🔮 SentiScan

Product Review Intelligence — ML-Powered Sentiment Analysis

Predict star ratings from product reviews using a TF-IDF + Linear Regression pipeline trained on 190,000+ real Flipkart reviews.

---

✨ Features

• Real-time Text Analysis — Paste any product review and instantly get a predicted star rating (1–5) with confidence score and animated visual feedback
• CSV Batch Processing — Upload any CSV file with review data and get a full sentiment distribution, star breakdown, and row-by-row preview
• Flexible Column Detection — Works with standard columns (ProductName, Review, Summary) or any custom column names automatically
• Production Ready — Deployed on Render with Gunicorn, CORS-enabled REST API, and a fully responsive frontend

---

🚀 Live Demo

https://sentiscan-product-review-intelligence.onrender.com

⚠️ Hosted on Render's free tier — first load may take 30–60 seconds to wake up from idle.

---

📸 Preview

Text Analysis	CSV Batch Results
Paste a review → get star rating + confidence	Upload CSV → full distribution + preview table

---

🧠 How It Works

Raw Review Text
      │
      ▼
Text Cleaning (lowercase, remove special chars)
      │
      ▼
TF-IDF Vectorization (5,000 features)
      │
      ▼
Linear Regression Model
      │
      ▼
Predicted Star Rating (1–5) + Confidence Score

The model was trained on 189,874 Flipkart product reviews combining ProductName, Review, and Summary columns into a single text feature. The predicted float is clipped and rounded to the nearest integer in the 1–5 range.

---

🗂️ Project Structure

SentiScan/
├── app.py                  # Flask backend — REST API
├── requirements.txt        # Python dependencies
├── sentiment_model.pkl     # Trained model + TF-IDF vectorizer
└── static/
    └── index.html          # Frontend — HTML, CSS, Vanilla JS

---

⚙️ API Reference

GET /api/health

Check if the server and model are running.

Response:

{
  "status": "ok",
  "model_loaded": true
}

---

POST /api/predict

Predict sentiment for a single text input.

Request:

{
  "text": "Amazing product! Totally worth the price."
}

Response:

{
  "score": 5,
  "raw": 4.823,
  "label": "Very Positive",
  "color": "#10b981",
  "confidence": 82.3
}

---

POST /api/analyze-csv

Batch analyze an entire CSV file.

Request: multipart/form-data with a file field containing a .csv file

Response:

{
  "total": 200,
  "distribution": {
    "1": { "count": 20, "pct": 10.0 },
    "5": { "count": 95, "pct": 47.5 }
  },
  "summary": {
    "good":    { "count": 130, "pct": 65.0 },
    "neutral": { "count": 30,  "pct": 15.0 },
    "bad":     { "count": 40,  "pct": 20.0 }
  },
  "preview": [...]
}

---

🏃 Run Locally

1. Clone the repo

git clone https://github.com/Samin-Saikia/SentiScan-Product-Review-Intelligence.git
cd SentiScan-Product-Review-Intelligence

2. Install dependencies

pip install -r requirements.txt

3. Run the server

python app.py

4. Open in browser

http://localhost:5000

---

📊 Dataset

Trained on the Flipkart Product Reviews dataset containing 189,874 entries across 812 unique products.

Column	Description
ProductName	Name of the product
Price	Listed price
Rate	Star rating (1–5) — target variable
Review	Short review title
Summary	Full review text

---

🛠️ Tech Stack

Layer	Technology
ML Model	scikit-learn — TF-IDF + Linear Regression
Backend	Python, Flask, Gunicorn
Frontend	HTML5, CSS3, Vanilla JavaScript
Deployment	Render

---

📈 Model Performance

Metric	Value
Algorithm	Linear Regression
Vectorizer	TF-IDF (5,000 features)
Train/Test Split	80% / 20%
Target	Star Rating (1–5)

---

👤 Author

Samin Saikia

• GitHub: @Samin-Saikia

---

📄 License

This project is open source and available under the MIT License.

---

_{Built with 🔮 and deployed to production}