ارﺗﻘﺎي ﺳﯿﺴﺘﻢ ﮔﻔﺘﮕﻮي وﻇﯿﻔﻪ ﮔﺮا از ﻃﺮﯾﻖ ﺗﻨﻈﯿﻢ ﺳﺮﯾﻊ و ﻧﻤﺎﮔﺮ ﺳﺎزي ﮐﺎرﺑﺮ

Task-oriented dialogue system (TODS) , UserProfile , Right To Be Forgotten , pro‎mp‎t-tuning , Collaborative Filtering

Improving task oriented dialog system using pro‎mp‎t tuning an‎d user profiling

مهندسي نرم افزار

Task-oriented dialogue systems play a central role in delivering personalized services through human-AI interactions. However، these systems face significant challenges، including insufficient personalization، the cold start problem، an‎d privacy concerns. These issues hinder their ability to deliver seamless an‎d reliable user experiences. Furthermore، limited user data often exacerbates the difficulty of creating effective personalized interactions، while lack of compliance with global privacy regulations complicates their deployment. To address these challenges، we propose an approach، named ”MindMeld،” that integrates advanced techniques for personalization، user profiling، an‎d privacy. The system leverages pro‎mp‎t-tuning of large language models to quickly adapt to specific user needs، even in data-limited scenarios. The system includes item-based collaborative filtering، enhanced with sentiment analysis to overcome the cold start problem an‎d improve recommendation accuracy. In addition، MindMeld implements the right to be forgotten، enabling users to dynamically delet‎e their data without compromising system performance. This ensures compliance with global privacy stan‎dards an‎d increases user trust. By combining few-shot learning with rapid tuning، the system achieves robust performance even in data-poor environments. The effectiveness of MindMeld was eva‎luated using comprehensive metrics، including Perplexity، Distinct، Success rate، Completion rate، an‎d User engagement score. The results show improvements over existing approaches. Specifically، the success rate increased by seventeen percent an‎d Distinction improved by twelve percent compared to similar studies. These findings highlight the system’s ability to deliver personalized an‎d efficient interactions while maintaining user privacy an‎d trust. Overall، MindMeld provides a scalable an‎d ethical solution for task-based conversational systems، bridging gaps in personalization، privacy، an‎d usability.

5

153996