Table of Contents
In pharmaceutical workflows, getting a medication name wrong isn’t just another transcription error. It can introduce incorrect information into critical workflows, from prescription refill automation and pharmacy IVRs to clinical documentation. A transcript can score highly on overall accuracy, but an error on a medication name can carry far greater consequences than errors elsewhere in the transcript.
That’s why we built Nova-3 Pharma, the first speech-to-text model purpose-built for the pharmaceutical industry.
Nova-3 Pharma is specifically designed to accurately transcribe drug names, pharmaceutical terminology, and medical vocabulary while maintaining the same high-quality transcription performance Deepgram customers expect from Nova-3. In our latest competitive benchmarks, Nova-3 Pharma ranks #1 in drug name recognition and #1 in overall transcription accuracy across both batch and streaming.
With Nova-3 Pharma, developers can build pharmacy IVRs, prescription pickup and refill automation, member services and contact center voice agents, and clinical documentation applications where drug name accuracy is critical.
It’s available now for both batch and streaming transcription through Deepgram’s hosted API and self-hosted deployments.
High accuracy for pharmaceutical terminology
General-purpose speech-to-text models are designed to perform well across a broad range of speech. Medical models add specialized clinical vocabulary. But pharmaceutical speech introduces another challenge: thousands of medication names that can be uncommon, phonetically similar, and difficult for general-purpose and medical STT systems to recognize reliably.
For pharmaceutical applications, aggregate accuracy alone doesn’t tell the full story. Critical terminology like medication names requires a higher level of precision.For example, if a patient says a medication name like “Benlysta” or “Dofetilide” and the model transcribes a different medication, a strong overall accuracy score can obscure a critical error.
Nova-3 Pharma is specifically designed for this level of specialized accuracy. The model combines Deepgram’s medical speech foundation with pharma-specific training to improve recognition of drug names and pharmaceutical terminology while maintaining strong general transcription performance.
The result is more accurate transcription where pharmaceutical terminology matters most.
Nova-3 Pharma is #1 in drug name recognition
Overall transcription accuracy and drug name recognition measure two different dimensions of performance. Word Error Rate (WER) measures accuracy across the full transcript, while Keyword Recognition Rate (KRR) measures whether specific terms spoken in the audio are correctly captured. For pharmaceutical speech, we evaluate both, using KRR to measure drug name recognition directly.
Nova-3 Pharma is trained on more than 5,000 drug names to improve recognition of pharmaceutical terminology. We evaluated the model against leading speech-to-text and medical speech-to-text models using real English medical audio across therapy, telehealth, clinical dictation, triage, and other medical use cases.
Nova-3 Pharma ranked #1 in drug name KRR across every model evaluated in both batch and streaming:
- 91.59% drug name KRR in batch
- 91.28% drug name KRR in streaming
Nova-3 Pharma is #1 in overall transcription accuracy
Nova-3 Pharma pairs industry-leading drug name recognition with industry-leading overall transcription accuracy. On the same benchmark, it delivered the lowest Word Error Rate of every model evaluated in both batch and streaming:
- 10.17% WER in batch
- 11.69% WER in streaming
Together, the results show that Nova-3 Pharma leads on both drug name recognition and overall transcription accuracy.
Power pharmaceutical applications with Nova-3 Pharma
Nova-3 Pharma gives developers a specialized speech-to-text foundation for voice applications across pharmaceutical workflows.
Build pharmacy IVRs and prescription refill automation that accurately capture medication names. Power payer and member-services voice agents handling medication-related requests. Transcribe clinical documentation where drug name accuracy is critical.
Deploy Nova-3 Pharma now
Nova-3 Pharma uses the same architecture and compute footprint as Nova-3 Medical and is available through Deepgram’s hosted API and self-hosted deployment.
It supports both streaming and batch transcription in eight English variants: en, en-US , en-GB, en-AU, en-CA,en-IE, en-IN, en-NZ.
Get started with: model=nova-3-pharma
