AI-Based Customer Support Chatbot for Intelligent and Real-Time Assistance Using Transformer Models and Reinforcement Learning

Abstract

Customer support services are increasingly adopting Artificial Intelligence (AI) technologies to improve service quality, reduce operational costs, and provide continuous assistance. Traditional rule-based chatbots often fail to understand complex user queries and maintain contextual conversations. This study proposes an AI-Based Customer Support Chatbot integrating Transformer-based Natural Language Processing (NLP), Bidirectional Encoder Representations from Transformers (BERT).The increasing demand for real-time customer support necessitates intelligent and adaptive conversational systems. This study proposes an AI-Based Customer Support Chatbot that integrates Transformer Models, Sentiment Analysis, and Reinforcement Learning to enhance customer service performance. The system employs a BERT-based intent recognition framework for accurate query classification, sentiment-aware response generation for personalized interactions, and reinforcement learning for continuous dialogue optimization. The chatbot was evaluated using 100,000 customer support queries collected from multi-domain datasets. Experimental results demonstrated superior performance compared to traditional approaches, achieving an intent classification accuracy of 97.8%, response time of 220 ms, and customer satisfaction score of 95%. Statistical analysis confirmed the effectiveness of the proposed framework, with ANOVA results indicating significant performance differences (F = 24.68, p = 0.0001) and regression analysis showing strong predictive capability (R² = 0.910, p = 0.0001).