Related Experiment Videos
[Edeka does all--machine speech recognition in social medicine expert testimony]
E Michel1, E M Michel, W Hägele
1Medizinischer Dienst der Krankenversicherung Westfalen-Lippe, Münster.
Gesundheitswesen (Bundesverband Der Arzte Des Offentlichen Gesundheitsdienstes (Germany))
|December 9, 1998
Summary
Low-cost speech recognition software shows potential for sociomedicine, but struggles with medical jargon. With specialized training, it can become an effective tool for medical transcription.
Area of Science:
- Medical Informatics
- Natural Language Processing
- Sociomedicine
Context:
- Automatic speech recognition (ASR) is established in limited vocabulary domains.
- Sociomedicine involves complex, specialized language, posing challenges for ASR.
Purpose:
- To evaluate the utility of affordable PC-based speech recognition software in sociomedicine.
- To assess the performance of ASR in transcribing medical malpractice expertise reports.
Summary:
- Dictating 11,000 words of medical text resulted in a 15.9% error rate for unknown text.
- Re-dictating corrected text reduced errors to under 3%, demonstrating significant learning.
- High error rates in novel medical text are attributed to specialized jargon, not system limitations.
Impact:
- ASR systems require extended training or specialization for effective use in medical fields.
- This technology offers a promising, cost-effective solution for medical transcription in sociomedicine.
- Optimized ASR could enhance efficiency in processing medical-legal documentation.