This project provides a full pipeline for sentiment analysis on Twitter data using a BERT-based model.
- Data loading and preprocessing
- Tokenization
- Model training and evaluation
- Reproducible results with fixed seeds
src/- Source code (data utilities, model, training, evaluation)data/- Training and validation CSV filesmodels/- Saved model checkpointsoutputs/- Evaluation logs and confusion matricesrun.py- Entry point for the full pipeline
- Install dependencies:
pip install -r requirements.txt
- Train and evaluate:
Use
python run.py
--helpfor configurable options.