Restaurant Review Sentiment Analyser

An AI-powered web application that analyses restaurant reviews and predicts whether the sentiment is positive or negative, complete with confidence scores and witty chef responses.

---

Features

Feature	Description
Sentiment Analysis	Multinomial Naive Bayes classifier trained on 1 000 restaurant reviews
Confidence Score	Probability-based confidence meter for every prediction
Voice Input	Speak your review using the browser's Speech Recognition API
Modern UI	Dark glassmorphism design with micro-animations (no Bootstrap)
FastAPI Backend	High-performance async Python backend
Docker Ready	Multi-stage Dockerfile for one-command deployment
Security	Non-root Docker user, env-based config, input validation

---

Architecture

+-------------------------------+   POST /api/predict   +-----------------------------+
|   React Frontend              | <-------------------> |   FastAPI Backend            |
|   (Vite + JSX)                |        JSON           |   (Uvicorn)                 |
|                               |                       |                             |
|   - Client Validation         |                       |   - TF-IDF Vectoriser       |
|   - Voice Input               |                       |   - MNB Classifier          |
|   - Animated Results          |                       |   - Error Handling          |
+-------------------------------+                       +-----------------------------+

---

Model Information

For full model documentation including architecture rationale, hyperparameter details, evaluation metrics, limitations, and future improvements, see MODEL.md.

Quick Summary

• Algorithm: Multinomial Naive Bayes (alpha=0.2)
• Features: TF-IDF Vectoriser with auto-selected vocabulary size
• Preprocessing: Text cleaning, contraction expansion, stop-word removal, lemmatisation
• Metrics: Accuracy ~77-79%, ROC-AUC ~0.84-0.86

---

Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• pip

1. Clone & Install

git clone https://github.com/Ishwarpatra/Restaurant-Reviews-Sentiment-Analysis-Deployment.git
cd Restaurant-Reviews-Sentiment-Analysis-Deployment
pip install -r requirements.txt

2. Train the Model (optional -- pre-trained .pkl files included)

python "Restaurant Reviews Sentiment Analyser - Deployment.py"

This will:

• Run vocabulary size analysis
• Train the final model with optimal features
• Print full evaluation metrics
• Save cv-transform.pkl and restaurant-sentiment-mnb-model.pkl

2b. Validate the Dataset

python data_validation.py

2c. Hyperparameter Tuning (optional)

python hyperparameter_tuning.py

Runs GridSearchCV with cross-validation over alpha and max_features.

3. Build the Frontend

cd client
npm install
npm run build
cd ..

4. Run the Server

python main.py

Open http://localhost:5000 in your browser.

Environment Variables

Variable	Default	Description
DEBUG	false	Enable debug mode & CORS wildcard
PORT	5000	Server port
ALLOWED_ORIGINS	http://localhost:5173	Comma-separated CORS origins

---

Docker Deployment

# Build the image
docker build -t restaurant-sentiment .

# Run the container
docker run -p 5000:5000 -e DEBUG=false restaurant-sentiment

Works on any cloud provider -- AWS ECS, Azure Container Instances, GCP Cloud Run, DigitalOcean, etc.

---

Testing

# Install test dependencies (already in requirements.txt)
pip install pytest pytest-asyncio pytest-cov httpx

# Run all tests
pytest

# Run with coverage report
pytest --cov=. --cov-report=term-missing

# Run a specific test file
pytest tests/test_api.py -v

---

Project Structure

├── main.py                        # FastAPI backend (API + static file serving)
├── preprocess.py                  # Text preprocessing module
├── data_validation.py             # Dataset quality checks
├── hyperparameter_tuning.py       # GridSearchCV tuning script
├── Restaurant Reviews Sentiment
│   Analyser - Deployment.py       # Training & evaluation script
├── Restaurant_Reviews.tsv         # Dataset (1 000 reviews)
├── restaurant-sentiment-mnb-model.pkl  # Trained classifier
├── cv-transform.pkl               # Fitted TF-IDF vectoriser
├── requirements.txt               # Python dependencies
├── environment.yml                # Conda environment definition
├── Dockerfile                     # Multi-stage Docker build
├── Procfile                       # Heroku / PaaS entrypoint
├── render.yaml                    # Render.com deployment config
├── MODEL.md                       # Model architecture documentation
├── CONTRIBUTING.md                # Contribution guidelines
├── tests/                         # Test suite (pytest)
│   ├── conftest.py                # Shared fixtures
│   ├── test_preprocess.py         # Preprocessing tests
│   ├── test_api.py                # API endpoint tests
│   └── test_data_validation.py    # Data validation tests
├── models/                        # Versioned model artefacts
│   └── README.md                  # Versioning instructions
├── client/                        # React frontend (Vite)
│   ├── index.html
│   ├── package.json
│   ├── vite.config.js
│   └── src/
│       ├── main.jsx
│       ├── App.jsx
│       ├── index.css
│       └── components/
│           ├── Navbar.jsx
│           └── Footer.jsx
└── README.md

---

Security Notes

• No debug=True in production -- Debug mode is toggled via the DEBUG env var (defaults to false)
• CORS -- Only allows configured origins in production; wildcard only in debug mode
• Input Validation -- Server-side via Pydantic validators + client-side JS validation
• Docker -- Runs as non-root appuser
• Dependencies -- Pinned to latest stable versions

---

Contributing

See CONTRIBUTING.md for setup instructions, coding standards, testing guidelines, and PR process.

---

Reproducibility

Method	Command
pip	pip install -r requirements.txt
conda	conda env create -f environment.yml
Docker	docker build -t restaurant-sentiment .

Model versioning is done via the models/ directory. See models/README.md for conventions.

---

Licence

MIT (c) Ishwarpatra

---

Star this repo if it helped you!