Development of an Intelligent Smart Campus Chatbot Based on Natural Language Processing and Retrieval-Augmented Generation for Integrated Academic Information Services

Abstract

The increasing volume and complexity of academic information in higher education institutions have created significant challenges for conventional information services. Students and prospective students often experience delays, inconsistent responses, and limited service availability when relying on manual administrative channels. To address this issue, this study develops and evaluates an Intelligent Smart Campus Chatbot that integrates Natural Language Processing (NLP) with Retrieval-Augmented Generation (RAG) to deliver contextually grounded academic information.

The proposed system employs LangChain as the orchestration framework, FAISS as the vector database, OpenAI text-embedding-ada-002 for document representation, and GPT-3.5-turbo or Groq LLaMA-3 as the response generation model. Institutional documents covering study programmes, tuition fees, scholarships, academic schedules, and admission procedures were used to construct the knowledge base. The chatbot was evaluated using 200 academic queries across five categories. The results indicate an overall accuracy of 91% with a macro-averaged Precision of 0.914, Recall of 0.908, and F1-score of 0.911. The system achieved an average response time of 3.2 seconds and a hallucination rate of 4.3%. User Acceptance Testing involving 50 respondents showed an overall satisfaction rate of 88%. These findings demonstrate that RAG-based retrieval significantly improves response accuracy and reliability compared to rule-based approaches, supporting the implementation of intelligent academic services within a Smart Campus environment.

GS Cited Analysis