Disaster Tweets Classifier

This project is a submission for Kaggle's Natural Language Processing with Disaster Tweets competition. The goal is to predict whether a tweet is about a real disaster or not using machine learning techniques.

Project Overview

The dataset consists of tweets labeled as either disaster-related or not. The main focus of this project was to preprocess the data, extract features using BERT embeddings, and build a binary classification model using PyTorch.

Key Features

• BERT Embeddings: Extracting deep contextualized embeddings from the tweets using a pre-trained BERT model.
• Feature Engineering: Additional features like sentiment analysis were added to enhance the prediction power of the model.
• PyTorch Model: A neural network was used for binary classification.

Setup

1. Clone the repository:

git clone https://github.com/yourusername/disaster-tweets-classifier.git
cd disaster-tweets-classifier

2. Install dependencies:

pip install -r requirements.txt

3. Run the main script:

python main.py

Data

The dataset is sourced from the competition's Kaggle page. Please download the dataset from there and place it in the appropriate folder.