Q&A Support Chatbot for Broadcast Advertising Documents

Korea Broadcast Advertising Corporation (KOBACO) developed a RAG-based Q&A service targeting relevant documents and integrated it with a chatbot system to enhance operational efficiency.

Land

Südkorea

Regierungsebene

National

Zeitleiste

1–2 Jahre

Projektkosten

TBD

Organisation

Korea Broadcast Advertising Corporation (KOBACO)

Umfang der Wirkung

Pilot/Kleines Team

Kategorien

Servicebereitstellung und BürgererlebnisDokumentenverarbeitung und Verwaltung

Technologieart

ChatbotsLLMs (z. B. Gemini, Claude, Microsoft Co-Pilot)

Overview

The project was initiated to improve the civil complaint handling process, as staff members faced difficulties in providing rapid responses due to repetitive and time-consuming manual searches. Korea Broadcast Advertising Corporation (KOBACO) developed a RAG-based Q&A service targeting relevant documents and integrated it with a chatbot system to enhance operational efficiency.

Project description

To support the broadcast advertising chatbot, relevant documents were restructured into search-optimized formats. Generative AI was used to automatically generate chatbot queries, leading to the development of a specialized chatbot service that combines RAG technology with KOBACO’s advertising data.

The service, which supports staff in responding efficiently to complaints, is classified into two types:

Chatbot-based responses: Provides answers based on pre-trained knowledge from existing documents (internal data, laws, precedents, etc.).
RAG-based SearchGPT: For queries outside the training set, it provides highly relevant content using RAG-powered search capabilities.

Key technological features

Data Collection: Collected internal data from KOBACO (Korea Broadcasting Advertising Corp.), along with external laws and legal precedents via APIs and public web pages
OCR Implementation: Applied Optical Character Recognition (OCR) models to process scanned documents and image data, enabling text extraction from existing archives.
RAG Utilization: Employed Document Layout Analysis (DLA) models to extract content and structure internal documents for efficient RAG (Retrieval-Augmented Generation) learning.

Expected outcomes

By replacing repetitive and exhaustive tasks with AI, the service is expected to create a work environment where employees can focus on more professional and high-value tasks.

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=29190