Sentiment Analysis REST API

A production-style REST API that predicts sentiment (positive/negative) from text input, built with FastAPI and a scikit-learn NLP pipeline trained on 50,000 IMDB reviews.

Tech Stack

Layer	Technology
Framework	FastAPI
ML Model	Logistic Regression + TF-IDF
NLP	NLTK (stopword removal, text cleaning)
Server	Uvicorn (ASGI)
Testing	pytest + HTTPX

Project Structure

sentiment-api/
├── app/
│   ├── main.py            # FastAPI app & routes
│   ├── model_pipeline.py  # ML inference logic
│   ├── schemas.py         # Request/response models
│   ├── config.py          # Environment config
│   └── ml_models/         # Saved .pkl model
├── notebooks/
│   └── train_model.py     # Model training script
├── tests/
│   └── test_api.py        # Full test suite (6 tests)
├── conftest.py
└── requirements.txt

Quickstart

# 1. Clone
git clone https://github.com/KaminiSengar22/sentiment-analysis-api.git
cd sentiment-analysis-api

# 2. Virtual environment
python -m venv venv
venv\Scripts\activate        # Windows
source venv/bin/activate     # Mac/Linux

# 3. Install dependencies
pip install -r requirements.txt

# 4. Train the model (first time only, takes ~3 mins)
python notebooks/train_model.py

# 5. Run the API
uvicorn app.main:app --reload

API docs: http://127.0.0.1:8000/docs

API Endpoints

Method	Endpoint	Description
GET	/	Health check
GET	/health	Model status
POST	/predict	Predict sentiment

Example

Request:

curl -X POST "http://127.0.0.1:8000/predict" \
  -H "Content-Type: application/json" \
  -d '{"text": "This product is absolutely fantastic!"}'

Response:

{
  "text": "This product is absolutely fantastic!",
  "sentiment": "positive",
  "confidence": 0.9732,
  "message": "Prediction complete with 97.3% confidence"
}

Model Performance

• Dataset: IMDB Movie Reviews (50,000 samples)
• Algorithm: Logistic Regression with TF-IDF bigrams
• Accuracy: ~91% on test set

Running Tests

pytest tests/ -v