RANCANG BANGUN QUESTION ANSWERING SYSTEM LAYANAN INFORMASI AKADEMIK BERBASIS RAG

Muhammad Surya Adhi Setiawan

Prosiding Seminar Nasional Indonesia
Vol. 3 No. 3 (2026): Prosiding Seminar Nasional Indonesia

The complexity of academic information in the New Student Admissions (PPMB) process often overwhelms conventional helpdesk services. This study aims to design and build a prototype Question Answering (QA) System based on Retrieval-Augmented Generation (RAG) that can automate information services accurately. The system is built using a three-layer architecture: Presentation Layer (Gradio UI), Application Layer (Python/LangChain), and Data Layer (ChromaDB). A key focus of the development is the data pipeline strategy, specifically handling "Indivisible Information Units" in PDF tables by setting a dynamic chunking limit of 3000 tokens. The prototype features a Knowledge Base Manager for dynamic document updates and a multilingual Chat Interface. Testing demonstrates the system's ability to process heterogeneous data from 30 sources and successfully retrieve specific procedural information, such as the "Golden Ticket" requirements, with high precision. The system is deployed using a reasoning model engine to ensure logical answer synthesis.