SONAR (Social Online Natural Disaster Alert Radar)

"Sonar an AI that finds the storm before it hits, detecting disasters from social media in real time."

Comment

Inspiration

Natural disasters can happen at any moment, and every second matters. We noticed that most alert systems rely on official news or reports, which can take hours to respond. But people on the ground are already posting about these events on Reddit way before the news picks it up. That’s what inspired us to build SONAR a tool that listens to real people online and helps send alerts faster. Our goal is to help communities and emergency responders take action sooner and save lives. What It Does

SONAR is a real-time disaster detection and alert system. It uses AI to scan social media (mainly Reddit) to figure out if a natural disaster is happening. If it detects something serious, it sends out alerts. These alerts show up on a live map through a website we built, and could also be sent to mobile devices.

Key Features:

  • Scans Reddit in real time for posts about natural disasters
  • Uses a custom AI model to tell how likely each post is about a real event
  • Gathers real data from Arduino sensors like temperature, smoke, and gas
  • Shows alerts on a live map with disaster clusters
  • Sends alerts when the system detects something risky

How We Built It

  • Reddit API: We used it to collect live posts from subreddits like r/CaliforniaNews.

  • AI/NLP Model: We trained a model using the CrisisNLP dataset to detect disaster posts. It reached about 96% accuracy.

  • Frontend: Built with React, next.js and Tailwind CSS for a clean, responsive UI.

  • Backend: Used Node.js and Python (Flask) to run the model and handle data and alerts.

Challenges We Faced

  • Accuracy: Social media can be messy, and it was hard to teach the AI to ignore jokes, fake news, or random posts.

  • Sensor Tuning: Getting accurate readings from our Arduino sensors took a lot of trial and error.

  • iPhone App: We tried building a mobile app but ran into issues and didn’t have enough time to finish it.

  • Rate Limits: Reddit doesn’t let you scrape too much data at once, so we had to work within their limits.

Accomplishments We're Proud Of

  • Built a working system that combines AI, live Reddit data, and real sensors

  • Got over 81% accuracy in disaster detection

  • Successfully connected all the parts — from hardware to the web UI

  • Created a real-time map that alerts people before the news does

What We Learned

  • NLP (Natural Language Processing) in the real world is tough but powerful

  • Crowdsourced data is valuable but needs to be filtered properly

  • Combining software and hardware makes a more reliable system

  • Working as a team helped us finish a big project in a short amount of time

What’s Next for SONAR

  • Add Twitter/X support to get even more real-time data

  • Build a full mobile app to send instant alerts

  • Team up with emergency services to try it in real life

  • Add more languages so it works worldwide

  • Connect with weather and satellite data for even better detection

  • Let users confirm alerts to improve accuracy over time

Built With

Share this project:

Updates