'Core of Data Dam' Development of AI Data Quality Standard Proposal

[Asia Economy Reporter Seulgina Jo] The Ministry of Science and ICT announced on the 5th that it has developed a standard draft establishing the concept, scope, and detailed requirements of quality related to artificial intelligence (AI) data, a core resource of the digital New Deal's 'Data Dam.'

On the 6th, the related standards will be officially proposed to the group standardization organization within the Korea Information and Communication Technology Association (TTA PG 1005, AI-based Technology), followed by a process of collecting opinions from related experts. In this case, the standard draft is expected to be finally adopted and confirmed by June next year.

The standard draft defines the quality of AI data as "the level at which data used in AI technology (models and algorithms) secures diversity, accuracy, validity, etc., to provide useful value to users."

It was developed as a universal standard applicable commonly across various fields such as natural language processing, autonomous vehicles, medical care, agriculture, fisheries, and manufacturing, and includes basic data specifications by defining necessary standard procedures and quality requirements at each stage of the entire data cycle?collection, refinement, processing, quality verification, and utilization?to systematically secure data quality and enhance interoperability.

Specifically, it consists of ▲ quality requirements such as diversity and factuality at the raw data collection stage and technical suitability requirements such as file format and resolution, ▲ requirements for preventing data duplication and de-identification measures at the refinement stage, ▲ requirements for object classification systems and labeling specifications at the processing stage, and ▲ verification requirements and methods such as validity at the quality inspection and utilization stages.

The Ministry of Science and ICT is already applying the developed standard draft to supplementary budget projects (10 major fields, 150 types of data). After collecting opinions from experts in industry, academia, and research, it will promptly promote group standardization and plans to propose related matters to the ‘Artificial Intelligence International Standardization Meeting (ISO/IEC JTC1/SC42)’ through institutions such as the National Radio Research Agency, aiming to lead to achievements in international standardization.

Furthermore, by the end of this year, based on the standard draft, the ministry plans to develop and distribute an ‘AI Data Quality Management Guide’ including detailed explanations to enable AI developers and public institution officials to plan and manage AI data quality more systematically.

This content was produced with the assistance of AI translation services.