A University of Bern network project integrating medicine, linguistics, law, patient safety, and data science to evaluate and build responsible AI documentation for real-world clinical practice.
Clinical documentation is essential for patient safety but increasingly burdens clinicians. This project evaluates AI-based documentation under real conditions such as dialectal variability, environmental noise and fragmented encounters involving multiple providers. Authentic Swiss data serve as a stress test for medicine at large. We combine legal, linguistic, clinical, and computational perspectives to study accuracy, safety, and compliance, and to deliver practical tools for deployment.
Healthcare is adopting AI rapidly, yet real-world evaluation in multilingual, high-pressure settings is scarce. We bridge that gap.
A truly interdisciplinary approach and open, reusable outputs designed for Swiss realities and transferrable beyond.
Modular tools for audio normalization, dialect-aware ASR post-processing, terminology correction, and LLM summarization.
De-identified evaluation sets and a Swiss-German medical term lexicon with common ASR misrecognitions.
Error-risk mapping, uncertainty visualizations, and checklists supporting safe human-in-the-loop use.
Consent templates and policy briefs on documentation, evidence, and accountability for AI-generated records.
Sociolinguistics & Dialects
Leads linguistic analysis and develops Swiss German medical lexicons and benchmarks across dialects.
Emergency Medicine & Digital Medicine
Leads clinical evaluation and safety benchmarking. Endowed professor of emergency telemedicine and digital triage.
Health Law & Data Protection
Leads legal analysis on consent, evidence and accountability and develops consent templates and policy briefs.
Patient Safety & Quality
Leads the safety framework, risk mapping and oversight toolkits for safe human-in-the-loop adoption.
Natural Language Processing, Data Science Lab
Leads the development, evaluation and adaptation of research-oriented AI and NLP pipelines for clinical speech data.
Emergency Medicine & Diagnostic Quality
Brings experience in interventional trials, diagnostic quality research and digital decision support.
Ambient documentation in multilingual clinical settings: early findings and challenges.
Publication PDFFor inquiries about collaboration, data access, or media, please reach out: