Layered Protocols in Voice Interaction with Computers,
Abstract
The Layered Protocol model for human-computer interfaces is described, with special reference to the problems of voice input and output. In a layered protocol, each level passes virtual messages back and forth between human and computer. These virtual messages are realized in the form of interchanges at the level below. The protocol at a level is analogous to the syntax of a sentence, in that it is the method by which the content of a message can be given an agreed interpretation. Each protocol can be designed or evaluated independently of all the others in an interface. The stability of a protocol is determined by its response delays and by the channel capacity of the lower-level protocols that support its messages. Sometimes an unstable protocol can be stabilized and speeded by reducing the message rate of supporting protocols. Users have been observed to do this intuitively. Voice input provides special problems because of the relatively high error probability inherent in the recognizer: errors in other modalities are likely to be due to operator fault. This tends to lead to unwarranted distrust of voice input, and to demands for types of feedback that are probably inappropriate to the level of protocol to which the recognizer is suited. Voice output can be used by the computer to initiate protocols, or to provide a response channel for protocols under conditions where the user's eyes are otherwise occupied.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1987
- Accession Number
- ADP005599
Entities
People
- M. M. Taylor
Organizations
- Defence Research and Development Canada