GRAPH RAG UNTUK MEMAHAMI PERATURAN TENTANG PAJAK KENDARAAN BERMOTOR DI PROVINSI BANTEN [GRAPH RAG TO UNDERSTAND THE REGULATION ON MOTOR VEHICLE TAX IN BANTEN PROVINCE]

Prihartono, Agung
Priantoro, Akhmad Unggul

Al-Ihtiram: Multidisciplinary Journal of Counseling and Social Research
Vol 4, No 1 (2025): Mei

Local regulation documents (Perda) often have a complex structure and intricate language, making them difficult for the general public to understand. Large Language Models (LLM) offer the potential to simplify this information but carry the risk of generating inaccurate or "hallucinated" information. This research proposes and tests the Graph Retrieval-Augmented Generation (Graph-RAG) method as a solution to build an accurate and explainable question-answering system. Using the Provincial Regulation of Banten Number 1 of 2024 on Regional Taxes and Levies as a case study, we built a knowledge graph mapping key entities related to Motor Vehicle Tax (PKB)—such as rates, taxpayers, and exemptions—along with their relationships. The results of the trial show that the Graph-RAG system can answer specific questions about the base rate of PKB, progressive rates, and excluded objects with 100% accuracy, while also providing direct citations to the relevant articles and paragraphs in the regulation. This method has proven effective in presenting complex regulatory information reliably and verifiably, demonstrating its potential as a tool to assist public administration and citizen information services.