Disaster Response Classification Pipeline using NLP and Machine Learning

Disasters, whether natural or human-made, generate vast amounts of text-based communications on social media and other digital platforms. Effectively classifying these messages is crucial for timely responses from emergency services and humanitarian organizations. This paper presents a Disaster Response Classification Pipeline using Natural Language Processing (NLP) and Machine Learning (ML) to categorize messages into 36 predefined disaster-related categories. The project follows a structured ETL (Extract, Transform, Load) pipeline to process raw messages, an ML pipeline to train a classification model, and a Flask-based web application for real-time message classification. The dataset, sourced from Figure Eight, contains multilingual messages from past disasters. Experimental results demonstrate the effectiveness of machine learning models in classifying messages with high accuracy, making this system valuable for real-world disaster management efforts.