Vietnamese Sentiment Analysis Assistant 🎭

A powerful hybrid sentiment analysis system for Vietnamese text using PhoBERT + Rule-based approach, achieving 95.80% accuracy on diverse test cases.

✨ Features

• Hybrid Model: Combines deep learning (PhoBERT) with rule-based analysis for optimal accuracy
• Advanced Preprocessing: Noise removal, teencode normalization, word segmentation
• Streamlit Web App: Beautiful, user-friendly interface
• History Management: SQLite-based storage with export/import capabilities
• Multi-format Export: CSV, JSON, HTML, ICS calendar format
• Comprehensive Testing: 1000+ diverse prompts validation

🚀 Quick Start

Local Development

# Clone repository
git clone https://github.com/d0ngle8k/Extract-Prompt-To-Emotion.git
cd Extract-Prompt-To-Emotion

# Create virtual environment
python -m venv .venv
.venv\Scripts\activate  # Windows
# source .venv/bin/activate  # macOS/Linux

# Install dependencies
pip install -r requirements.txt

# Run the app
streamlit run app.py

🌐 Deploy to Streamlit Cloud

1. Fork this repository on GitHub
2. Go to Streamlit Cloud
3. Connect your GitHub account
4. Select this repository
5. Deploy! (Streamlit will automatically detect app.py as the entry point)

⚠️ Streamlit Cloud Notes:

• Free tier limitations: Model loading may take time on first run
• Data persistence: History data doesn't persist between sessions (cloud limitation)
• Memory usage: PhoBERT model requires ~2GB RAM, ensure adequate resources

📊 Performance

• Overall Accuracy: 95.80%
• Positive: 85.71%
• Negative: 98.72%
• Neutral: 98.70%

Tested on 1000 diverse Vietnamese prompts including standard text, toxic language, slang, and edge cases.

🏗️ Architecture

app.py              # Streamlit web interface
├── preprocessing.py # Text cleaning & segmentation
├── phobert_module.py # Hugging Face PhoBERT integration
├── rule_based.py    # Lexicon-based sentiment analysis
├── fusion.py        # Conditional model fusion
└── db_connector.py  # SQLite database operations

📁 Project Structure

Extract-Prompt-To-Emotion/
├── app.py                 # Main Streamlit application
├── preprocessing.py       # Vietnamese text preprocessing
├── phobert_module.py      # PhoBERT sentiment analysis
├── rule_based.py          # Rule-based sentiment analysis
├── fusion.py              # Model fusion logic
├── db_connector.py        # Database operations
├── requirements.txt       # Python dependencies
├── .streamlit/config.toml # Streamlit configuration
├── test/                  # Test suite
│   ├── test_1000_prompts.py
│   └── test_1000_prompts_refined.txt
├── phobert_finetuned/     # Pre-trained model files
└── README.md

🔧 Dependencies

• streamlit: Web app framework
• transformers: Hugging Face models
• torch: PyTorch for deep learning
• underthesea: Vietnamese NLP toolkit
• pandas: Data manipulation
• numpy: Numerical operations

🎯 Usage

1. Classification Tab: Input Vietnamese text and get instant sentiment analysis
2. History Tab: View past classifications with export options
3. Export Data: Download history in CSV, JSON, HTML, or ICS format
4. Import Data: Upload CSV/JSON files to add historical data

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

This project is open source. Feel free to use and modify.

🙏 Acknowledgments

• PhoBERT by VinAI Research

• Underthesea Vietnamese NLP toolkit

• Streamlit for the amazing web app framework

  🤝 Collaborator

Special thanks: chotxx - Vu Quoc Bao