Dialog Studio: An Example Based Spoken Dialog System Development Workbench

In recent years, spoken dialog systems have been exploited by many researchers, but it is still too early to apply it for practical use in commercial fields. In addition to the inaccuracy of each component of dialog systems including automatic speech recognizer (ASR), spoken language understanding (SLU), and dialog management (DM), the inconvenience and the high cost of maintaining and upgrading the spoken dialog systems are one of the biggest problems that prevents the use of spoken dialog systems in commercial applications. This paper introduces a dialog workbench tool, Dialog Studio, which is a back-end workbench system for example-based spoken dialog systems. It is designed for reducing human efforts of tutoring the dialog systems with new dialog examples, supporting semiautomatic learning by recommending a dialog administrator with proper candidate answers and synchronizing all dialog system’s components for upgrading spoken dialog systems.