公關危機是指某些會使組織形象遭受到損害的突發事件，若對於這些危機不妥善處理，可能將會造成災難性的後果。組織為了避免危機的發生，必須全面性的了解組織在社會大眾間的評價。在網際網路發達之後，網際網路成為一個新的傳播媒介。網路社群的討論有別於傳統媒體，其發表速度更快、傳播更廣。有別於傳統公關危機，本研究所關切的是網路論壇所形成的社群公關危機。 在社群中，「聲量」代表著民眾對於某議題的討論度。對組織而言，民眾對於組織的評價是重要且不可忽視的。當民眾對於某議題有著大量的負面意見或因為議題本身的爭議性而導致正反兩面意見的爭論不休，進而顯現在聲量上的大幅或持續的上升，對於組織而言，這就是一種預警訊號。因此組織必須隨時監控聲量來關注民眾對於組織的評論，以儘早知道異常的發生。 在本研究的方法中，運用網路爬蟲抓取與組織有相關的所有文章，建立管制圖對聲量進行監控，並以尼爾森法則為依據，訂定異常訊號的偵測規則。透過文章的來源分佈及頻道來得知最主要的發言群，讓組織能夠針對性的提出解決對策化解危機。在組織提出解決方案後，繼續追蹤事件的發展，以確保解決方案的有效性。 本研究結果顯示，以每日為單位的管制圖，雖然無法偵測出所有即將發生的危機事件，但仍有約50%的案例被成功偵測出，且在後續的追蹤中可以有效的判斷組織的解決方案是否有效。未來的研究將朝向縮短監控的時間間隔，以期提高本研究方法的成功率。 Public relations crisis occurs when unexpected events damage the organization’s image. The organization may face a disastrous consequence if the crisis is not attended. To avoid the crisis, the organization has to know the public’s appraisals constantly and comprehensively. The advance and popularity of Internet has made Internet a new media. Unlike traditional media, the discussions in online communities spread faster and broader. This study focus on the public relation crisis incurred by online community forum other than the traditional media. “Volume” indicates how animated a topic is discussed in the community. The public’s appraisal is very critical and cannot be ignored by the organization. The volume may increase dramatically or rise continuously when there is a great deal of negative opinions to the organization or the topic is controversial, which signal a potential crisis to the organization. Therefore, the organization should watch the public’s appraisal by monitoring volume in online community to early detect the occurrence of unusual events. This study uses the web crawler to collect articles relevant to the case organizations, constructs control charts to monitor the volume, which is measured from the collected articles, and then formulates detection rules based on the Nelson rules to detect abnormal signals. Main spoken groups are identified based on the threads and channels of article sources, which enable the organization to formulate targeting actions to resolve the crisis. The organization keeps tracking the development of the event to ensure the effectiveness of its solutions. Though the constructed control charts are unable to detect all upcoming crises by our case studies, our approach indeed detect 50% of the cases and successfully tracts the follow-up development to valid the organization''s solution. Our future study will adopt a shorter sampling interval to construct the control charts to expect a greater accuracy of event detection.
