Rapid message composition for large vocabulary speech output aids: A review of the possibilities

Abstract

There are now a variety of synthetic-speech aids designed to alleviate the communication difficulties of people unable to speak. To be effective, such devices should have a large if not unlimited repertoire of output messages. Perhaps the major remaining unsolved problem in this area (even more than achieving good speech quality and intelligibility) is that of composing the required message in anything like “real time”, that is, so as to give conversational ability. This paper reviews the possibilities for rapid message composition in large-vocabulary speech synthesis systems, with attention being restricted to keyboard input. To have any chance at all of achieving the required speed, it is necessary to limit ourselves to dextrous people for whom the inability to speak is effectively their only handicap. Although a less than ideal scheme, the possibilities for rapid input can be classified according to the nature of the message elements selected and the type of input device used. Size of the input device is argued to be an extremely important consideration as it has implications for portability of the aid and for selection times. Approaches based on message-element selection using sequential keyboards, touch-sensitive input devices and parallel (chord) keyboards are reviewed. Message elements considered are single letters, iconic representations of “concept,” and syllables. Word selection using abbreviated “shortforms” is also discussed. Although prospects for truly verbatim message composition appear limited, there does seem to be scope for useful research in the future. Perhaps the most fruitful avenue would be to explore the potential of a chord keyboard employing orthographic syllabic coding and used in conjunction with a sizeable shortform dictionary.