Text-Mining Analysis of Firefighters’ Evaluation Narratives in the Ambulance Activity Logs of the Yangsan Fire Station

This study employs text-mining techniques to analyze evaluation narratives from the emergency activity logs of the Yangsan Fire Station 119 Emergency Medical Service unit (Gyeongsangnam-do, South Korea) for 2018∼2023. To reflect the progression of coronavirus disease 2019 (COVID-19), the dataset was divided into three phases: pre-pandemic (2018∼2019), pandemic (2020∼2021), and post-pandemic recovery (2022∼2023). Data preprocessing was performed using MS Excel, IBM SPSS, R, and the RHINO morphological analyzer. Scheffé’s post-hoc test assessed differences in text volume across phases. Term frequency–inverse document frequency analysis, word-cloud visualization, and latent Dirichlet allocation topic modeling identified shifts in salient keywords and thematic patterns. The analysis showed a significant increase in narrative volume following the onset of COVID-19, with more frequent references to infection control, personal protective equipment, and advance hospital notification. Topic modeling demonstrated a transition from pandemic-era infection-control procedures to more detailed physiological descriptions in the endemic phase. These results underscore the evolving informational content of emergency medical documentation and its value as an empirical basis for improving record-keeping systems and documentation efficiency among emergency medical personnel responding to infectious diseases and other disasters.