2023 Volume 21 Issue 1 Pages 59-67
Learning critical situations during flood emergency response is useful for learning lessons from past disasters and anticipating possible difficulties before disasters. Based on after-action review reports published by local governments, “The Collection of Critical Situations during Flood Emergency Response for Local Governments” catalogs the past examples of critical situations in which local government officers had a hard time making sensible decisions because they panicked, did not know what to do, were confused or in a dilemma, etc., during an emergency response. Although it needs updating periodically, the challenge is how we can efficiently collect new examples of critical situations and reflect them in the collection, as the number of review reports is rapidly growing because of frequent floods in recent years. Today’s innovative text-mining technologies may provide solutions to this challenge. In this study, we analyzed the trends of examples in recent after-action reports and identified points that should be considered when deep learning technologies are applied to the automatic extraction of examples from the reports by text mining. When examples are extracted with the typical keywords in the sentences, high risk of duplication of extraction is anticipated. At the application of deep learning technologies, this consideration is recommended.
