This was our teams entry for the firsts every HackSLU hackathon. The theme was healthcare and we specialized in the track of telehealth and mental health.
Oftentimes it can take weeks or months to get into a therapists office, and much of the first visit is centered around the psychiatrist asking a series of standardized questions to rank the severity of the patient's condition, if there is any. Due to the sheer number of people utilizing therapy, this initial visit contributes to a large backlog of patients. Our project was designed to help midigate this backlog by determining the severity of the patients conditions before the appointment, and to provide resources on where to seek help, should they need it.
Important
Context
For most of our team, this was their first hackathon and first real experience programming outside of class. Because of this, we descided to keep our scope small. We descided to create a custom sentiment analysis model to analize the patients first chat, and from there descided which of the psychiatric standardized questionairs to present to the patient. To make the experienced more personalized, we used Google Gemeni to tailor the questions asked in a way that feels conversational and natural. After the patient has completed the questionair, we then use simple prompt engineering to get a general score of the patients wellbeing, possible courses of action, and resources near the patient.
Note
Improvements
Currently we rely on gemeni for the final assessment and scoring, simply due to the 24 hour time constraint and limited team experience. Ideally this would use a custom model, hosted with our existing models API.
