Ambient Clinical Scribes
Training voice AI to document medicine as it happens
Industry Challenge
Physician documentation is one of the greatest drivers of clinical burnout. Ambient AI scribes — which listen to doctor–patient conversations and generate structured clinical notes automatically — have the potential to give clinicians hours back every week. But training these models requires massive volumes of real, multilingual clinical conversation audio paired with expert transcriptions and structured clinical note annotations.
How SCILabel Serves This Industry
Data Collection
We source doctor–patient consultation audio from clinical research partnerships, teaching hospitals, and consent-based community recording programmes. Recordings span general practice, internal medicine, paediatrics, obstetrics, and specialist consultations. Multilingual coverage includes English, Swahili, French, and other regional languages. All audio is collected under explicit informed consent with full IRB-equivalent review.
Data Annotation & Labeling
Our clinical transcriptionists — nurses, clinical officers, and health records specialists — produce verbatim and clean-read transcriptions with speaker diarisation (doctor vs. patient) and clinical entity tagging (symptoms, diagnoses, medications, procedures). We also annotate structured clinical note fields: History of Presenting Complaint, Examination Findings, Assessment, and Plan (SOAP format).
Data & Model Evaluation
Clinical evaluators assess transcription word error rate (WER), entity recognition accuracy against gold-standard annotations, and clinical note completeness and accuracy benchmarked against physician-authored notes.
Annotation Types & Formats
- Verbatim and clean-read transcription with speaker diarisation
- Named entity recognition: symptoms, diagnoses, medications, allergies, procedures
- SOAP note structure annotation from consultation transcripts
- ICD-10 and SNOMED CT coding of clinical entities in transcripts
- Dialogue act labeling (question, instruction, reassurance, history-taking)
Specialist Workforce Tracks
Track 1 (Medical NLP) and Track 3 (RLHF): Clinical Officers, Nurses, Medical Records Officers, General Practitioners.