結合文字探勘與社會網絡分析尋找虛擬社群中之意見領袖

Identification of opinion leaders in virtual community using the integration of text mining and social network analysis

葉姵妏(Pei-Wen Yeh)

指導教授 ： 洪智力

中原大學/商學院/資訊管理研究所/碩士(2012年)

https://doi.org/10.6840/cycu201200085

摘要

電子口碑傳遞過程中的關鍵人物是很重要的，其扮演著散佈資訊的角色。意見領袖就是這其中的關鍵人物之一，並且在網絡中屬於橋樑的溝通位置。在過去的研究中，相當多數的研究是利用社會網絡分析整個網絡的關係，例如分析網絡的大小、中心性等關係。而其中的網絡中心性，則可計算分析出網絡中的重要節點為何。若某一節點落在網絡中心性上，則這個節點成為意見領袖的可能性就相對的大。但，網絡中意見領袖所發表的文章內容也是識別意見領袖的重要標的，從過去的相關研究發現，並未加以從意見領袖文章內容特質進行質與量的分析，若只是利用社會網絡分析去判斷網絡中的意見領袖尚有不周全之處。因此，本研究透過三個模組來進行實驗，(1)社會網絡分析模組、(2)文字探勘模組、(3)整合模組，透過社會網絡分析與文字探勘技術，計算了節點的仲介中心性與識別意見領袖文章的因素，並加入了專家權重增加意見領袖因素的可靠性。實驗結果顯示，本研究提出的文字探勘模組與整合模組的確有較好的效果，準確率皆比社會網絡分析模組更佳。證實加入意見領袖的文章等因素加以判斷，更能提升找尋到關鍵意見領袖的準確率。

關鍵字

意見領袖 ； 社會網絡分析 ； 文字探勘

並列摘要

In the process of electronic word-of-mouth, key people play an important role in the dissemination of information. Opinion leaders are one type of these key people, holding a position in the network for building bridges for communication. The majority of previous studies analyzed relationships within the context of the entire network, e.g. analyzing the relationship between centrality and the size of the network, etc. Given a network’s centrality, one can then calculate and analyze the network’s key nodes. If a node falls within the network’s centrality, this node has a considerably greater likelihood of being an opinion leader. In the network, content published by opinion leaders is also a major indication for recognizing opinion leaders. However, previous related studies in this area have not taken content written by opinion leaders as factors for quantitative or qualitative analysis, relying only on social network analysis to determine the identity of opinion leaders. Therefore this study conducts experiments from three approaches: (1) a social network analysis module, (2) a text mining module and (3) an integrated module. Through social network analysis and text mining techniques, we calculate factors of the nodes’ agency centrality and the identification of content provided by opinion leaders, thus increasing the reliability of expert-weighted factors for opinion leaders. Experimental results indicate that the text mining and integrated modules proposed by this study provide more accurate results than the social network analysis module. This confirms that adding factors related to articles written by opinion leaders improves the accuracy of determination of opinion leaders.

並列關鍵字

social network analysis ； opinion leader ； text mining

參考文獻

48. 吳齊殷、莊庭瑞（2004），超連結網絡分析：一項分析網路社會結構的新方法，資訊社會研究，第6卷，頁127-148。

53. 王思峰、劉麗萍（2007），虛擬社群中誰比較可以信任？，中華心理學刊，49卷1期，頁1-17。

54. 朱國明（2008），以網路的資訊豐富環境與訊息框架探討網路購物行為之前因與後果模型研究，中華管理研究學報，第8卷，頁81-202。

2. Armstrong, A. G., and Hagel, J. (1996). The real value of on-line communities. Harvard Business Review, 134-141.

3. Armstrong, A. G., and Hagel, J. (1997). Net gain: expanding markets through virtual communities. McKinsey & Company.