How prayesh13/youtube-comments-analysis works
Project type and purpose
• This is an end‑to‑end machine‑learning pipeline for YouTube comment sentiment analysis, packaging a trained model behind a Flask API and exposing it via a Chrome plugin. It uses DVC for data/model versioning and GitHub Actions for CI/CD.
File structure
• Top‑level orchestration and config:
• Makefile, Dockerfile, dvc.yaml/dvc.lock → pipeline stages (data→train→evaluate)
• GitHub Actions (cicd.yaml) → CI/CD
• Data science pipeline:
• notebooks/ → exploration & prototyping
• src/data/ → data_ingestion.py, data_preprocessing.py
• src/features/ → feature engineering
• src/model/ → model_building.py, model_evaluation.py, register_model.py
• src/visualization/ → visualization scripts
• models/ & vocab.pkl → serialized artifacts
• Service layer:
• flask_app/ → app.py, requirements.txt → inference microservice
• Packaging & testing:
• setup.py, requirements.txt, tox.ini
• Documentation & reports: docs/, reports/
README for additional insights
• Follows cookiecutter‑datascience template
• Make targets (make data, make train) reflect DVC stages
• Emphasizes src/ modular structure