Open AccessComputer-Assisted Databasing of Disaster Management Information Through Natural Language Processing
Author(s) -
Kentaro Inui,
Yuji Watanabe,
Kenshi Yamaguchi,
Shingo Suzuki,
Hiroko Koumoto,
Naoko Kosaka,
Akira Koyama,
Tomohiro Kokogawa,
Yuji Maeda
Publication year - 2015
Publication title -
journal of disaster research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.332
H-Index - 18
eISSN - 1883-8030
pISSN - 1881-2473
DOI - 10.20965/jdr.2015.p0830
Subject(s) - computer science , schema (genetic algorithms) , information system , emergency management , natural disaster , management information systems , information management , natural language , database , computer security , artificial intelligence , engineering , information retrieval , geography , meteorology , law , political science , electrical engineering
During times of disaster, local government departments and divisions need to communicate a broad range of information for disaster management to share the understating of the changing situation. This paper addresses the issues of how to effectively use a computer database system to communicate disaster management information and how to apply natural language processing technology to reduce the human labor for databasing a vast amount of information. The database schema was designed based on analyzing a collection of real-life disaster management information and the specifications of existing standardized systems. Our data analysis reveals that our database schema sufficiently covers the information exchanged in a local government during the Great East Earthquake. Our prototype system is designed so as to allow local governments to introduce it at a low cost: (i) the system’s user interface facilitates the operations for databasing given information, (ii) the system can be easily customized to each local municipality by simply replacing the dictionary and the sample data for training the system, and (iii) the system can be automatically adapted to each local municipality or each disaster incident through its capability of automatic learning from the user’s corrections to the system’s language processing outputs.