The design of voice-driven interfaces

The design of voice-driven interfaces Alexander I. Rudnicky School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 Abstract This paper presents some issues that arise in building voice- driven interfaces to complex applications and describes some of the approaches that we have developed for this purpose. To test these approaches, we have implemented a voice spreadsheet and have begun observation of users interacting with it. Introduction Interaction with a complex system imposes a substantial cognitive load on a human user. Voice, being a natural communication modality for humans, allows the user to spend less time on translating requests into keyboard actions and more time on problem solving activities. The benefits of voice input will very likely be realized through such ob- servable consequences as shorter completion times, fewer errors in finished work, and increased user satisfaction. The design of interactive systems that use continuous speech for complex problem-solving activity, however, is not well un- derstood. This purpose of this paper is to describe an ap- proach to the problem of carrying out such designs. To study the variables that affect the behavior of a user performing a task using voice, we must construct a suitable environment. The choice of task is critical: simulations, for example, while the allow the experimenter to control all aspects of the environment might not produce findings that are easily Ixanslatable to other applications. On the other hand, some' 'real" tasks may not necessarily produce be- havior that is "interesting" from a spoken language system perspective. Simple data entry tasks might fall into this category. We believe that the spreadsheet environment has the desirable characteristics we are seeking in that it will permit us to study a wide range of behavior. We will describe it in greater detail below. Before doing so, we would like to review the issues that must be addressed in the course of building an interface. Speech Interface Design Designing a functioning spoken language system requires attention to a number of factors. This section reviews these and describes some of the approaches we have taken. Language design Language design refers to the creation of a spoken lan- guage suitable for human interaction with an application. The goal is to create a habitable language that captures the range of expression to be expected from the typical user and maps it into the capabiifies of the application. From the perspective of the system we have been exploring, we can make the further distinction between core language and task language. Core language is a component that is not ex- pected to change from one circumstance to another and en- compasses certain common modes of expression. A good example of such a core is the numbers. There are many ways to express numeric information and in fact capturing these leads to the construction of a fairly complex grammar. However, once created, it is clear that the number grammar fits not only into, say, the spreadsheet task but can be em- bedded in any number of …

Content truncated to 500 words