Total Accidental Information in Kyoto University and Comparative Analysis by using Natural Language Processing

Abstract

Since the incorporation of national universities in Japan, Kyoto University has prioritized safety management. To clarify the causes of accidents at Kyoto University since 2010, a comparative analysis is conducted by applying natural language processing. Seventy percent of reported accidents can be classified into five categories: “poked with a needle,” “fell down,” “traffic accident,” “exposure to body fluid,” and “cut or scratch.” The plot of the number of monthly accidents is bimodal with maximums in June and November. An analysis based on natural language processing reveals that “fell down” tends to occur when people move from a driveway to a sidewalk on a bicycle, but the typical analysis comparing categories does not reveal such a trend. Additionally, the category, “fell down” can be classified to “contact type” and “slip type.”