Day_90 : (Re)Evacuation research literature analysis-A Text mining

Evacuation’s research literatures are divided into two categories for this analysis. One is natural disaster’s research literature conducted by all specialties. The other is social scientist’s research literature on natural disasters. The database, the Springer link, is selected to conduct all field’s evacuation research literature analysis. The E. L.Quarantelli Resource Collection (See the Website), Disaster Research Center of the University of Delaware was chosen as a target database for analyzing social science literature. The collection is one of the world’s most complete ones on the social and behavioral science aspects of disasters. These two databases’ literatures were analyzed by a text mining. To conduct the text mining, the RH Corder was used.

The following is just one result example, a content analysis of the springer link database.

  1. Search words are “evacuation AND urban AND (tsunami OR flood OR typhoon OR hurricane)”
  2. The number of extracted literature is 824 (2000-2014)
  3. The titles, key words, and abstracts of the 824 were combined into one text file
  4. The extracted words which appear over 20 times in the text are shown in Table1
  5. A co-occurrence network analysis result is indicated in Figure 1

Table 1       Extracted Words (over 20 times) and Frequencies (sorry, original Japanese version’s words are left)

wordsfreq

140715_村上先生_共起ネットワーク2

Figure 1 A co-occurrence network analysis result

In Figure1, the circle sizes around the words (Nodes sizes) mean the frequencies of the words appeared in the text. Edges mean the connections between the words. Then, you can see the above analysis (by color) result.

For instance, emergency response-preparedness-decision-support with “event” are combined with evacuation as key words. Climate-change-impact was also detected with coastal-adaptation. We can estimate that detected Taiwan-assess-community-resilience represents the Typhoon Morakot disaster. (Then, this is confirmed by returning to the original text.)

Murakami et al. (Murakami, Nakasu, Shimamura, Goto, and Ogawa, 2015) is referred.