A Deep Learning based Chatbot for Academics using NLP

Artificial Intelligence (AI) and Natural Language Processing (NLP) have significantly enhanced human-machine interaction, particularly in the education sector. This study presents the implementation of an AI-driven chatbot designed to assist learners on an elearning platform using deep learning techniques. The chatbot employs a feedforward neural network to classify user intent, trained on a structured dataset categorized across specific learning domains. Preprocessing methods such as tokenization, stemming, and bag-of-words representation are applied to refine the data for classification. The model, built with PyTorch and optimized using the softmax function, selects the most relevant response with high precision. Performance metrics, including precision, recall, and F1-score, indicate a robust and efficient chatbot, achieving an accuracy of 99.20% in managing student inquiries. The chatbot successfully produced relevant responses for 98.7% of test queries, demonstrating its adaptability across multiple academic subjects. Additionally, the system integrates the Gemini API to enhance response accuracy for unfamiliar queries, ensuring a more dynamic and interactive learning experience. The chatbot offers real-time user engagement, making it an effective tool for personalized and context-aware education assistance.