Medical Transcription, also referred to as “Healthcare Documentation” is one of the fastest growing medical industries. Transcription involves the transfer of audios recorded by doctors and skilled nursing staff, pertaining to the patient’s visit. While medical transcription is often considered a mundane task, nevertheless, the physician’s observations, orders, and conversations with patients must be accurately transcribed.
Medical Transcription: Google’s Newest Technology
Google tested whether the voice recognition technologies already available in Google Assistant, Google Home, and Google Translate could be used to fully automate the transcription process. Automating medical transcription can help doctors and transcribers take notes more quickly.
In a recent experiment, Google developed and used two automatic speech recognition models:
- Connectionist Temporal Classification (CTC) phoneme-based model,
- Listen Attend and Spell (LAS) grapheme-based model
Google researchers entered over 14,000 hours of recorded speech into both systems.
The result was a pretty respectable word error rate of 20.1% for the CTC model and 18.9% for the LAS model. However, the CTC model required the researchers to clean up noise in the recordings before processing it.
Google, will now work with Stanford University doctors to fine tune clinically relevant information for automated medical transcription.
Automating medical transcription will reduce processing time, increase transcription accuracy, and increase the physician’s productive time with patients.