Indonesia AI-Based Chatbot Development for Academic Information Services at Universitas Negeri Medan and an Analysis of Indonesian Language Usage

Muhammad Rafli Wijaya, Alya Namira, Ananda Syafika, Syairal Fahmy DalimuntheNon-Scopus15 Juni 2026Afiliasi: ID

Abstrak

The efficiency of academic information services remains a major challenge in higher education institutions, particularly in responding to student inquiries quickly and independently. This study develops an artificial intelligence-based chatbot system for academic information services at Universitas Negeri Medan (UNIMED) by utilizing the LLaMA 3.1 8B Instant Large Language Model (LLM) via the Groq API within a Retrieval-Augmented Generation (RAG) framework. The system was built using a three-tier architecture consisting of a React.js (Vite) frontend, a Node.js with Express backend, and a Supabase (PostgreSQL) database serving as the academic FAQ knowledge base. The Cross-Industry Standard Process for Data Mining (CRISP-DM) was adopted as the research methodology. System evaluation was conducted using Black-Box Testing across four main scenarios: questions available in the FAQ, follow-up questions requiring conversational context, off-topic questions, and reference link validation all of which yielded a pass status. Furthermore, Indonesian language testing demonstrated that the system is capable of understanding diverse student language variations, including formal language, informal expressions, academic abbreviations, and ambiguous queries, while maintaining appropriate academic communication etiquette. The results indicate that the RAG approach is effective in reducing AI hallucination risks, and that this web-based chatbot offers broader accessibility compared to previous messaging platform-based chatbot systems.

Identifier

Ekspor BibTeX
Scroll to Top