A Retrieval-Augmented Generation-based Chatbot using Llama 3
Humaira, Yori Adi Atma, Rika Idmayanti, Alvin Fadli Dwi Mulya

Department of Information Technology, Politeknik Negeri Padang, Jl. Kampus Limau Manis, Padang 25164

Abstract

Artificial intelligence is widely used in a variety of sectors, including chatbot technology. This study creates a chatbot utilizing the Natural Language Processing (NLP) technique. This method enables the system to work more intelligently in understanding the user^s language and to respond in a natural language. This chatbot examines a case study of Question/Answer (Q/A) information from the Information Technology (IT) Department at Padang State Polytechnic (PNP). This chatbot employs Large Language Model (LLM) technology, notably Llama 3 and Retrieval-Augmented Generation (RAG). RAG serves to improve the performance of Llama 3. The use of RAG in this situation enables the chatbot^s knowledge to be dynamically updated. This research resulted in the successful development of an interactive chatbot that uses natural language. The natural language spoken is Indonesian. The system testing was conducted using the researchers^ intuitive approach. Users provided questions to the chatbot system, and the chatbot responded by giving answers. These responses were successfully aligned with the context of the available documents.

Keywords: RAG, LLM, Llama 3, Q/A, NLP, Chatbot, Information

Topic: Artificial Intelligence (AI)