Text mining of practical disaster reports: Case study on Cascadia earthquake preparedness

Area of Science:

• Disaster Management
• Text Mining
• Natural Language Processing

Background:

• Practical disaster reports are crucial for learning from past events but are underutilized due to analysis challenges.
• Synthesizing and analyzing this vast literature is essential for improving future disaster mitigation and preparedness.

Purpose of the Study:

• To present a corpus of practical disaster reports for text mining.
• To introduce and validate an approach for extracting insights from these reports using text mining tools.
• To explore opportunities and challenges in text mining practical disaster reports.

Main Methods:

• Developed a corpus of practical disaster reports.
• Applied text mining tools to extract insights.
• Conducted a case study on U.S. Pacific Northwest earthquake preparedness.
• Performed a user survey on text mining tool utility.

Main Results:

• Identified differences in emergency management priorities between Washington and Oregon.
• Uncovered latent sentiments within disaster reports.
• Highlighted inconsistent vocabulary across the disaster management field.
• Survey indicated advanced AI tools like GPT offer more accessible insights than simpler methods.

Conclusions:

• Text mining offers a viable method for extracting valuable insights from practical disaster reports.
• Advanced AI tools can enhance accessibility of disaster-related data analysis.
• Addressing vocabulary inconsistencies and understanding sentiment are key for improved disaster communication and planning.