PoC for Educational Administration Response via Hyper-scale AI

Overview

To address the rapid surge in educational activity protection consultations* and the heavy burden of school administration, the Gyeonggi Provincial Office of Education (GOE) aims to provide an environment where faculty can focus on teaching by utilizing Hyper-scale AI for work support.
*Consultation Volume: (2021) 1,424 → (2022) 4,393 → (2023) 9,761 cases.

While many administrative tasks have been digitized, faculty members still struggle with repetitive tasks centered on searching through manuals and regulations. This pilot project focuses on School Student Records (SSR)—a task that is consistently time-consuming at the beginning and end of semesters—applying AI to streamline the process among 410 administrative tasks.

Project description

The project selected six core document types, including the "School Student Record Entry Manual" for elementary, middle, and high schools and the "Academic Achievement Management Guidelines." To ensure high accuracy, over 2,000 Q&A pairs were collected and pre-processed. A Vector Database was constructed from this refined data, and the RAG (Retrieval-Augmented Generation) system was optimized through continuous validation of answers and meticulous adjustment of LLM parameters.

The service allows teachers to inquire about SSR entries through an interactive AI interface. When a teacher enters a question, the AI provides a natural language response based on the pre-trained manuals and Q&A datasets. To ensure credibility, the system lists the specific reference materials (guidelines, manuals) and presents the corresponding content, allowing teachers to verify the information instantly.

Key technological features

• Automated Data Pipeline: Utilization of a Hyper-scale AI platform’s automated pipeline for document parsing, chunking, pre-processing, and RAG optimization
• Accuracy Optimization & Viewer Integration: Enhancing response accuracy through LLM parameter tuning and integrating a document viewer to provide the exact source pages within the original documents

Expected outcomes

By shifting from keyword-based searches to context-driven natural language retrieval, the system delivers responses with over 97% accuracy. Previously, obtaining answers through the School Student Records Support Portal could take several days; now, the Generative AI service provides the necessary information instantly. This creates a robust support system for faster and more accurate record-keeping.

When applied to the broader spectrum of educational administration, this service will significantly reduce the workload of school faculty, allowing them to dedicate more time to educational activities. Looking ahead, the GOE plans to establish the "Gyeonggi Education Digital Platform"—incorporating AI and Big Data—to innovate administrative workflows and enhance communication within the educational community.

Launch year: 2024

Attribution

Case study courtesy of the National Information Society Agency, Republic of Korea, Apolitical's Content Partner. Sourced from Use Cases of Public AI Service, Vol.2, https://eng.nia.or.kr/site/nia_eng/ex/bbs/View.do?cbIdx=31975&bcIdx=29182&parentSeq=29233