A countless number of people are affected by mental health and have their lives drastically impacted. Suicide accounts for a quarter of all deaths in young people, most suicide attempts come from lower-income individuals and sessions to help individuals suffering from mental health problems are costly; ranging from $100 to $200 per session.
We designed a product that makes mental health support accessible to ANYONE at ANY TIME.
What it does
The software provides a text interface and a voice interface for communication with the end-user (depending on their preference). Individuals suffering from mental health problems can speak with a conversational AI about their problems. This can circumvent the problem of long wait times to see therapists or people feeling uncomfortable speaking to another human about their problems.
How we built it
The back-end conversational AI uses a GPT transformer-based approach to converse. A pre-trained conversational GPT model was used along with transfer learning to fine-tune the model to behave similar to a therapist. We used a dataset with 800 therapist questions and answers to provide additional training for the model.
We used a Flask API to connect this back-end to our front end.
The front-end consists of a React-Native mobile application that is very simple to use. It incorporates a Twilio online text communication platform and a Dasha.AI voice agent to interface with the conversational AI.
Challenges we ran into
- Time management
- Getting the correct format for transfer learning
- Setting up Twilio and Dasha.AI
Accomplishments that we're proud of
We provide a product that can help other people and this is something we couldn't be more proud of.
What we learned
- Communication is important in dev projects
- Splitting up the work is necessary - don't step on each others' feet
- Pre-trained conversational AI models are powerful as-is
- There are many resources out there for end-user interaction such as Twilio and Dahsa.AI
What's next for MentAIly
We are committed to continuing the development of MentAIly. We look to improve the back-end through additional learning for conversational AI. We also hope to improve the user interface after receiving feedback.