Acoustical Science and Technology | |
A multilingual-supporting dialog system across multiple domains | |
Yasuhisa Niimi2  Masahiro Araki2  Yunbiao Xu1  | |
[1] College of Computer Science and Information Engineering, Hangzhou University of Commerce;Department of Electronics & Information Science, Kyoto Institute of Technology | |
关键词: Multilingual; Dialog system; Multiple domains; Dialog controller; Case frame; Topic frame; | |
DOI : 10.1250/ast.24.349 | |
学科分类:声学和超声波 | |
来源: Acoustical Society of Japan | |
【 摘 要 】
References(13)Cited-By(2)This paper presents a multilingual-supporting dialog system that was implemented in Chinese and Japanese in tasks of sightseeing, accommodation-seeking and PC assembling guidance. Such a dialog system benefits from three main methods we proposed, including case frame conversion, template-based text generation and topic frame driven dialog control scheme. The former two methods are for improving the portability across languages, and the last one is for improving the portability across domains. The case frame conversion is used for translating a case frame described in a particular language into that described in a pivot language. The template-based text generation is used for generating text responses in a particular language from abstract responses described in the pivot language. The topic frame driven dialog control scheme makes it possible to manage mixed-initiative dialog based on a set of task-dependent topic frames. Both Chinese and Japanese experiments in several domains showed that the three methods proposed could be used to improve the portability of the dialog system across domains and languages.
【 授权许可】
Unknown
【 预 览 】
Files | Size | Format | View |
---|---|---|---|
RO201912080715376ZK.pdf | 387KB | download |