Movie Review Sentiment Analyzer

A BERT-based sentiment analysis web application that classifies IMDb movie reviews into Positive or Negative. It supports manual review input, live IMDb scraping for sentiment prediction, and dataset-based batch predictions with downloadable results. Deployed via Docker and Kubernetes, accessible via http://172.212.76.116

---

Features

• Manual Input Sentiment Analysis
• IMDb Review Scraper + Analysis
• Dataset Upload with Review Column Prediction
• Download Results as CSV
• Interactive Plots and Feedback

---

Model Details

• Architecture: BERT (bert-base-uncased)
• Trained on: IMDb 50K Dataset
• Frameworks: PyTorch, Transformers (HuggingFace)
• Preprocessing: Tokenization with padding, truncation (max length: 256)
• Training Time: 3 epochs on stratified 80/20 split

Performance Metrics

• Accuracy: 92.85%
• Precision/Recall/F1:

              precision    recall  f1-score   support

           0       0.94      0.92      0.93      5000
           1       0.92      0.94      0.93      5000

    accuracy                           0.93     10000
   macro avg       0.93      0.93      0.93     10000
weighted avg       0.93      0.93      0.93     10000

---

Dataset

• Source: IMDb 50K Movie Review Dataset
• Balanced: 25K Positive / 25K Negative
• Split: 80% Train, 20% Test

---

Project Structure

├── app.py                      # Streamlit web app
├── bert_sentiment_model/      # Saved BERT model + tokenizer
│   ├── config.json
│   ├── pytorch_model.bin
│   ├── tokenizer_config.json
│   ├── vocab.txt
│   └── special_tokens_map.json
├── Dockerfile
├── requirements.txt
├── README.md
├── notebook.ipynb             # Training and evaluation
├── test_scraper.py            # Unit tests for scraper functionality
├── utils/
│   ├── scraper.py             # IMDb scraping logic
│   └── predict.py             # Inference logic for prediction

---

Setup Instructions

1. Clone the repository

git clone https://github.com/your-username/movie-review-sentiment-analyzer.git
cd movie-review-sentiment-analyzer

2. Create environment & install requirements

pip install -r requirements.txt

3. Run locally

streamlit run app.py

---

Deployment

Docker

# Dockerfile
FROM python:3.10
WORKDIR /app
COPY . .
RUN pip install -r requirements.txt
CMD ["streamlit", "run", "app.py", "--server.port=8501", "--server.enableCORS=false"]

docker build -t bert-sentiment-app .
docker run -p 8501:8501 bert-sentiment-app

Kubernetes (AKS or Minikube)

• Use Kubernetes deployment YAMLs
• Deploy Docker image to AKS via Azure Container Registry or GitHub Actions

---

Endpoints (App Sections)

Feature	Description
Manual Input	Type any review manually and get sentiment
IMDb Scraper	Enter IMDb movie URL, scrape 10 reviews, analyze
Dataset Upload	Upload CSV with review column, get batch sentiment
Download Prediction	Download the annotated CSV

---

Tech Stack

• Frontend: Streamlit
• Backend: Python, Transformers, PyTorch
• Scraping: Selenium (ChromeDriver)
• Deployment: Docker, Kubernetes (AKS)
• Notebook: Jupyter, Kaggle Kernel

---

Sample Predictions

predict_sentiment("I love this movie!")  # Positive
predict_sentiment("This movie was terrible.")  # Negative

---

App Screenshots

Screenshot 1: App Overview

This is the main interface of the app, where users can choose between different input options.

Screenshot 2: IMDb Scraping

This screenshot shows the IMDb scraping feature where users can enter an IMDb movie URL to fetch reviews and analyze sentiment.

Screenshot 3: Manual Input Sentiment Analysis

Here, users can type any movie review manually, and the app will predict its sentiment as positive or negative.

---

External Access

• Public IP: http://172.212.76.116
• Domain: https://moviereviewsentimentanalyzer.com.np

---

Author

Sangam Paudel

• Email: sangampaudel530@gmail.com
• Kaggle: @sangampaudel530

---

License

MIT License. See LICENSE for more information.

---

Acknowledgements

• IMDb Dataset
• HuggingFace Transformers
• Streamlit for app interface
• Azure Kubernetes Service (AKS) for deployment