Homepage
Dashboard
Whatsapp Notification
Architecture

Inspiration

Based on data collected by BPS, Central Java in top of list number of suicide by 331 of 812 cases in 2016. According to that data, we need to do prevention so the cases can be reduced. And it’s everyone responsibility to prevent their loved one to stop hurting their self, even anyone around the location of suicide attempt have to help.

What it does

Detect suicidal tweets from specific location
Push notification of suicidal tweets to registered users via whatsapp
Classify tweet as suicidal or not

How we built it

Clone the repository
Build the docker image on each directory
Run

We use github workflow to automate deployment, every commit to main branch will trigger the workflow to deploy in our server

Stacks

Frontend Stack

Django

Backend Stack

Go -> Backend for Frontend and Pushing Notification
Flask -> Classification Engine
Twillio -> Pushing Notification

ML Stack

Python
Tensorflow
Glove Embedding

Deplyoment

Docker
Kubernetes
Github Action CI/CD

Challenges we ran into

Hard to collect dataset
Twitter API limitation
Need to wait of verification to registering the number of whatsapp business in Twillio

Accomplishments that we're proud of

This is first hackaton for our several member and we finished it!
We used a lot of stacks in this project and we happy we can practice what we have learned
Tried our very best to always stick with the best practices

What we learned

Train model from dirty data of text
Number of suicide in Indonesia especially in Central Java is high
Manage multiple services in only 2 days

What's next for Supriadi

Adding list of locations
Improving classification engine accuracy
Use to manage tweet from wide world
Buy premium twitter and whatsapp API
Auto direct message of the suicidal tweet user to give professional recommendation