Inspiration

We initially had the idea for an emotional well-being AI but it wasn’t thought out enough. Then we noticed the Best Veteran’s Hack category and it clicked. Our immediate family members are veterans and we’re all aware of posts talking about veterans taking their own lives. After doing some research and seeing how alarming the statistics are, we decided to focus this app on mental health. By having the interaction over more normal channels such as text or Facebook messenger as opposed to it’s own app, it’s much less invasive/intrusive and a lot easier to adopt.

What it does

Interacts with user over text or Facebook Messenger. Will check on user periodically throughout the day, can be set or randomized to a degree. Will ask how the user is doing, will ask if the user is prepared for events coming up on their Google calendar, will ask the user how their day went (typically at end of day). Learn mood based on text replies. Try to hold conversation to a degree. Utilize leading questions to get user to vent (healthily).

How we built it

User will text number to sign up for service (with Twilio API). Adds user to database, ask if user would like to talk over Facebook as well, sends link to Facebook. App will send text messages to user, takes reply, gives option to continue conversation naturally (any input besides ’bye’). On backend side analyze user replies with Microsoft Cog Sci API and Watson to analyze emotional response for better feedback. Keep count of replies in conversation chains and send messages at times where they have longer conversations more frequently. Give option to have similar experience with Amazon Alexa. Host website with Microsoft Azure and IBM Blue Mix.

Challenges we ran into

Integrating multiple APIs, namely Google Calendar API with Python.

Accomplishments that we're proud of

First time we attempted an AI chatbot.

What we learned

How granular metrics work and why they’re important.

What's next for MOO:D

Implementing deep learning to create more natural conversation flow with recursive and convolutional neural networks.

Share this project:
×

Updates