Using crowdsourcing to process low-quality disaster information

With the popularity of mobile devices, the vast amount of disaster responses from the crowd results in the error and duplication of the disaster responses, which is not conducive for the government to relief disaster. This research aims to eliminate the error and duplication of the disaster responses through crowdsourcing. We used community website to recruit crowd and integrated both artificial intelligence (AI) and crowd intelligence (CI). We first used the AI filter with keywords to delete the inaccurate responses and structuralized the accurate responses by geocoding and GIS spatial analysis. Then, in the CI filter, designed the website with the user-friendly user interface to guide the crowd in completing the mission of consolidating duplicated responses and redefining unstructured responses correctly. We verified the feasibility of the process in an actual case during Typhoon Soudeloron August 2015 and Typhoon Dujuan on September 2015 in Taiwan. 210 volunteers from the Internet participated in the crowdsourcing missions and completed the 5906 groups of mission integration. The results showed that crowd are pleased to complete crowdsourcing mission through community website and have the ability to consolidate duplicated disaster responses accurately and quickly with the guidance of the platform.