Carl Rebman

Speech recognition in the human-computer interface

Researchers have studied human speech interaction with computers for many years. Much of the focus in this area has been on creating better technical speech recognition (SR) systems, and almost all of the testing has centered on accuracy and productivity gains. However, there has been little study of other issues, such as user acceptance. This paper reports the results of an experiment investigating word generation rates, word error rates, and user acceptance of a speech recognition program as compared to typing. Although the subjects made more errors when using the speech recognition software, they were able to generate more than twice as much text in the same amount of time. However, this relative efficiency was not enough to overcome the inaccuracy and annoyance in fixing so many errors.