Korea Data Agency
AI Data Processing for the Data Voucher Programme
As a designated AI-processing supplier under the data voucher programme, we prepared data held by client companies so it could be used for AI training.
Nov 2022 – ongoing
Project
- Nov 2022 – present — designated AI-processing supplier for the data voucher programme.
- Jun–Nov 2023 — delivered two AI-processing projects.
- Jun–Nov 2024 — delivered one AI-processing project.
Challenge
- Raw data held by a client cannot be used for training as it is — it has to be collected, cleaned and processed first.
- Every source is laid out differently, so batch processing alone does not classify it correctly; some judgement has to stay with a person.
- When several people work on it, any drift in the criteria shows up directly as lower training data quality.
Development
Data preparation
- Prepared client data into a form usable for AI training.
- Defined the processing rules per source type first, then classified and tagged against those rules.
- Assembled and reviewed deliverables to the requirements of the government programme.
Projects delivered in 2023
- Kodes — an online AI image generation system for visual journalism.
- Geospatial Information Technology — an AI crack detection model for airfield runways, from drone footage.
- Generative AI used to produce the processed data each field needed.
How the work was run
- Wrote a processing standard so that different workers made the same judgement calls.
- Recruited and trained local crowd workers, many returning after a career break.
- Sampled the output for review and sent anything off-standard back for rework.
Screens
These are the live product screens. The interface is in Korean.
Technology
Related links
Other projects
A&A
2023 – ongoing
Smart RPC Real-time Grain Inventory System
Korea Press Foundation
2021–2025
KPF-BERT and AI-based News Services
Codnut
2023–2025