Chatbot assistant based on Large-Language Models for University students

• Published in: 2025 IEEE 29th International Conference on Intelligent Engineering Systems (INES) pp. 000077 - 000082
• Main Authors: Otsuka, Ami, Hajdu, Csaba, Kovacs, Katalin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 11.06.2025
• Subjects: chatbot; Chatbots; Cloud computing; High performance computing; human assistance; Hungarian langauge; large language models; on-premise; Regulation; Retrieval augmented generation; Scalability; Stress; User experience; Visualization; Web services
• DOI: 10.1109/INES67149.2025.11078205

Large-language models (LLMs) have recently gained significant traction in natural language processing (NLP) by accurately modeling and imitating human-like conversations. One standout application area involves chatbots, which leverage LLMs to provide context-aware, natural language interactions. However, existing solutions often target English and rely on external cloud-based platforms, raising concerns about data privacy and language coverage. In contrast, this paper presents a locally deployable, Hungarian-language chatbot developed to assist university students with education and examination regulations. The proposed system ensures in-house deployment, facilitating compliance with institutional data policies and offering cost-effective scalability. Beyond offering straightforward answers on deadlines and academic rules, our chatbot is designed to handle more nuanced student inquiries, enhancing user experience and administrative efficiency. Preliminary testing demonstrates robust performance in Hungarian context. Future plans include extending the chatbot's domain to more complex subjects, broader document sets, and additional institutions, as well as integrating high-performance computing resources for large-scale deployments.