Text Tone Classifier

This is a fine-tuned model that checks if a sentence sounds toxic or not. The project uses a small dataset of online comments. It fine-tunes a pretrained DistilBERT model to do binary classification: toxic or not toxic.

📁 Important Project files

├── main.py             # Trains the model
├── predict.py          # Usage of trained model
├── download_dataset.py # Downloads dataset
├── model/              # Saved model after training (ignored by git)

📊 Dataset

Dataset: Jigsaw Toxic Comment Classification Challenge from Kaggle.
Only train.csv is needed.
Use download_dataset.py to download it into the project folder.

🚀 How to run

1. Install libraries

pip install transformers datasets scikit-learn pandas accelerate kagglehub

2. Download dataset
   python download_dataset.py

3. Train the model
   python main.py

4. Check tone of your sentence
   python predict.py

Then type a sentence in the console. The model will say if it is toxic or not.