Sentiment Analysis 2025

End-to-End Sentiment Analysis on Amazon Reviews using AWS SageMaker and HuggingFace Transformers

---

Project Overview

This project demonstrates a full machine learning workflow for sentiment analysis on Amazon product reviews, including:

• Cloud-based training on AWS SageMaker with HuggingFace Transformers
• IAM and resource management for real-world deployment
• Model artifact management and local inference (batch and single predictions)
• Visualization of sentiment predictions

---

Project Structure

sentiment-sagemaker/ ├── input/ # Input dataset (CSV)

├── sagemaker_trained/ # Trained model artifact from SageMaker

├── src/ # Source scripts

├── Scripts/ # Utility scripts

├── local_inference.py # Script for extracting and running local predictions

├── batch_prediction.py # Batch inference & visualization script

├── .gitignore

└── README.md

---

How to Run Locally

1. Clone the repo:

   git clone https://github.com/<your-username>/Sentiment_analysis_2025.git
   cd Sentiment_analysis_2025

2. Install requirements (use virtualenv!):

   pip install -r requirements.txt

   (Add your requirements.txt using pip freeze > requirements.txt if needed)

3. Run local inference:

   python local_inference.py

   • Extracts the SageMaker model and predicts sample reviews.

4. Batch prediction & visualization:

   python batch_prediction.py

   • Processes the whole dataset and shows result charts.

---

Screenshots

---

Key Learnings

• Real-world AWS resource, quota, and IAM troubleshooting
• How to train and deploy Transformer models at scale
• Automating ML pipelines for reproducibility
• Visualizing and interpreting results

---

Tech Stack

• Python 3.11+
• HuggingFace Transformers & PyTorch
• Pandas, Matplotlib
• AWS SageMaker (training), S3 (storage)
• IAM roles and permissions

---

Author

• Souritra Banerjee

---

For questions, open an issue or contact me on LinkedIn