KAIST and KBS Collaborate to Build Korean-Style News Video AI Database

[Asia Economy Reporter Kim Bong-su] The Korea Advanced Institute of Science and Technology (KAIST) announced on the 20th that it will collaborate with the Korean Broadcasting System (KBS) to build a Korean-style news video artificial intelligence (AI) database and metadata.

A consortium led by the KAIST Graduate School of Culture Technology research team, including Professors Park Ju-yong, Lee Won-jae, and Noh Jun-yong, has embarked on developing technology that uses AI to easily extract and edit video information. By combining the vast news video data owned by KBS, the executing agency, they plan to actively create an AI-based video content industry in the post-COVID era.

Currently, AI is rapidly expanding beyond areas that complement human mechanical movements and judgment, such as disease diagnosis and autonomous driving, into creative fields like media content production. However, AI technology that allows anyone, even non-experts, to easily create and edit content through speech still requires significant development. Experts agree that the lack of Korean language-based datasets is a major obstacle in South Korea.

KAIST, together with KBS, plans to build high-quality video data for AI model training and develop a ‘Korean-style news video metadata standard model.’ The public open news archive field based on metadata is currently led by the US’s AP and the UK’s BBC.

Professor Park Ju-yong stated, “Through this project, researchers and related startups who have struggled with copyright issues and the lack of universal metadata will find relief,” adding, “Furthermore, it will lay the foundation for a Korean-style content platform.”

The project is funded with a total budget of 4.2 billion KRW, including the 2021 AI training data construction project budget from the Ministry of Science and ICT and the Korea Intelligent Information Society Agency, as well as private investments from the consortium led by KAIST.

The KAIST Graduate School of Culture Technology is leading the project, with KBS, Metrics Research, Action Power, Sori Java, Data Maker, and Inter Minds participating as joint research institutions. All research outcomes developed in this project, including databases, AI training models, and programming codes, are public goods and can be used by anyone for research and business purposes.

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