Continuous Speech Recognition in Radiology Reporting.

Background. The early 1990s saw the development of the first viable voice recognition software.' The software was fraught with problems requiring pauses between words and poor word recognition. These characteristics were unacceptable to most individuals and proved impractical in applications other than minimal note writing of frequently repeated phrases. Recent advances in the area of continuous speech recognition have resulted in software with the capability to be viable for everyday use. An implication for the use of such software for the generation of medical records is readily apparent. Speech recognition systems offer a popular alternate computer input method for two principal reasons: speech is a natural form of communication, and speech recognition systems can recognize speech at a rate faster than many people can type.2