Using Natural Dialogs as the Basis for Speech Interface Design

Abstract

The main premise of this chapter is that speech applications will be more usable if they follow the conventions of natural human dialog. Understanding how natural dialogs take place in the application domain serves as an effective starting point for a speech user interface design. This chapter reports on natural dialog studies that take place in the earliest stages of an application life cycle, prior to any system design or functional specification. Not only do these studies help in the design of grammars, feedback, and prompts, but they also point out instances where speech technology cannot be effectively applied. Natural dialog studies help to uncover common patterns of behavior and can reveal unexpected insights that significantly improve application design. Occasionally, observing participants in a natural setting can even trigger the rethinking of basic design assumptions. The natural dialog studies described in this chapter were conducted in the process of designing four experimental applications built by the Speech Applications Group at Sun Microsystems Laboratories. Each of these applications attempted to push the bounds of conversational interaction, given the constraints of current speech technology. Of the four case studies described, two are for speech-only systems and two involve multimodal systems: (1) The SpeechActs Calendar, a telephone-based system to access an online calendar, (2) The Office Monitor, a walk-up speech interface which offers information and takes messages when an office occupant is out, (3) The Automated Customer Service Representative, a multimodal version of the Lands’ End catalog that allows users to query with speech and see the results on an interactive television display, and (4) Multimodal Drawing, a speech-enhanced user interface to a drawing editor. The study descriptions conclude with a summary of lessons learned about general speech user interface concepts.