Deepgram works with a wide range of healthcare clients and platforms, including Lyrebird, Tortus, Middletown Medical, Twilio, Five9, and Phonely, to create advanced voice AI automations and experiences designed for safe, secure, and scalable deployment. We’re deepening our support for healthcare with improved medical model accuracy that addresses many of the common pain points that both large businesses and developers face when automating EHR and SOAP notes, transcribing patient sessions, handling patient intakes, and more.
Since the launch of Deepgram Nova-2, our industry-leading speech-to-text (STT) model, we’ve been hard at work enhancing its capabilities, including expanded language support as well as the release of a number of fit-for-purpose use case models specifically trained for different speech domains like meetings, phone calls, finance, and content/video streaming. These options deliver improved accuracy and are optimized to handle domain-specific vocabulary, jargon, and abbreviations as well as different environmental audio conditions and flavors of background noise.
Today we are excited to unveil our latest innovation in speech recognition technology specifically designed for the healthcare sector—the Nova-2 medical model. Our Nova-2 medical model is proficient in transcribing healthcare terminology, including symptoms, diagnoses, treatments, medications, and clinical jargon. This advanced model is not just an upgrade; it's a giant leap forward, boosting keyword accuracy and enhancing communication in medical domains in several notable ways:
- Enhanced Recognition of Medical Terms: Experience a remarkable 16% relative improvement in word recall rates (WRR) for medical terminology compared to the previous model, with an average relative WRR improvement of 20.5% vs. leading competitors. This enhancement ensures that important details are captured accurately, supporting healthcare professionals in delivering top-notch care.
- Superior Overall Accuracy: For pre-recorded (batch) transcription, the new medical model has an 11% relative improvement in overall word error rate (WER), with an average relative improvement in WER of 42.8% vs. benchmarked alternatives. With the Nova-2 medical model, you won’t sacrifice general speech recognition performance for medical terminology accuracy.
- Faster turnarounds for faster operations: Nova-2’s groundbreaking architecture yields a significant speed advantage compared to alternative speech recognition solutions, resulting in transcription speeds that are 5 to 40 times faster than comparable vendors and one of the only options available on the market for real-time application performance.
- Cost Efficiency: The Nova-2 medical model was meticulously engineered to help you save money. Our state-of-the-art, next-generation model sets new standards in efficiency, delivering significant cost savings for your operations.
“We have been using the Nova-2-medical model for the last few months, and have had a very positive experience. Having medical speech recognition with good accuracy makes a significant difference in the notes that doctors create using our AI Scribe. We are definitely impressed with its ability to recognize medical terms and the blazing processing speed of our audio files.”
–Gerardo Guerra Bonilla, CEO Chartnote
Figure 1: Median file word error rate (WER) for pre-recorded English transcription across all benchmarked medical domain test sets.
Why Choose the Deepgram Nova-2 Medical Model
Our medical model customers are transforming the healthcare industry by advancing medical documentation and patient interaction with voice AI, giving the healthcare providers they serve a number of key benefits:
- Higher Precision: Minimize errors in patient records and improve the accuracy of clinical documentation.
- Efficient Workflow: Spend less time on paperwork and more time with patients, thanks to our model's swift and accurate recognition capabilities.
- Compliance and Confidentiality: Adhere to stringent medical documentation standards and ensure HIPAA compliance by ensuring patient information remains confidential and secure.
- Flexible Deployment Options: Available as a managed service, or securely self-host on your own VPC or on-premises.
- Custom Model Training: Improve accuracy of uncommon keywords (e.g. new drug names) with rapid custom model training services that boost the Nova-2 medical model’s already impressive, out-of-the-box performance.
“Deepgram's speech-to-text has been far and away the best we've seen for medical transcription accuracy. Plus their approach makes it easy to add in new, uncommon words outside of their training domain.”
–Will Bodewes, CEO Phonely AI
Getting Started
We hope you’ll visit our API Playground or sign up to try our Nova-2 medical model for yourself.
To access the model, simply use model=nova-2-medical in your API calls. To learn more, please visit our API Documentation.
Join us in leading the healthcare revolution with superior speech recognition technology that understands your needs. For more information and to see how the Nova-2 medical model can integrate into your system, contact us today!
If you have any feedback about this post, or anything else around Deepgram, we'd love to hear from you. Please let us know in our GitHub discussions or contact us to talk to one of our product experts for more information today.