AI Chatbot

Technical Challenges to Solve

You can use these datasets for training your intent classification and slot filling models, or create your own custom dataset for the e-commerce domain.

Your project must include all of the following components. This is a comprehensive NLU project covering the full chatbot development pipeline.

Intent recognition is the core NLU component that classifies user messages into predefined categories. Modern approaches use transformer-based models fine-tuned on domain-specific data.

Model Architecture Options

Evaluation Metrics

Slot filling extracts relevant entities from user messages using sequence labeling. The BIO (Beginning-Inside-Outside) tagging scheme is the standard approach.

BIO Tagging Scheme

Example Annotation

Multi-turn conversations require tracking dialogue state across user turns. The dialogue manager maintains context, filled slots, and conversation history.

Dialogue State Components

Conversation Flow Example

Deploy your chatbot as a REST API using Flask or FastAPI. The API should handle stateful conversations with session management.

API Endpoints

Request/Response Format

{
  "session_id": "abc123",
  "message": "Check order #12345",
  "metadata": {
    "user_id": "user_001",
    "channel": "web"
  }
}

{
  "session_id": "abc123",
  "response": "Order #12345 is shipped!",
  "intent": "order_status",
  "confidence": 0.95,
  "slots": {"order_id": "#12345"}
}

Create a public GitHub repository with the exact name shown below:

Required Project Structure

ai-chatbot/
├── notebooks/
│   ├── 01_data_preparation.ipynb       # Data loading and preprocessing
│   ├── 02_intent_classification.ipynb  # Intent model training
│   ├── 03_slot_filling.ipynb           # NER model training
│   └── 04_evaluation.ipynb             # Model evaluation
├── src/
│   ├── nlu/
│   │   ├── intent_classifier.py        # Intent classification module
│   │   └── slot_filler.py              # Entity extraction module
│   ├── dialogue/
│   │   ├── state_tracker.py            # Dialogue state management
│   │   └── response_generator.py       # Response templates
│   ├── api/
│   │   ├── app.py                      # Flask/FastAPI application
│   │   └── routes.py                   # API endpoints
│   └── utils.py                        # Helper functions
├── data/
│   ├── intents.json                    # Intent training data
│   └── entities.json                   # Entity annotations
├── models/
│   ├── intent_model/                   # Saved intent classifier
│   └── ner_model/                      # Saved NER model
├── tests/
│   └── test_chatbot.py                 # Unit tests
├── requirements.txt                    # Python dependencies
├── Dockerfile                          # Optional: containerization
└── README.md                           # Project documentation

Your project will be graded on the following criteria. Total: 800 points.

Grading Levels