ERRORS AND ERROR-CORRECTION IN AUTOMATIC SPEECH RECOGNITION SYSTEMS

Errors, whether created by the user, the recognizer, or inadequate sys tems design, are an important consideration in the more widespread and successful use of automatic speech recognition (ASR). An experiment i s described in which recognition errors are studied under different ty pes of feedback. Subjects entered data verbally to a microcomputer acc ording to four experimental conditions: namely, orthogonal combination s of spoken and visual feedback presented concurrently or terminally a fter six items. Although no significant differences in terms of error rates or speed of data entry were shown across the conditions, analysi s of the time penalty for error correction indicated that as a general rule, there is a small timing advantage for terminal feedback, when t he error rate is low. It was found that subjects do not monitor visual feedback with the same degree of accuracy as spoken, as a larger numb er of incorrect data entry strings was being confirmed as correct. Fur ther evidence for the use of 'second best' recognition data is given, since correct recognition on re-entry could be increased from 83.0% to 92.4% when the first choice recognition was deleted from the second a ttempt. Finally, the implications for error correction protocols in sy stem design are discussed.