How to add custom words, medical terms to nvidia riva ASR

AI & Data Science Deep Learning (Training & Inference) Riva
1

I want nvidia riva ASR to correctly transcribe medical terms like histopathology, mandibular, fissue speciment, microscopic diagnosis etc.
Is there any tutorial/resource which shows how to do this ?

Thanks
Prashant

2

Hi @yogpatri ,
Please refer to the sample links below.
https://docs.nvidia.com/deeplearning/riva/user-guide/docs/notebooks/Riva_speech_API_demo.html#1.-Offline-ASR-Example
https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/datasets.html#preparing-custom-asr-data

Thanks!

3

Thanks for the reply.
At present, for training, both audio and corresponding text data are required.
Since medical text is easily available , would it be useful if I use the text to speech service to produce training data for speech to text.

It would be great if anyone who has tried that can share the experience.

4

Hi @yogpatri ,
Can you please try using a LM trained with medical terms and refer to the Riva documentation for instructions on how to deploy with a custom LM.
Here’s an example LM with medical terms.: NVIDIA NGC
Thanks!

5

Hi @AakankshaS , thanks for responding. The above NVIDIA NGC link is not working, seems like it may have changed.
It would be really great if you can share the updated url or exact name of the language model so that I can search it.

6

Hi @yogpatri ,
Apologies for the invalid link.
You should be able to create your own language model with the words you want.
Please refer to Riva — NVIDIA Riva for instructions on how to create a LM with either TAO or KenLM.

Thanks!