DrBaymaxMini's journey started as an automated chatbot for suicide detection. In today's digital world, your phone probably knows the state of your mental health better than your spouse. This is our attempt to create a chatbot for studying your mental health, diagnose from among a range of mental illnesses and provide suggestion or recommendations for getting better.

What it does

You can talk to DrBaymaxMini, who will ask you a bunch of probing questions to gauge the state of your mental health. If you are feeling sad, DrBaymaxMini will be activated from your own Google-home device and will understand what you need to heal. :) It will also try to help you out to improve on, and will guide you through your this phase.

How we built it

  1. We started our journey with a traditional chatbot using a python NLP library - nltk - and a tensorflow abstraction - tflearn. This chatbot communicated via text.
  2. We introduced speech to text and text to speech using DialogFlow to integrate our chatbot with Google Assistant. We plan to use DrBaymaxMini with our Google Home mini device.

Challenges we ran into

  1. Setting up Dialog-flow to accept dynamic responses from our chatbot engine was challenging. We faced a lot of integration issues, but all's well that ends well!
  2. Collecting accurate mental health symptoms and diagnosis information and mapping conversation flows was challenging.

Accomplishments that we're proud of

The team spent an incredible amount of effort in getting Google Assistant as a part of DrBaymaxMini. We did not give up!

What we learned

This was our very first experience working with a chatbot and google assistant (which is one of the best chat robots ever!). We learned the intricacies of integrating with google assistant. We also learnt how to accurately map conversation flows with precise contexts.

What's next for DrBaymaxMini

We plan to expand the scope of mental health diagnosis to a wide variety. Adapting to flexible communication. Better and robust natural language processing.

Built With

Share this project: