Japanese radiological report creation with continuous speech recognition

Ten Japanese radiological reports consisting of 1381 characters (681 words) were created by two board-certified radiologists who used conventional typing and a continuous speech-recognition system called AmiVoice (Advanced Media, Inc., Tokyo, Japan). The two radiologists had not had any special trai...

Full description

Saved in:
Bibliographic Details
Published inNihon Igaku Hoshasen Gakkai zasshi. Nippon acta radiologica Vol. 62; no. 1; p. 23
Main Authors Takahara, Taro, Nakajima, Mika, Nitatori, Toshiaki, Hachiya, Junichi
Format Journal Article
LanguageJapanese
Published Japan 01.01.2002
Subjects
Humans
Radiology Information Systems - standards
Software
Speech Acoustics
Voice
Online AccessGet more information

Cover

More Information
Summary:Ten Japanese radiological reports consisting of 1381 characters (681 words) were created by two board-certified radiologists who used conventional typing and a continuous speech-recognition system called AmiVoice (Advanced Media, Inc., Tokyo, Japan). The two radiologists had not had any special training prior to their use of the continuous speech-recognition system. The model of speech-to-text analysis was generated from 22,589 radiological reports (5.7 MB). Dedicated pronunciations for loan words (i.e., English words) were registered by a board-certified radiologist in consideration of variations in Japanese pronunciation. Misrecognition occurred in 40 of 1362 words, corresponding to a 97.1% rate of accuracy of recognition. The average speech recognition time per report was 31.3 sec, and the additional time required for corrections was 25.0 sec. The total speech input time of 56.2 sec was much less than the conventional input time of 142.8 sec for typing. Continuous speech recognition is faster than typing, even considering the additional time required for corrections, and is acceptable in view of the overall reduction in report turn-around time.
ISSN:0048-0428