Amrit74/chaybot-nlu
0
๐ค Chatbot NLU Trainer & Evaluator
A comprehensive platform for training, evaluating, and managing NLU (Natural Language Understanding) models for chatbots.
Features
๐ฏ Core Features
- Intent Classification Training - Train models to understand user intentions
- Entity Recognition - Extract key information from user messages
- Multi-Backend Support - Train with HuggingFace, Rasa, or spaCy
- Model Evaluation - Comprehensive metrics and confusion matrices
- Active Learning - Improve models with uncertain predictions
- Model Versioning - Track and manage different model versions
๐ Analytics & Monitoring
- Real-time training progress
- Performance metrics visualization
- Confidence score analysis
- Intent distribution charts
๐ง Built With
- Frontend: Gradio for interactive UI
- Backend: Python with scikit-learn, transformers
- Visualization: Plotly for charts and graphs
- Storage: JSON-based data management
How to Use
1. Training Tab
- Upload your training data (JSON format)
- Select backend (HuggingFace/Rasa/spaCy)
- Configure training parameters
- Start training and monitor progress
2. Evaluation Tab
- Test your trained model
- View performance metrics
- Analyze confusion matrix
- Check per-intent statistics
3. Prediction Tab
- Enter text to classify
- View predicted intent and confidence
- See alternative predictions
- Get entity extraction results
4. Active Learning
- Review uncertain predictions
- Provide correct labels
- Retrain model with feedback
- Improve model accuracy
Sample Data Format
[
{
"text": "I want to book a flight to New York",
"intent": "book_flight",
"entities": [
{"entity": "destination", "value": "New York"}
]
},
{
"text": "Cancel my reservation",
"intent": "cancel_booking",
"entities": []
}
]Links
- GitHub Repository: Chatbot-NLU-Trainer--Evaluator
- Full Application: React + Node.js Version
Author
Amarjit Kumar
- GitHub: @Amarjit99
License
MIT License - See LICENSE file for details
This is a demo version optimized for Hugging Face Spaces. For the full-featured application with MongoDB integration, user management, and advanced features, check out the GitHub repository.
