在網際網路發達的時代，隨著社群媒體的快速發展，人們可以很方便地透過社群媒體，發表各人對企業所提供的產品或服務的意見。企業也必須懂得從社群網站中挖掘商機，最普遍的行銷手法就是直接在線上與潛在的消費族群作互動，Facebook粉絲專頁為中小企業最常使用的互動平台，當企業發布最新的行銷資訊，粉絲可以直接針對行銷資訊提出個人的看法，透過社群擴散的模式，將行銷資訊分享給更多人知道，而本研究提出的核心機制是根據企業在粉絲專頁所提供的消費資訊，以人工標記的方式，將貼文分成四大類型，包含產品資訊、活動促銷資訊、實體店面資訊、與企業合作資訊，透過粉絲之間在評論上的互動，運用社群網路中內向度的計算與情緒分析，找出具有影響力的評論，從篩選結果可以發現，具有影響力的評論都是針對企業所發出行銷資訊提出自己的看法及疑問，而未篩選過的評論主要都是文不對題、不具意義的評論，透過篩選前後的比對，可以有效的找出對企業有幫助的評論，進而提升企業服務品質、改善經營績效。

With the advent of social media, people can express their opinions on the products or services offered by a company. A company should also make use of his social website to attract potential customers. The most effective marketing strategy is to interact directly with the consumer groups through a social website. Among different social networks, Facebook fan page is the most commonly used website for small and medium companies. A company can publish his latest marketing information on a Facebook fan page, customers then directly leave comments on the marketing information. Through the diffusion of information among the community, more and more people know about the products and services of the company. Our research is proposed a mechanism to analyze the comments and to filter out unimportant comments. To this end, we first manually divide the posts on a Facebook fan page into four different types, including the product description, the promotional campaign information, the bricks and mortar information and the cooperation information. Then, we use the UCINET to calculate the in-degree of a comment. The comments with small number of in-degrees are insignificant comments and are therefore deleted. From the remained posts after the filtering operation, we found that most of the influential comments are belong to the comments on promotional campaign information and the filtered comments are usually irrelevant or meaningless. By filtering out the unimportant comments, a company can focus on improving his products or services based on important comments.

[1] 資訊工業策進會創新應用服務研究所 (2017年5月)。台灣人愛黏社群網站 8成有這個帳號。聯合新聞網。Retrieved from :　https://udn.com/news/story/7266/2436253
[2] Doris Lin (2016年9月)。電商的Facebook粉絲專頁經營，6步驟打好基本功。TransBiz。Retrieved from : http://transbiz.com.tw/facebook-%E7%B2%89%E7%B5%B2%E5%B0%88%E9%A0%81%E7%B6%93%E7%87%9F/
[3] 馬安奇 (2014年8月)。臉書上一個讚值多少錢？社群行銷價值公式，看這裡！。經理人月刊。Retrieved from :
https://www.managertoday.com.tw/articles/view/44244
[4] Facebook官方 (2016年3月)。全球5000萬企業有粉絲專頁，台灣八成有按讚。今日新聞。Retrieved from : https://www.nownews.com/news/20160304/2016377
[5] James G.Anderson 2002. Evaluation in health informatics: social network analysis, Computers in Biology and Medicine, Vol. 32 , pp. 179-193.
[6] K. Haythommthwaite 2005. Characterized social networks as having the following components, Actors. New York: Nodes.
[7] Peter V. Marsden and Karen E. Campbell (1984年12月) Measuring Tie Strength Social Forces, Vol. 63, No. 2, pp. 482-501
[8] 蔡武憲、許家倫、陳佑宗、高雪華、林政德、官怡萍(2010年1月)。創意團隊與創意績效的社會網路分析。龍華科技大學，企業管理系專題。
[9] Wasserman S. 1994. Social Network Analysis：Methods and Applications. Cambridge: Cambridge University Press, pp. 173~196
[10] Freeman, Linton 1979. Centrality in social networks: Conceptual clarification. Social Networks 1, pp. 215-239.
[11] 張俊龍、彭達楓(2017年)。SNA分析中關於中心度的研究。澳門科技大學商學院。
[12] 葉姵妏(2012年)。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。中原大學，資訊管理所碩士論文。
[13] 李在錫(2008年6月)。汽車社群討論區之社會網絡關聯分析。國立屏東商業技術學院，資訊管理所碩士論文。
[14] 張巨昇(2015年5月)。結合資料探勘與社群網路分析於貨架指派問題。元智大學，工業工程與管理學系碩士論文。
[15] Y. Lu and J. Chen. (2014). Public Opinion Analysis of Microblog Content. 2014 International Conference on Information Science & Applications (ICISA), Seoul, 2014, pp. 1-5.
[16] Lei Zhang, Riddhiman Ghosh, Mohamed Dekhil, Meichun Hsu, & Bing Liu. (2011). Combining Lexicon-based and Learning-based Methods for Twitter Sentiment Analysis. HP Laboratories, HPL-2011-89.
[17] H Tsai, C. (1998). MMSEG: A Word Identification System for Mandarin Chinese Text Based on Two Variants of the Maximum Matching Algorithm. ( http://technology.chtsai.org/mmseg/ )
[18] NTUSD 為台灣大學資訊工程學研究所自然語言處理實驗室所建置的中文情緒詞辭典，詳細資料參見計畫網站 Retrieved from : http://nlg18.csie.ntu.edu.tw:8080/opinion/index.html