DefinePK

Abstract

Universities often struggle to efficiently handle student queries related to admissions, academics, and administration through traditional methods like emails, calls, or help desks. This research introduces an AI-powered chatbot designed to provide instant and accurate responses to university students using Natural Language Processing (NLP) techniques. The system combines Semantic BERT (SBERT), TF-IDF, and Cosine Similarity to enhance contextual understanding. A custom dataset of over 500 student queries was compiled through online and in-person surveys, with answers verified by university departments. The backend, developed in Python, processes queries and retrieves relevant responses, while the Streamlit-based frontend allows real-time interaction. The chatbot was deployed in a local environment and tested through unit, integration, and system-level testing. It demonstrated high accuracy, speed, and user satisfaction. Comparative results show SBERT outperformed the T5 model in terms of relevance and understanding. To efficiently handle students' queries, chatbots are a suitable consideration for future improvements to support multi-turn dialogue, multiple languages, real-time data updates, and cloud deployment for wider access. .
Keywords : Chatbot, Natural Language Processing (NLP), Semantic BERT (SBERT), TF-IDF