Development of a System for Structuring Disaster Management Workflow Using Generative AI
Journal of Information Processing · 2026 · 34 · 322-337
DOI: 10.2197/ipsjjip.34.322 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Information Processing, 2026. Verfasst von Koki Asami und 1 weiteren Autor:innen. Die Arbeit nennt 15 Literaturangaben.
Abstract (Originalsprache)
This paper presents a generative-AI pipeline that converts local disaster-management plans into interactive, explainable workflows. Claude 3.7 Sonnet extracts tasks, responsible departments, and dependencies; GPT-4.5 renders the result, while a chatbot-based human-in-the-loop interface lets officials revise it in natural language. In a study with fifteen municipal officers, the system extracted 177 tasks and 180 dependencies from a prefectural earthquake plan—9.4× and 18.4× the manual baseline (18.9 and 9.8). Per-task effort fell from 95s manually to 8.9s, and participants rated coverage 4.07±0.83 and chatbot usefulness 4.13±0.83 on a five-point scale. The simple and versatile output format supports plan revision, inter-municipal benchmarking, and linkage to simulation and scenario-generation tools, offering a scalable, transparent foundation for evidence-based disaster governance. Source code is available at https://github.com/koki-asami/disaster-management-workflow-extraction-system.
Zitieren
APA
Asami, K., & Hatayama, M. (2026). Development of a System for Structuring Disaster Management Workflow Using Generative AI. Journal of Information Processing, 34(0), 322-337. https://doi.org/10.2197/ipsjjip.34.322
Vancouver
Asami K, Hatayama M. Development of a System for Structuring Disaster Management Workflow Using Generative AI. Journal of Information Processing. 2026;34(0):322-337. doi:10.2197/ipsjjip.34.322
BibTeX
@article{Asami2026,
author = {Asami, Koki and Hatayama, Michinori},
title = {Development of a System for Structuring Disaster Management Workflow Using Generative AI},
journal = {Journal of Information Processing},
year = {2026},
volume = {34},
number = {0},
pages = {322--337},
doi = {10.2197/ipsjjip.34.322}
}
EUVIMED