Inspiration

Social media plays a key role in crises, and automating the classification of tweets can help with faster decision-making and response.

What it does

It classifies tweets during crises to identify critical information like evacuation updates, aid requests, and damage reports.

How we built it

We experimented with Random Forest, Naive Bayes, CNN, and DistilBERT. DistilBERT performed best due to its ability to capture contextual meaning in text.

Accomplishments that we're proud of

We're most satisfied with achieving nearly 80% accuracy using the DistilBERT model, which handled crisis-related language effectively.

What's next

We would explore adding a map to visualize the geographical distribution of crisis-related tweets to aid real-time decision-making.

Built With

Share this project:

Updates