NLU Dialogue Management — Multi-Turn Conversation State Tracking

I will implement a dialogue management system for your conversational AI — covering conversation state tracking across multi-turn exchanges, slot filling with clarification prompts for missing required information, context carry-forward for references to earlier conversation turns, fallback handling for low-confidence predictions, and conversation reset logic. Multi-turn conversation management is where most NLU implementations fall apart — handling "I want to book a flight from London... to Paris... for next Tuesday" as three separate turns filling slots progressively requires state tracking that single-turn intent classification alone cannot provide.

The dialogue management covers conversation state schema design, slot filling with configurable required/optional slots, clarification dialogue for missing required slots, context carry-forward for pronouns and references, fallback and disambiguation dialogues, and integration with your NLU classifier and backend action layer.