Takuya Nishimoto, Yutaka Kobayashi and Yasuhisa Niimi
In current spoken dialog systems, the design of user interfaces strongly depends on tasks and domains. We intend to develop a general purpose spoken dialog system. It extends existing Graphical User Interface (GUI) design used on-line database services on the World Wide Web (WWW), so that the design of user interface does not require special knowledge about spoken dialogs. Web documents are written in Hyper Text Markup Language (HTML) and Common Gateway Interface (CGI), which naturally guide some dialog interactions as intended by the authors and by the browsers. In other words, each document has its own dialog control strategy in itself. Making use of such strategies enables task-oriented spoken dialogs, provided a single set of task-independent client modules. The dialog management information is extracted from mark-up tags, accordingly free from natural languages of text. Therefore, one could even provide a cross-language feature to the system by combining machine translation systems.