A classifier module for analyzing community responses on disaster preparedness

In this paper, we implemented, trained, and evaluated a module to automatically classify community responses of respondents of an e-participation toolkit on disaster preparedness. Responses were automatically classified into 10 different categories, which were derived from a codebook created for the content analysis of the responses. Results show that most responses were classified in the category Local Government Unit (LGU) accountability. The classifier achieved a 93.3% mean accuracy with ten-fold cross validation. This analysis shows that the created classifier module can successfully decode responses and can be further improved by applying multi-label classification. However, the classifier achieved 63.8% agreement with the manual coding, and a revision of the code book may be deemed necessary as a future work. The application of pre-processors such as spelling checkers can also be considered.