1:45 PM - 3:15 PM
[O11-P112] Disaster Prevention Map Creation Using Free Digital Tools on Block Walls Using the Ichikawa Gakuen neighborhood in Ichikawa City
Keywords:Disaster Prevention Map, Google Street View, Block Walls, Generation AI
1. Background and Purpose
Hazard maps help identify dangerous areas during disasters but may not address students' daily safety needs, such as commuting routes. This study surveyed students at Ichikawa Gakuen to determine how hazard maps could be improved. The results highlighted a key issue: difficulty in quantitatively analyzing free-response surveys due to human bias during tabulation. Students commonly requested disaster-related information on block walls. However, previous studies raised concerns about the reliability of traditional survey methods. This study thus aimed to improve questionnaire analysis and block wall surveys using free digital tools.
2. Research Methods
2.1 Enhancing Questionnaire Analysis
Students answered a Google Form survey using a hazard map of the school area, suggesting improvements. To quantitatively process responses, text mining was applied. Traditional morphological analysis is labor-intensive and can reflect analyst bias (Koshinaka, 2012). Therefore, this study employed four generative AIs to reduce subjectivity and compare analytical consistency. Each AI performed morphological analysis and grouping, followed by co-occurrence network visualization using KH Coder to identify map improvement themes.
2.2 Investigating Block Wall Distribution with Street View
Based on student feedback, the focus shifted to block walls. Traditional on-site surveys (e.g., Kawakami, 1999) are time-consuming. Instead, a 7.5 km² area around the school was surveyed using Google Street View, identifying block walls with four or more horizontal and six or more vertical layers requiring reinforcement.
3. Results
645 valid questionnaire responses were collected. AI-generated grouping lists and co-occurrence network diagrams (Figures 3 and 4) showed frequent mentions of block wall-related safety concerns. Although full analysis by other AIs (Gemini, Deepseek, Copilot) and the author is not shown, results confirmed the lack of block wall disaster-prevention information.
Street View confirmed the locations of many block walls (Table 1), though some features like rebar presence were indiscernible. These locations were plotted on a blank map (Figure 5), and safe evacuation routes were derived and marked in green.
4. Discussion
4.1 Consistency Across AI Analyses
Table 2 summarizes common themes found by each AI. All emphasized the need for clearer evacuation routes, danger zone markers, and barrier information. These findings were consistent across all AIs, demonstrating a certain reliability in AI-assisted text analysis.
4.2 Validity of Street View Surveys
Table 3 compares block wall counts from actual walking surveys with those obtained from Street View along the green evacuation route. Differences were minimal (fewer than five), and all indicated a decrease in wall numbers. Thus, Street View was deemed effective for identifying potential hazards.
5. Conclusion
This study successfully used AI to analyze student feedback on hazard maps and validated the use of Street View for block wall surveys. The methods showed reliability, offering a cost-effective and replicable approach to community disaster prevention research.
Hazard maps help identify dangerous areas during disasters but may not address students' daily safety needs, such as commuting routes. This study surveyed students at Ichikawa Gakuen to determine how hazard maps could be improved. The results highlighted a key issue: difficulty in quantitatively analyzing free-response surveys due to human bias during tabulation. Students commonly requested disaster-related information on block walls. However, previous studies raised concerns about the reliability of traditional survey methods. This study thus aimed to improve questionnaire analysis and block wall surveys using free digital tools.
2. Research Methods
2.1 Enhancing Questionnaire Analysis
Students answered a Google Form survey using a hazard map of the school area, suggesting improvements. To quantitatively process responses, text mining was applied. Traditional morphological analysis is labor-intensive and can reflect analyst bias (Koshinaka, 2012). Therefore, this study employed four generative AIs to reduce subjectivity and compare analytical consistency. Each AI performed morphological analysis and grouping, followed by co-occurrence network visualization using KH Coder to identify map improvement themes.
2.2 Investigating Block Wall Distribution with Street View
Based on student feedback, the focus shifted to block walls. Traditional on-site surveys (e.g., Kawakami, 1999) are time-consuming. Instead, a 7.5 km² area around the school was surveyed using Google Street View, identifying block walls with four or more horizontal and six or more vertical layers requiring reinforcement.
3. Results
645 valid questionnaire responses were collected. AI-generated grouping lists and co-occurrence network diagrams (Figures 3 and 4) showed frequent mentions of block wall-related safety concerns. Although full analysis by other AIs (Gemini, Deepseek, Copilot) and the author is not shown, results confirmed the lack of block wall disaster-prevention information.
Street View confirmed the locations of many block walls (Table 1), though some features like rebar presence were indiscernible. These locations were plotted on a blank map (Figure 5), and safe evacuation routes were derived and marked in green.
4. Discussion
4.1 Consistency Across AI Analyses
Table 2 summarizes common themes found by each AI. All emphasized the need for clearer evacuation routes, danger zone markers, and barrier information. These findings were consistent across all AIs, demonstrating a certain reliability in AI-assisted text analysis.
4.2 Validity of Street View Surveys
Table 3 compares block wall counts from actual walking surveys with those obtained from Street View along the green evacuation route. Differences were minimal (fewer than five), and all indicated a decrease in wall numbers. Thus, Street View was deemed effective for identifying potential hazards.
5. Conclusion
This study successfully used AI to analyze student feedback on hazard maps and validated the use of Street View for block wall surveys. The methods showed reliability, offering a cost-effective and replicable approach to community disaster prevention research.