Question answering in an oral dialogue system

In this paper we describe how a computer dictation system can answer certain questions from a physician. Using speech recognition, dictation systems can automatically generate medical reports. Our dictation system, DictaMed, models the medical entities using an object oriented representation. These entities are created along with the dictation when they are mentioned by the physician. The dictation system is complemented with a questioning module so that the physician can have access to the values of the previously created entities and possibly to supplementary information. The answer to a question is synthesized using a text-to-speech converter.