Introduction: Because messages are broadcast on social media timelessly and efficiently, it is a critical disaster management survey tool. Taiwan’s Regional Emergency Medical Operation Center (REMOC) has responded to nationwide disaster messages in real time since 2005. To construct an early warning system to reduce the delay between the outbreak of accidents and messages received by REMOC. Methods: Timely data, often from first responders and social media, is critical. A system was developed to collect disaster related posts from social media platforms, focusing on keywords like typhoon, flooding, and traffic accidents. Given the mix of topics, informal language, and frequent errors, text analytics and the SVM-based tool LibShortText were used to classify relevant messages. REMOC’s system identified 3,022 disaster-related messages out of 66,588 total posts. We calculated event frequency in 2019, response speed, and engagement rate (ER) to assess the responsiveness of traditional media and individual users. Results: Most messages collected were related to earthquakes (23%), followed by traffic accidents (17%), fires (12%), and floods (3%). Of 702 earthquake-related messages (142 episodes), social media users shared 35 messages about 18 earthquakes, while traditional media posted 667 messages about 124 earthquakes. Social media consistently posted 100% of earthquake messages faster than traditional media (“early dissemination efficacy”: 124/124). For 251 fires (367 messages), social media shared 122 messages on 82 fires, with an early dissemination efficacy of 0.57 hours. For 32 flood events (117 messages), social media posted 55 messages on 16 floods, achieving an early dissemination efficacy of 0.65 hours. Conclusion: Results show that social media users responded faster than traditional media to disaster news, especially for earthquakes, fires, and floods. Earthquakes and floods had the highest response rates from social media. Integrating REMOCs and social media information systems can greatly enhance preparedness and response, particularly for earthquakes and floods.
Kao et al. (Sun,) studied this question.