The model is based on BLOOMZ-3B a large language model trained by the BigScience team.
The full paper is available here.
The model is a created using peft, transformers, and bitsandbytes.

Using a transcript of a patient visit with his/her general health care provider (GP) recommend where that patient should be referred to and extract the main keywords from the transcript.
For example, if a patient has allergy-related concerns, then GP needs to refer that patient to the allergy/immunology department or an immunology doctor available.

Extracted keywords are there to help the doctor in making the final decision of what to recommend to the patient.