Large Language Model and Retrieval-Augmented Generation Model for Indonesian Publication

Abstract

Garba Rujukan Digital (GARUDA) is a platform for publications and references in scientific articles, journals, and theses in Indonesia. However, to be able to find specific information in many articles and journals, of course, it is necessary to develop a system to make it easier to find this information. Therefore, a chatbot system with Large Language Model (LLM) and Retrieval Augmented Generation (RAG) was developed which is used to retrieve information through data-based chatbots on GARUDA. To find out the results of this study, a matrix evaluation was carried out using the ROUGE score with an average result of the value range from 42.68% to 68.03%. Thus, the evaluation showed that the output worked quite well in answering questions in scientific articles in the GARUDA Computer Science & IT indexed journal, especially on web-based subtopics.

Keywords: chatbot, RAG, LLM, GARUDA Kemdikbud