What question did this study set out to answer?

The study aimed to evaluate the transcription accuracy of various automatic speech recognition systems for orthodontic clinical records.

March 12, 2026Open Access

Transcription Accuracy of Automatic Speech Recognition for Orthodontic Clinical Records

View Full Paper

Ask AI

Bookmark

The study aimed to evaluate the transcription accuracy of various automatic speech recognition systems for orthodontic clinical records.
Analyzed 200 orthodontic diagnostic and treatment planning summaries narrated by specialists.
Utilized 4 commercial ASR systems and 5 application programming interfaces for transcription.
Compared transcription outcomes, including domain word error rates, lexical accuracy, and semantic similarity.
Implemented a two-stage pipeline with a large language model for generative error correction.
Conducted qualitative error analysis and assessed the impact of background noise and narrator accents.
GPT4oTranscribeCorrected showed the highest transcription accuracy (DWER = 3.5%).
Significant variability in accuracy was observed across systems, particularly with technical vocabulary.
Hallucinations occurred in Whisper and DigitalTCO systems.
Background noise negatively impacted transcription accuracy across all systems.
Clinically significant transcription errors ranged from 2% to 66%.

No takes yet. Share an insight, caveat, or question.

Investigates transcription accuracy of speech recognition systems in orthodontic records, revealing significant clinical errors.

O’Kane et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark