This paper explores the application of reinforcement learning (RL), specifically utilizing the Deep Q-Network (DQN) algorithm to improve chatbot performance in the context of emotional support and mental health care. Although chatbots are widely used across various fields, achieving human-like conversational abilities remains a significant challenge, especially in sensitive areas like therapy and emotional well-being. This paper seeks to overcome these challenges by developing a therapist chatbot designed to offer empathetic and supportive interactions for individuals managing their emotional health. The framework proposed involves a chatbot agent that learns to optimize cumulative rewards through dialogue interactions, which is essential for enhancing user satisfaction, engagement, and task completion in therapeutic conversations. It tackles key challenges such as state representation, action space definition, and reward shaping while adapting the DQN algorithm to the dialogue domain. By effectively balancing exploration and exploitation and improving policy, the chatbot is designed to serve as a supportive tool, providing users with personalized guidance and emotional support.