ASR Systems Achieve Varied Orthodontic Transcription Accuracy
Researchers in southern England and Zurich published a cross-sectional study in the Journal of Dental Research evaluating transcription accuracy of 10 automatic speech recognition (ASR) systems for narrated orthodontic clinical records. Using Domain Word Error Rate (DWER), they found GPT4o TranscribeCorrected had 3.47% DWER, GPT4o Transcribe and Heidi Health had 7.6% and 6.1% respectively, while Dragon systems reached 29–48% and background noise increased errors.
Key Points
- 1Found GPT4o TranscribeCorrected produced lowest DWER at 3.47%, outperforming other tested ASR systems
- 2Showed significant performance variability, with commercial clinical ASR having DWERs up to 48%
- 3Indicates clinicians must verify and correct ASR transcripts due to mistranscriptions and hallucination risks
Scoring Rationale
Peer-reviewed study with actionable comparative results, but limited scope to orthodontic recordings and selected systems.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
