This Project is about to develop parsing model for TOD system that can accurately detect intent and do slot fullfilling
• Developed hybrid parsing model combining RoBERTa and T5 model for intent detection and slot filling in TOD system. • Implemented preprocessing technique (to remove ASR disfluency), prefix tuning optimization technique, prompt engineering which significantly improve model performance acheiving 89% exact match accurcay