Hostname: page-component-7b7567c8c-z4zkx Total loading time: 0 Render date: 2026-11-17T19:54:35.930Z Has data issue: false hasContentIssue false

Construction of a System for Retrieving Notification Documents Issued When There Is a Disaster Using Multimodal AI

Published online by Cambridge University Press:  23 March 2026

Yoshino Sugihara
Affiliation:
Emergency and Disaster Medical Pharmacy, Faculty of Pharmaceutical Sciences, Fukuoka University, Japan
Rina Hirai
Affiliation:
Emergency and Disaster Medical Pharmacy, Faculty of Pharmaceutical Sciences, Fukuoka University, Japan
Soichiro Ushio
Affiliation:
Emergency and Disaster Medical Pharmacy, Faculty of Pharmaceutical Sciences, Fukuoka University, Japan
Megumi Katsuta
Affiliation:
Emergency and Disaster Medical Pharmacy, Faculty of Pharmaceutical Sciences, Fukuoka University, JapanNippon Medical School CHIBAHOKUSOU Hospital, Japan
Toshinobu Hayashi
Affiliation:
Emergency and Disaster Medical Pharmacy, Faculty of Pharmaceutical Sciences, Fukuoka University, Japan
Takashi Egawa
Affiliation:
Emergency and Disaster Medical Pharmacy, Faculty of Pharmaceutical Sciences, Fukuoka University, Japan
Rights & Permissions [Opens in a new window]

Abstract

Core share and HTML view are not available for this content. However, as you have access to this content, a full PDF is available via the 'Save PDF' action button.
Introduction:

Information management for large-scale disasters is at the heart of logistics. When there is a disaster, the MHLW issues a notification document regarding the measures to be taken by the MHLW and the relevant local governments, depending on the damage situation. However, it takes time to summarize and understand the notification documents. Therefore, we developed a system using Gemini to search for notification documents issued in the event of a disaster.

Methods:

Gemini is a multimodal AI that can make advanced decisions by combining various types of information and can quickly summarize and search information from Google Drive. Notification documents were stored in Google Drive, and questions were created from keywords such as “dispatch of medical professionals” and “handling of insurance treatment,” and the search results were verified.

Results:

As a result of searching for the question, “I want to know about the dispatch of medical personnel when there is a disaster,” three materials were matched. In addition, the search results included a summary of the materials and links to all related materials stored in Google Drive.

Conclusion:

The advantages of the notification document retrieval system are as follows: it is possible to check the notifications that are common to multiple disasters at once, anyone can use it free of charge by sharing an account, and it is possible to summarize the contents of the materials as well as extract related materials. Conversely, as an improvement, it was mentioned that since it is not an exact match, documents that are not necessary are extracted, and only a part of the required documents is extracted. Although it is necessary to devise ways to share accounts and update data, a notification document retrieval system using multimodal AI can be expected to manage information in the event of a disaster.

Information

Type
Meeting Abstracts
Copyright
© The Author(s), 2026. Published by Cambridge University Press on behalf of World Association for Disaster and Emergency Medicine