Online suicidal tendencies reporter

Helps report suicidal tendencies using NLP sentiment analysis, and connects users with help.

Comment

Inspiration

Sometimes it's way too late before someone discovers the suicidal tendencies of a person. There seem to be some patterns, in terms of writing style, frequency etc might help figure out if the mood of some person is drastically changing over time for the worse.

What it does

This is an attempt to identify suicidal tendencies of users online before it's too late. It identifies change in mood (similar to suicidal) and then tries to connect the user with Suicide prevention authorities

How I built it

I used React for the frontend, along with Socket.io for abstracting the socket protocol features in the chatroom. Flask for building the backend APIs to integrate it with the running Pytorch model. The Flask backend handles requests and returns a JSON response when triggered.

Challenges I ran into

  • Finding ways to detect sentiment over a period of time, ways to enumerate and formulate when to trigger help messages to the user.

  • Deploying the model for inference, issues with implementing the backend for running the model.

Accomplishments that I'm proud of

Running the PyTorch model locally and connecting with the frontend for inference.

What I learned

Building an API endpoint for the Pytorch model to work with the frontend.

What's next for Online suicidal tendencies reporter

  • Code release, open-source the underlying code which can be integrated into other apps by developers.
  • A tutorial to get people started with running the code in production on Cloud services like Azure or Google Cloud Platform.

Built With

Share this project:

Updates