DOI: 10.3778/j.issn.1673-9418.2304022

• 前沿·综述 • 上一篇    下一篇

知识驱动的对话生成模型研究综述

许璧麒,马志强,周钰童,贾文超,刘佳,吕凯   

  1. 1. 内蒙古工业大学 数据科学与应用学院,呼和浩特  010080
    2. 内蒙古工业大学 内蒙古自治区基于大数据的软件服务工程技术研究中心,呼和浩特  010080
  • 出版日期:2024-01-01 发布日期:2024-01-01

Survey of Research on Knowledge-Driven Dialogue Generation Models

XU Biqi, MA Zhiqiang, ZHOU Yutong, JIA Wenchao, LIU Jia, LYU Kai   

  1. 1. College of Data Science and Application, Inner Mongolia University of Technology, Hohhot 010080, China
    2. Inner Mongolia Autonomous Region Engineering & Technology Research Centre of Big Data Based Software Service, Inner Mongolia University of Technology, Hohhot 010080, China
  • Online:2024-01-01 Published:2024-01-01

摘要: 知识驱动的对话生成模型旨在利用不同形式的知识来强化对话生成模型,使得对话生成模型不仅能从对话数据中学习语义交互,而且还能深入理解用户输入、背景知识和对话上下文,生成更合理、更具多样性、更富含信息量和拟人的回复,进而推动对话系统的发展。目前相关工作仍处于初期探索阶段,并且很少有对现有成果的全面梳理和系统总结。对知识驱动的对话生成模型研究展开综述,首先,针对现有的研究成果,梳理并介绍了当前知识驱动的对话生成任务和主要遇到的问题,并且给出详细的任务定义和问题定义;其次,整理并介绍了知识驱动的对话生成模型建模所需的数据集;然后,对目前知识驱动的对话生成研究过程中知识获取、知识表示、知识选择和知识融入相关研究中每个模型的改进、研究现状、模型涉及的评价指标和模型的性能进行重点介绍;最后,对知识驱动的对话生成模型研究未来的发展方向进行展望。

关键词: 对话生成模型, 对话系统, 外部知识, 知识驱动的对话生成

Abstract: Knowledge-driven dialogue generation models aim to enhance dialogue generation models by using different forms of knowledge, so that dialogue generation models can not only learn semantic interactions from dialogue data, but also deeply understand user input, background knowledge and dialogue context to generate more reasonable, diverse, informative and anthropomorphic responses, and thus promote the development of dialogue systems. Currently, the related work is still in the early stages of exploration, and there is a lack of comprehensive reviews and systematic summaries of existing results. This paper provides a comprehensive review of the research on knowledge-driven dialogue generation models. Firstly, in response to the existing research results, it sorts out and introduces the current knowledge-driven dialogue generation tasks and the main problems encountered, and provides detailed task definitions and problem definitions. Secondly, it organizes and introduces the datasets required for the modeling of knowledge-driven dialogue generation models. Then, it focuses on the improvement, research status, evaluation indicators involved, and performance of each model in the process of knowledge-driven dialogue generation research, including knowledge acquisition, knowledge representation, knowledge selection, and knowledge integration-related studies. Finally, the future development directions of knowledge-based dialogue generation models are prospected.

Key words: dialogue generation model, dialogue systems, external knowledge, knowledge-driven dialogue generation