S.A.M

• 

  Initial Screen

• 

  Generated Response

Inspiration

• Inspired by one of out group member's experience working with vulnerable individuals
• Access to mental health services is not easily available. Counselling can be expensive, and the friction to get into sections can be a big obstacle for many
• The third United Nation's Sustainable Development Goal, Good Health and Well-being

What it does

• Initially assesses a user's mental status to gauge how the conversation should go
• Interacts with the user while continuing to gather data about their emotional situation

How we built it

• Integrated Jupyter Notebooks APIs with a django front-end
• Based on Transformers models BART and DistilBERT

Challenges we ran into

• For a while the text generator model was spewing short and inaccurate answers
• The classification model for the responses was not able to discern between positive and negative emotions for some time while testing

What's next for S.A.M

• Adding support to languages other than English
• Tracking changes in users' mental health status over longer periods of time
• Implement text-to-speech in order to increase accessibility

Built With

• css
• html
• javascript
• jupyter
• pandas
• python
• transformer