透過您的圖書館登入
IP:54.160.133.33
  • 學位論文

建立眾包成雙比對法辨識重複災情通報

A Crowdsourcing Pairwise Comparison Method for Recognizing Duplicated Disaster Reports

指導教授 : 康仕仲

摘要


災情通報是災害應變過程中最重要的第一手資訊,決策者會根據這些資訊來安排人力確認、搶救應變或分配救災資源。然而,每當災害發生,經常會有重複通報的情況,造成額外的人力派遣或資源分配不足,不僅升高人員暴露在災害下的風險,同時也增加災害應變的人力及資源需求。為解決上述議題,本研究建立眾包成雙比對法辨識重複災情通報,並設立兩個研究目標:(1) 將辨識重複通報的工作有系統地分配給群眾;(2) 建立量化眾包需求的評估方法。眾包成雙比對法能有系統地減少比對數量,並分配重複通報的項目給不同的群眾,其中,通報比對的眾包結果會經過演算法驗證其正確性。接著,提出量化眾包需求的評估方法,包括眾包工作量、群眾個體能力、眾包需求人數。本研究建立P+眾包平台,並於2015杜鵑颱風中實際驗證,結果在30小時內蒐集到5,656件災害通報,從Facebook募集共93位群眾,總共完成1,980組通報比對。透過P+,群眾平均需要11秒才能判斷出重複通報,最多25秒內便可完成判斷。此外,將通報分組後再互相比對同組內的通報,能有效降低九成以上的通報比對數量,且群眾個體能力的估算僅有-1.45%的誤差,這些數據顯示P+平台有潛力辨別重複災害通報。根據驗證結果,本研究發現災害通報呈現指數型成長,但群眾募集量卻是線性增加,因此未來可將通報細分成更多組別,以減少眾包工作量,或是增加群眾通報比對數量,透過分析社群媒體的募集效益,來提升眾包人數。同時,本研究建議平時要訓練一群積極的參與者,以利災時應變能提升比對數量。整體而言,本研究建立的眾包平台具有發展性,而眾包需求評估法則有助於建立階段性目標。

並列摘要


Disaster reports play an important role in acquisition of information for disaster responses. In the disaster emergency operation, decision-makers need to decide the allocation of labor and resources according to disaster reports. However, some reports are duplicated so that decision-makers might allocate resources again on the same event. This causes not only higher risk of losing relief members' lives but also short of relief resources. To resolve the problems, this research addresses two issues: (1) develop a crowdsourcing platform for recognizing duplicated disaster reports; (2) develop an evaluation method for crowdsourcing requirements. We developed a crowdsourcing pairwise comparison method to systematically crowdsource the work of recognizing duplicated reports with lower workload. Specifically, results of comparison tasks are determined based on validated feedback from crowds. Then, we raised a crowdsourcing requirement evaluation method, which quantizes the workload of crowdsourcing tasks and the crowd individual capacity for the estimation of required number of crowds in our project. Finally, we established an online platform P+ and applied it during a real disaster, Typhoon Dujuan in 2015 in Taiwan. Statistically, 5,656 disaster reports were collected in 30 hours, and 1,980 tasks were assigned to 93 registered people gathered from Facebook. The crowds could complete a comparison task in 25 seconds. We found that reports grouping was effective in decreasing over 90% crowdsourcing workload in P+. Besides, the estimation of crowd individual capacity had the accuracy of -1.45% error. These results show that our platform has potential for the recognition of duplicated reports. Moreover, we had an important observation that disaster reports had an exponential growth but volunteers increased linearly. Therefore, it is necessary to divide reports into more groups with lower workload. Another solution is to determine the factors of increasing crowds in social media. Further, we suggest to regularly gather active volunteers in social media for more crowdsourcing capacity. In conclusion, P+ is a feasible crowdsourcing platform, and the crowdsourcing requirement evaluation method helps set up terms of project goals.

參考文獻


Acar, A., and Muraki, Y. (2011). “Twitter for crisis communication: lessons learned from Japan’s tsunami disaster.” International Journal of Web Based Communities, 7(3), 392.
Comfort, L. K. (2004). “Coordination in Rapidly Evolving Disaster Response Systems: The Role of Information.” American Behavioral Scientist, 48(3), 295–313.
Cutter, S. L. (2003). “GI Science, Disasters, and Emergency Management.” Transactions in GIS, 7(4), 439–445.
Gao, H. (2011). “Harnessing the Crowdsourcing Power of Social Media for Disaster Relief.” IEEE Intelligent Systems, 26(3), 10–14.
Goodchild, M. F., and Glennon, J. A. (2010). “Crowdsourcing geographic information for disaster response: a research frontier.” International Journal of Digital Earth, 3(3), 231–241.

延伸閱讀