Issues in the design of a voice man machine dialogue system generating written medical reports

In this paper, we describe the syntactic and semantic issues of a voice man machine dialogue system. This system, featuring a voice recognition circuit, allows the dictation of medical analysis reports and automatically generates a written document. An object oriented kernel is designed to represent the linguistic and medical entities. The syntactic parser is adapted to voice input and enables the physician to dictate using natural language. A semantics analyzer manages the medical entities and the agenda of the report. This analyzer allows to detect certain errors or inconsistencies, and to warn the user, using synthesized audio messages. The audio interface has been merged into a graphical environment to allow keyboard and pointing device controls, as well as graphical alerts.