49% of those who suffer from anxiety and depression never reach out for help. We wanted a way we could find and classify people who are at risk of depressive behaviour and reach out to them when they need help the most.

What it does

FriendlyEar is a noninvasive, passive browser extension that checks to see if what you're typing signals depressive behaviour. When it notices warning signs, it sends you a text to check in and asks you how you're doing.

How we built it

FriendlyEar uses a natural language sentiment analysis of a dataset of online activity by depressed people. It then passes those sentiment scores through a neural network that models the sentiment features of depression. With that classification, FriendlyEar triggers a text message if it notices a pattern of depressive behaviour in real time.

What's next for FriendlyEar

The eventual goal is for a fully integrated chat bot that can respond to text messages based on user input rather than simply offering blanket advice / help. We also want to explore the possibility of introducing a human component -- helping people connect with therapists to make taking the first step toward getting help much easier.

Share this project: