Advanced Chatbot Development for National Digital Transformation

Through this advanced chatbot, we provided a consultation support service to respond to inquiries and data requests regarding national digital projects and budgets from central ministries and public institutions.

kraj

Korea Południowa

poziom rządu

Krajowy

oś czasu

3–4 lata

koszt projektu

Brak danych

organizacja

NIA (National Information Society Agency)

skala wpływu

Pilotaż/mały zespół

kategorie

Przetwarzanie i administracja dokumentówAnaliza danych i prognozowanie

rodzaj technologii

ChatbotyLLM-y (np. Gemini, Claude, Microsoft Co-Pilot)

Overview

While guidelines, templates, and reference materials are currently distributed via official documents, there are limitations in service response due to a small workforce and an Excel-based data collection system. To overcome this, we built an automated and efficient monitoring and analysis service for national digital projects using Hyper-scale AI technology.

Project description

For AI model training, data was first extracted from 10 initial Q&A sets, internal budget documents, and laws. The development followed these procedures:

Preprocessing of training data (700 passages)
SmartChat-based query generation and verification
Training and indexing of preprocessed data into SearchGPT/SmartChat
Development of a demo page for functional testing

Key technological features

SearchGPT: Completed preprocessing of 700 data passages, followed by training and indexing; integrated with a demo page to provide search services.
SmartChat: Completed generation and verification of 1,800 queries; applied to the chatbot solution and integrated with a demo page for automated query generation.

Expected outcomes

As a result of the chatbot advancement, we reduced processing time for identifying accurate information and follow-up tasks through rapid and consistent responses. We successfully provided an intelligent search service capable of answering various inquiries based on internal data. Furthermore, we expect to enhance performance through future improvements, such as converting tables and numerical data into text or structured formats for additional training.

Launch year: 2023

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=29191