Custom Language Models (CLM) with AWS HealthScribe

I apologize, but there seems to be a misunderstanding in your question. AWS HealthScribe does not currently support the use of Custom Language Models (CLMs). CLMs are a feature of Amazon Transcribe, the general-purpose transcription service, but they are not directly integrated with AWS HealthScribe.

AWS HealthScribe is a specialized service that combines speech recognition and generative AI technologies specifically optimized for medical conversations. It is designed to automatically generate preliminary clinical notes and consultation transcripts from patient-clinician conversations in healthcare settings. The service is currently optimized for US English in general medicine and orthopedics practices.

While AWS HealthScribe does not support CLMs, it uses custom-trained speech recognition technology tailored for medical conversations. This means it's already optimized for healthcare terminology and context without the need for additional custom language models.

As for creating a pipeline between Amazon Transcribe (with CLMs) and AWS HealthScribe, this is not currently possible. AWS HealthScribe is an end-to-end solution that handles both the transcription and the generation of clinical notes. It's not designed to accept external transcripts as input.

If you need highly specialized vocabulary or language patterns that aren't covered by AWS HealthScribe's built-in capabilities, you may want to consider using Amazon Transcribe Medical with Custom Vocabularies instead. However, this would not provide the clinical note generation features of AWS HealthScribe.

For the most accurate and up-to-date information on AWS HealthScribe's capabilities and potential future features, I recommend checking the official AWS documentation or contacting AWS support directly.
Sources
AWS Healthscribe - AWS AI Service Cards
AWS HealthScribe - Amazon Transcribe