🥡 WaimaiAnalyzer

WaimaiAnalyzer is a machine-learning-powered sentiment analysis tool for Chinese food delivery (外卖, waimai) reviews. Enter a review in any language — it gets translated to Chinese automatically — and the app instantly tells you whether the sentiment is positive 😋 or negative 😡, complete with a confidence percentage.

---

✨ Features

• 🌏 Multi-language support — Accepts reviews in any language and auto-translates them to Chinese (Simplified) via Google Translate before analysis.
• ✂️ Chinese text segmentation — Uses Jieba to tokenise Chinese text the same way the model was trained.
• 🤖 Naive Bayes classifier — A lightweight, fast CountVectorizer + MultinomialNB pipeline trained on ~10,000 real-world Waimai reviews.
• 📊 Confidence scores — Shows the exact probability for both positive and negative predictions so you always know how certain the model is.
• 🎈 Visual feedback — Balloons for positive reviews, clear error styling for negative ones.
• ⚡ Streamlit web UI — A clean, interactive browser-based interface that anyone can use without writing a single line of code.

---

🖼️ App Preview

Positive Review	Negative Review
😋 Positive! (92.3% confident)	😡 Negative! (87.1% confident)

---

🗂️ Project Structure

WaimaiAnalyzer/
├── main.py               # Streamlit web application
├── train.py              # Model training script
├── requirements.txt     # Python dependencies
├── data/
│   └── waimai_10k.csv   # ~10,000 labelled Waimai reviews (0 = negative, 1 = positive)
└── model/
    └── waimai_model.pkl # Trained Naive Bayes pipeline (generated by main.py)

---

🚀 Getting Started

1. Clone the repository

git clone https://github.com/Flo1632/WaimaiAnalyzer.git
cd WaimaiAnalyzer

2. Install dependencies

pip install -r requirements.txt

3. Train the model (optional — a pre-trained model is already included)

python main.py

This reads data/waimai_10k.csv, trains the classifier, prints the accuracy, and saves the model to model/waimai_model.pkl.

4. Launch the web app

streamlit run app.py

Open your browser at http://localhost:8501 and start analysing reviews!

---

🧠 How It Works

1. Input — The user types a review (any language) into the text field.
2. Translation — If the auto-translate checkbox is enabled, the text is sent to Google Translate and converted to Simplified Chinese.
3. Tokenisation — Jieba splits the Chinese text into individual word tokens (e.g. "外卖很棒" → ["外卖", "很", "棒"]).
4. Vectorisation — CountVectorizer converts the token list into a bag-of-words feature vector.
5. Classification — MultinomialNB predicts the sentiment label (0 = negative, 1 = positive) and returns class probabilities.
6. Display — The app shows the result, confidence score, and a per-class probability breakdown.

---

📦 Dependencies

Package	Purpose
streamlit	Interactive web UI
pandas	Data loading and manipulation
scikit-learn	ML pipeline (CountVectorizer + MultinomialNB)
joblib	Model serialisation
jieba	Chinese text segmentation
deep-translator	Automatic translation via Google Translate

Install all dependencies with:

pip install -r requirements.txt

---

📊 Dataset

The model is trained on waimai_10k.csv, a publicly available dataset of approximately 10,000 Chinese food-delivery reviews labelled as:

• 1 — Positive review
• 0 — Negative review

---

🤝 Contributing

Contributions, ideas, and feedback are welcome! Feel free to open an issue or submit a pull request.

---

📄 License

This project is open-source. See the repository for license details.