A Computational Analysis of Korean News Discourse on Generative AI Sora

초록

Recent advances in generative artificial intelligence have intensified public debate on creativity and cultural production, yet little is known about how such technologies are discursively framed in Korean media, particularly in relation to film and creative labor. Existing studies on AI discourse have largely focused on qualitative ethical discussions, leaving a gap in large-scale, empirical analyses of Korean media discourse. This study examines how Korean news media represent OpenAI’s text-to-video model Sora in relation to film, with particular attention to the discursive asymmetry between technological innovation and creative anxiety. Using computational text analysis, this study analyzes 240 Korean news articles published between January 2024 and July 2025, collected from Big Kinds, the news database of the Korea Press Foundation. The analysis employs keyword frequency analysis, time-series trend analysis, and co-occurrence-based semantic network analysis. Centrality measures are used to identify dominant and peripheral actors within the discourse. The results show that technology-related keywords such as AI, artificial intelligence, and OpenAI occupy the highest centrality positions. In contrast, terms related to creators, artists, and film production remain structurally peripheral. Time-series analysis reveals a sharp peak in media attention immediately following Sora's announcement in February 2024, followed by a decline and subsequent smaller peaks associated with practical applications such as AI film festivals. These findings demonstrate that Korean news discourse prioritizes technological innovation driven by global technology firms while marginalizing concerns about creative labor. By combining computational methods with media and film discourse analysis, this study contributes to discourse studies and film studies by empirically mapping the power structure embedded in media representations of generative AI.

키워드