# 💬 CommentSense — Customer Feedback Classification App

Try the live demo: **[CommentSense on Streamlit](https://cgbr5grkj7uddchfrspqk8.streamlit.app/)**  

CommentSense is a web app to automatically classify customer feedback into **praise**, **complaint**, or **toxic**. It supports both real-time text input and batch CSV uploads, and is built using **Streamlit** + **scikit-learn** with TF-IDF + Linear SVM (calibrated) under the hood.

---

## 🚀 Live Demo

- **Single Text** tab: Paste or type a comment and get immediate classification + confidence.  
- **Batch (CSV)** tab: Upload a CSV with a `text` column; the app shows predictions side-by-side and lets you download the results as a new CSV.  
- (Note: The live demo URL currently shows an internal error, which may be due to redeployment; you can clone and run locally for full experience.)

---

## 🛠 Technology Stack

| Component         | Technology / Library        |
|-------------------|-------------------------------|
| Web UI            | Streamlit                     |
| Feature Extraction| TF-IDF (from scikit-learn)    |
| Model             | Linear SVM (Calibrated)       |
| Persistence       | joblib / pickle                |
| Data Handling     | pandas, numpy                  |
| Additional Support| pyarrow (for Streamlit dataframe rendering) |

---

## 📁 Project Structure

commentsense/ ├─ app.py # Streamlit app for UI & serving ├─ train.py # Script to retrain the model ├─ requirements.txt # All Python dependencies ├─ README.md # This file ├─ .gitignore ├─ data/ │ └─ customer_feedback.csv # Example training data └─ models/ ├─ vectorizer.pkl └─ classifier.pkl

---

## 🛠 Setup & Run Locally

1. **Clone the repo**
   ```bash
   git clone https://github.com/<your-username>/commentsense.git
   cd commentsense

2. Create Python environment (3.11 recommended)

   python -m venv .venv
   .venv\Scripts\activate     # on Windows
   # or
   source .venv/bin/activate  # on macOS/Linux

3. Install dependencies

   pip install --upgrade pip
   pip install -r requirements.txt

4. (Optional) Retrain the model

   python train.py

5. Run the app

   streamlit run app.py

   Visit http://localhost:8501 in your browser.

---

✅ Features

• Single comment classification: Type or paste feedback and get instant label + confidence
• Batch prediction: Upload CSV with a text column → get predictions + download results
• Retraining pipeline: Use your own dataset to re-fit the model
• Clean, robust pipeline: TF-IDF + calibrated Linear SVM ensures interpretable and reliable predictions
• Visual display: Streamlit’s nice table UI with pyarrow support

---

🎯 Use Cases & Applications

• Customer support automation (detect complaints, escalate toxic feedback)
• Marketing / product feedback analysis
• Sentiment segmentation for surveys
• Dashboard integration in BI tools

---

📜 License

MIT — see LICENSE

Ah, and if you like, I can also generate a Deploy-to-Streamlit badge (HTML + Markdown) link you can put at the top of your README so viewers can instantly launch your app. Do you want me to add that?