Time-Series Analysis of Media Discourse on Nutrition Teachers (2016-2024) : BigKinds-Based Text Mining and Word Co-Occurrence Network Analysis

Nutrition teachers are essential to school meal operations, nutrition education, and safety management, but the evolution of public discourse on their responsibilities has not been tracked systematically. This study mapped the structure and temporal shifts of media discourse on nutrition teachers, identified crisis-driven reframings, and derived policy implications for role design and resource allocation. Using BigKinds (Korea Press Foundation), this syudy analyzed Korean-language news on “nutrition teachers” from January 2016 to March 2024 with text-mining methods—term frequency (TF) and word co-occurrence network analysis. After preprocessing (tokenization with KoNLPy-Okt, stopword removal, normalization), 3,674 articles were retained. Across the period, a stable school meals-school-teacher axis dominated. During COVID-19 (2020-2021), agricultural products, support, and safety surged, reflecting supply-chain responses (e.g., produce packages) and operational adaptations beyond routine service. In 2022-2023, the industrial-safety framing strengthened, with safety, health, and inspection rising alongside the coverage of kitchen work-environment risks. Early 2024 (Jan-Mar) included incident-linked terms (e.g., death), warranting cautious interpretation. The network results were consistent. The strongest tie was school-school meals, followed by school-teacher and nutrition-teacher, indicating concurrent expectations for education, nutrition guidance, and operational safety within the school-meals ecosystem. Policy and practice implications included the following: staffing expansion and administrative-load reduction; strengthened occupational health and safety (ventilation, exposure control, periodic health checks, and mental-health support); codified crisis protocols with standing inter-organizational support; and digital safety management (monitoring/record automation and risk prediction) with continuous professional development. The limitations of this study include media-reporting bias, a partial year for 2024, the non-causal nature of co-occurrence metrics, and sensitivity to preprocessing/normalization. Nevertheless, this study offers a data-driven map of discourse evolution to inform role design, resource allocation, and risk communication.