Transformer-Based Sentiment Analysis (YouTube Comments)
I fine-tuned BERT for YouTube-comment sentiment classification, served it through a Flask API, containerized it with Docker, and automated image publishing with GitHub Actions.
Result
F1 = 0.81 on 10,000+ imbalanced YouTube comments, CI/CD to Docker Hub
I fine-tuned BERT with PyTorch and Hugging Face on an imbalanced dataset of more than 10,000 YouTube comments.
I reached an F1 score of 0.81, exposed the model through a Flask REST API, and published repeatable Docker images through GitHub Actions. I also deployed an interactive Gradio demo on Hugging Face Spaces.
BERTPyTorchFlaskGitHub Actions