Demonstration of Public Inquiries for Workers' Compensation Insurance Decisions

Overview

Workers' compensation administrators repeatedly search relevant materials to process over 160,000 annual workers' compensation cases. To address processing delays caused by repetitive searches and the need to learn case knowledge, the introduction of AI was considered.
Accordingly, the Korea Workers' Compensation & Welfare Service aimed to enhance service reliability by training AI on workers' compensation insurance decision judgments, review decisions, disease determination certificates, etc. This would streamline workers' compensation insurance administration and improve public service response by answering inquiries from injured workers.

Project description

The project was conducted to implement a business chatbot for intelligent document search and analysis, and a public-facing chatbot for industrial accident processing, by unifying internal and external data.

Relevant laws, ordinances, regulations, and other materials, along with internal data from the Korea Workers' Compensation & Welfare Service (case precedents, review decisions, disease determination results, etc.), were utilized for training the service model. The model training procedure involved: 1. parsing and refining document data, 2. vectorization via chunking followed by an embedding model, 3. storing the vectors in a database, and providing responses through a Retrieval-Augmented Generation (RAG) system.

Key service features include providing AI-based customer support services to assist workers with administrative decisions regarding work-related injuries. Additionally, workers' compensation case handlers can easily search for recognized work-related injury cases.

Key technological features

• API Interoperability Specification Development Based on Pulmonary Function Test Guidelines Using Merge Engine
• Generation of training data via Retriever, GPT, etc., and customer service chatbots and search services

Expected outcomes

By utilizing GPT to provide chatbot and search services for customer support of workers' compensation claimants, case handlers can establish investigation directions for work-related injury claims filed by workers based on existing precedents, review decisions, and disease determination information, and use this as a basis for judgment. Furthermore, by providing a service allowing users, such as injured workers, to search for recognized work-related injury cases, the reliability of injury recognition outcomes can be enhanced.

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.1,  https://eng.nia.or.kr/site/nia_eng/ex/bbs/View.do?cbIdx=31975&bcIdx=28997&parentSeq=29020