BROWSE
Detail
Conference Paper
Design and Collection of Bilingual Spoken Dialog Corpus for Speech-to-speech Translation System
Cited - time in 
Share 
Share
- Issue Date
- 2008-11
- Citation
- International Conference on Speech Databases and Assessments (COCOSDA) 2008, pp.1-4
- Language
- English
- Type
- Conference Paper
- Abstract
- In this paper we report about our designing and developing a bilingual spoken dialog corpus for a traveling domain speech-to-speech translation system (S2ST). The final goal of the project is to develop a translation device which can be used by a Korean traveler traveling foreign countries where English is used as a primary language. To meet the goal, we designed 592 scenarios and tools for speech collection and annotation. Two speakers, one Korean speaker and one English speaker, participate in 20 dialogs and each dialog scenario consists of about 10 utterances. 200 spoken dialogs of pairs of speakers were collected based on two different Wizard-of-Oz (WOZ) methods. In the one method, English and Korean speech synthesizers are used as WOZ, whereas a human interpreter acts as WOZ in the other. The designed scenarios, developed tools and collected results are reported on in this paper.
- KSP Keywords
- Foreign countries, Korean speech, Primary Language, Speech Synthesizer, Speech-to-Speech Translation System, Spoken dialog, Wizard-of-Oz