Identifying key topics and network structures: A case in Korean AI value chain

Abstract

This study provides results of key topics and actor relationships in the Korea AI value chain using news articles. We build a dataset of 5200 news articles (2021–2024). Using LDA topic modeling, we identify 14 key topics and track yearly changes in the number of news articles in four value chain domains: the Hardware, Data, Model, and Solution domains. The results show a surge in the Hardware domain, a majority of articles in the Solution domain, gradual growth in the Model domain, and a small, stagnant Data domain. We then conduct a network analysis of actors within each domain. The Data domain has a policy driven structure with a relatively high share of government actors. The Hardware, Model, and Solution domains exhibit market driven structures dominated by big-tech firms. University and research institutes play a peripheral role across domains. Connections between domains also change over time. In the early years of the period, the strongest connection is between Solution-Data. By 2024, the strongest connection is between Solution-Hardware. Taken together, these findings suggest that the Korean AI value chain is dominated by large firms that mainly practice closed open innovation. The ecosystem remains in a transformation stage and shows weak integration in and around the Data and Model domains and a limited role for academic actors. Based on this, we argue that Korean AI policy should promote market open innovation through SMEs and start-ups, social open innovation, and closer links across domains and between universities, research institutes, and industry.