KISTI launches AI-based curation system for science and technology data

The science and technology information curation system [Courtesy of KISTI]

The Korea Institute of Science and Technology Information (KISTI) announced on Thursday that it officially launched an information curation system for science and technology that employs artificial intelligence (AI) models a day earlier.

The use of AI models saves the cost of metadata and reference building by 14.5 percent compared to manual work. It also reduces the construction time by around 3,592 hours annually.

A total of five AI models have been applied to the science and technology information curation system—automated extraction of theses and dissertations metadata, automated separation of reference attributes, automated extraction of tables and figures, automated classification of document topics, and automated object identification. KISTI has been developing AI models since 2018 to automate the construction of databases on science and technology information. It introduced the first science and technology information curation system using its AI models in 2022, which had an average accuracy score of 99 percent in the test conducted by the Telecommunications Technology Association (TTA).

“By introducing AI models, we will respond quickly and efficiently to the annually increasing task of building science and technology data,” Lee Hye-jin, head of KISTI’s Digital Curation Center, said.