级联对话状态跟踪模型在口语对话中是否抢话？
计算与语言
2023-11-21 v1 人工智能
音频与语音处理
信号处理
摘要
在面向任务的对话（TOD）系统中，正确更新系统对用户需求的理解是顺畅交互的关键。传统上 TOD 系统由若干相互交互的模块组成。尽管这些组件各自都是活跃研究团体的焦点，但它们在交互中的行为可能被忽视。本文对最先进系统在复杂设定（如高度依赖对话上下文的对话状态跟踪）中的错误进行了全面分析。基于口语 MultiWoz，我们确定非类别槽位值的错误是弥合口语与基于聊天的对话系统之间差距所必须解决的关键。我们探索了潜在的解决方案以改进转写并帮助对话状态跟踪生成模型纠正此类错误。
引用
@article{arxiv.2311.04922,
title = {Are cascade dialogue state tracking models speaking out of turn in spoken dialogues?},
author = {Lucas Druart and Léo Jacqmin and Benoît Favre and Lina Maria Rojas-Barahona and Valentin Vielzeuf},
journal= {arXiv preprint arXiv:2311.04922},
year = {2023}
}
备注
Submitted to IEEE ICASSP 2024{\copyright} 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works