Deepgram nova-3-medical model mis-transcribes medical terminology & drug names — how to improve accuracy?

[Harsh772005]

Hi everyone,

We are evaluating Deepgram’s nova-3-medical ASR model, but we are not getting the required accuracy for medical terminology. Many drug names and clinical terms are mis-transcribed, which is not acceptable for our use case.

Issue Description

Even with recommended settings, the transcription output fails on multiple important medical terms.

We tested using this publicly available dataset:
🔗 https://observer.med.upenn.edu/dataset/explore

Below is a comparison between the reference transcript and our Deepgram output:
(Example: anonymized_deepgram_enhanced_diff_1_1.html)
anonymized_deepgram_enhanced_diff 1 1.html

Several drug names and medical terms are incorrectly transcribed.

Current Request Configuration

request = audio_data
model = "nova-3-medical"
language = "en-US"
smart_format = True
diarize = True
utterances = True

We also tried alternate formatting and diarization settings, but accuracy did not improve significantly.

What We Tried

• Smart formatting on/off

• Diarization on/off

• Verified audio sampling rate

• Tested premium/enhanced settings

• Multiple files from the dataset

Still no major improvement in medical-term accuracy.

Looking for Suggestions

Does anyone have advice or experience improving transcription accuracy for medical terminology/drug names using Deepgram?

Specifically, looking for:

• boosting options

• Any additional nova-3-medical parameters

• Preprocessing tips for clinical audio

• Alternative Deepgram models for medical ASR

• Known limitations of nova-3-medical

Any recommendations or insights would be greatly appreciated!

[deepgram-community[bot]]

Thanks for asking your question. Please be sure to reply with as much detail as possible so the community can assist you efficiently.
_{Consider joining our Discord community for more opportunity to engage with your fellow Deepgram users. You can earn points which can be redeemed for cool stuff by being active in our communities!}

[deepgram-community[bot]]

Hey there! It looks like you haven't connected your GitHub account to your Deepgram account. You can do this at https://community.deepgram.com - being verified through this process will allow our team to help you in a much more streamlined fashion.

[deepgram-community[bot]]

It looks like we're missing some important information to help debug your issue. Would you mind providing us with the following details in a reply?

• The deepgram product you are using (e.g Speech to Text, Agent API)
• A request ID that triggered your error or issue.

[Harsh772005]

It looks like we're missing some important information to help debug your issue. Would you mind providing us with the following details in a reply?

• The deepgram product you are using (e.g Speech to Text, Agent API)
• A request ID that triggered your error or issue.

1️⃣ Which Deepgram product you are using
2) Request ID : f2030da0-39a8-49c8-bbb9-57f525ccf7cf
We are using Deepgram Speech-to-Text (nova-3-medical model) via the API.

[Rajan-Gupta1]

Any update from DeepGram on this thread. We are experiencing similar issues with DeepGram with medical grade transcriptions. e.g. Blood Pressure spoken as onehundred over eighty is being returned as 100 over 80. This far from what a medical transcription should be doing. I am sure we are doing something wrong. Kindly advise.