Reference: Johnson, K.; Poon, A.; Shiffman, S.; Lin, R.; & Fagan, L. M. A History-Taking System that Uses Continuous Speech Recognition. Washington D.C., 1992.
Abstract: Q-MED is an automated history-taking system that uses speaker independent continuous speech as its main interface modality. Q-MED is designed to allow a patient to enter her basic symptoms by engaging in a dialog with the program. Error-recovery mechanisms help to eliminate findings resulting from misrecognitions or incorrect parses. An evaluation of the natural language parser that Q-MED uses to map user utterances to findings showed an overall semantic accuracy of 87 percent; Q-MED asks more specific questions to capture findings that were not volunteered, or that were unable to be parsed in their initial, open-ended form.