The gigantic increase in data amounts over the timeline leads to deep learning, as it is a promising approach to resolving the purpose of data processing and management. Due to the limitless scope in all sectors, chatbots have recently been in keen interest and are on the roadmap of improvement. Additionally, deep learning models are nowadays a promising option for feature learning. This paper has evaluated a hybrid deep feature learning model to implement for Chatbot development. The paper evaluates this proposed approach with varying batch sizes for feature learning to measure the impact of batch size on performance. It concluded that the model performance is uplifted for batch size 64 and performs least for batch size 256.