Depression is a big problem in our student community. It can be triggered by stress from hard exams, by relationship problems and many other issues that are common in the life of a university student. We wanted to make something that will help prevent depression and support people who suffer from it.
What it does
Psychologists say that it is often difficult to self-diagnose depression. Moreover, even if you feel depressed, you can be hesitant to talk to your friends about it, due to social stigma and apathy.
Our app gathers statistics of your activity on the Internet and detects first signs of depression. If it is possible that you are feeling unhappy it sends your trusted friends, whom you have selected beforehand, a reminder to talk to you about your feelings.
To motivate users to register we created a small text quest that explains depression in interactive form.
How we built it
Our app consists of 2 main components: Telegram bot that lets you register trusted friends and notifies them if you are feeling bad. It also has motivational text quest and a button that you can press if you are feeling depressed. The other component is a Chrome extension that gathers statistics of your activity and sends it to our server based on Django.
Our machine learning model takes into account your visited pages, Google search requests and Twitter posts to detect if you show any signs of depression. It can also analyze your photo to check if you look sad. All personal data are stored only in aggregated form to ensure privacy.
Challenges we ran into
Speaking of technical challenges, none of us has built Chrome extensions before, so creating one was a real challenge. We ran into several strange issues but were able to get around them.
Building a model that assesses a user's mental state was another big challenge: we take a few different factors into account, and it was quite difficult to calibrate all of them.
Last but not least, we also had to come up with a way to motivate people to register in our app and add trusted friends. Usually, people are not too concerned about depression when they I feeling happy. We thought that we can engage users with an interesting mini-game that will elicit interest in our app and show them why the issue is important.
Accomplishments that we're proud of
We've built quite a good model and integrated it into a neat app. We also came up with an interesting solution for how to motivate users to register in our app.
What we learned
We learned a lot about signs of depression and its treatment. We also learned how to build Chrom extensions and improved our skills in Natural Language Processing, Telegram Bot API and Django.
What's next for Depression < Tech
We think that our project can help people deal with mild depression induced by stress. Using our application we plan to gather a dataset labeled with self-reported "I am feeling depressed" responses from users, which will allow us to make even better model fined tuned especially for the purpose of detecting depression-associated behavior. We also want to analyze that data to find tips that help you prevent depression.