Institutional Research Data Utilization for Policy Support

Overview

The Korea Federation of SMEs (KBIZ) sought to introduce AI technology to discover policy indicators through internal data searches and to provide similar reference materials when responding to requests from institutions and members. Accordingly, this project aimed to verify the feasibility of a SearchGPT/SmartChat-based service to support policy-related tasks using key internal documents such as reports, news articles, and legal data.

Project description

To train this service model, 859 queries were generated using a dataset consisting of 1,963 documents, 10,000 news articles, and 1,963 statutes. The model training procedure began with the pre-processing of internal data, followed by the generation and review of queries based on SmartChat and the indexing of the processed data. Subsequently, a demo page was developed to verify the integration of the trained model with the service interface. This system is designed as a search and chatbot service utilizing major internal document data, specifically aimed at supporting staff members during the policy proposal process.

Key technological features

• SearchGPT/SmartChat-Based Service: Learning and indexing pre-processed data within SearchGPT and generating queries based on SmartChat
• Hybrid Integration: A service approach for the chat interface that combines the respective strengths of search engines and chatbot solutions through a hybrid integration method

Expected outcomes

This service supports the discovery of policy indicators for the Korea Federation of SMEs, the proposal of new policies, and the drafting of press releases by learning from key research materials. By converting existing statistical information from table formats into structured text, the system allows staff to easily secure various quantitative data points through simple searches. Furthermore, instead of reviewing the entire content of reference materials like research reports, staff can engage in Q&A with the chatbot to find specific evidence for policy formulation. This is expected to significantly reduce the time required for data processing and improve overall administrative efficiency.

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=29222