SilencioNetwork/medical-speech-dataset

Medical Speech Dataset

A specialized speech dataset for healthcare AI applications featuring real medical terminology, clinical conversations, and domain-specific vocabulary.

This dataset is curated from the complete-voiceai-speech-dataset and focuses specifically on medical domain speech data collected from real healthcare contexts.

Dataset Overview

• Total audio files: 33 recordings
• Total duration: ~42 minutes
• Languages: English (native) + Global Medical (multilingual)
• Domain: Medical terminology, clinical documentation, patient-provider conversations
• Audio format: WAV files
• Sample rate: 48 kHz
• License: CC BY-NC 4.0 (free for research, non-commercial use)

Target Applications

This dataset is designed for:

• Medical ASR systems (ambient clinical documentation, medical dictation)
• Healthcare AI assistants (Abridge, Suki, Nabla, Ambience Healthcare)
• Medical voice note transcription
• Clinical conversation analysis
• Medical terminology recognition models
• Healthcare dialogue systems

Dataset Structure

medical-speech-dataset/
├── english_medical/
│   └── medical/
│       ├── data/           # 8 audio files
│       └── metadata.csv    # Speaker metadata
└── global_medical/
    └── medical/
        ├── data/           # 25 audio files
        └── metadata.csv    # Speaker metadata

Data Splits

English Medical (Native Speakers)

• Files: 8 recordings
• Context: Native English speakers discussing medical topics
• Use case: High-accuracy medical ASR training, US/UK clinical documentation

Global Medical (Multilingual)

• Files: 25 recordings
• Context: Medical speech from diverse linguistic backgrounds
• Use case: Accent-robust medical ASR, global telehealth applications

Key Features

✅ Real medical terminology - Conditions, medications, procedures, anatomical terms
✅ Natural speech patterns - Disfluencies, hesitations, clinical conversation flow
✅ Diverse accents - Global medical professionals and patients
✅ Domain-specific vocabulary - Not available in general speech datasets
✅ Ethical data collection - Consent-based, privacy-preserving

Use Cases

1. Ambient Clinical Documentation

Train models to transcribe doctor-patient conversations in real-time (similar to Abridge, Suki, Nabla).

2. Medical Dictation Systems

Improve accuracy for physicians dictating clinical notes, discharge summaries, and prescriptions.

3. Telehealth Transcription

Build ASR systems for virtual healthcare consultations across diverse accents and languages.

4. Medical Voice Assistants

Develop voice-enabled healthcare tools for symptom checking, medication reminders, and patient education.

5. Clinical Research

Analyze speech patterns in medical contexts, study communication dynamics between providers and patients.

Loading the Dataset

from datasets import load_dataset

# Load full dataset
dataset = load_dataset("SilencioNetwork/medical-speech-dataset")

# Load specific split
english_medical = load_dataset("SilencioNetwork/medical-speech-dataset", data_dir="english_medical")
global_medical = load_dataset("SilencioNetwork/medical-speech-dataset", data_dir="global_medical")

Sample Metadata

Each recording includes:

• file_name: Audio file identifier
• birth_place: Speaker's country/region of origin
• language: Primary language spoken
• context: Medical (clinical terminology, healthcare conversations)

Medical Speech Characteristics

This dataset captures real-world medical speech features:

• Medical jargon: "hypertension", "myocardial infarction", "differential diagnosis"
• Clinical abbreviations: Spoken medical shorthand (BP, HR, PRN, etc.)
• Provider-patient dynamics: Turn-taking, clarification requests, empathy markers
• Multilingual medical contexts: Healthcare delivery across linguistic boundaries

Ethical Considerations

All data was collected with explicit informed consent. No protected health information (PHI) is included - all recordings contain general medical terminology only, not patient-specific data.

Need More Medical Speech Data?

This is a sample dataset from Silencio's larger Off-the-Shelf (OTS) medical speech inventory:

📊 Available in full inventory:

• 300+ hours of medical domain speech
• 15+ languages
• Specialized domains: cardiology, radiology, surgery, pharmacy, etc.
• Provider + patient perspectives

Contact us for access: alex@silencioai.com

Citation

If you use this dataset in your research or commercial product, please cite:

@dataset{silencio_medical_speech_2026,
  title={Medical Speech Dataset},
  author={Silencio Network},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/SilencioNetwork/medical-speech-dataset}
}

Related Datasets

• Complete Voice AI Speech Dataset - 39 language/accent variants
• Indian Languages Speech - 9 Indian languages
• European Languages Speech - 5 European languages
• Global English Accents Speech - 20 English accent variants

License

CC BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0 International)

✅ Free for research and non-commercial use
❌ Commercial use requires licensing (contact us)

About Silencio

Silencio is a voice AI data sourcing company with 2M+ contributors across 180+ countries. We provide scaled sourcing of real-world audio and speech data for AI labs, robotics companies, and healthcare AI developers.

🌐 silenciai.com
📧 sofia@silencioai.com

---

Tags: medical speech, healthcare AI, clinical documentation, medical ASR, medical dictation, ambient scribe, domain-specific speech, medical terminology, healthcare NLP, voice health