透過您的圖書館登入
IP:18.117.76.7
  • 學位論文

建構以情感為基礎之社群影響力模式

Constructing A Sentiment-Based Social Influence Model

指導教授 : 張瑋倫

摘要


根據市調機構美國皮尤中心(Pew Research Center)於2015年3月的的報告指出,美國民眾透過社群媒體接收新聞或生活訊息的比率達到63%,代表Facebook與Twitter已不是單純與朋友間交換訊息的平台,而是用戶們接收新訊息最重要的平台,但平台上每日生產鉅量的資訊,已經超出一般人的負荷程度而根據Katz與Lazarsfeld(1955)所提出的兩級傳播理論指出當用戶處於資訊相對混亂的情況下,會轉而尋求網絡中所謂的影響力者或意見領袖的意見來做為自身的判斷依據。 而本研究目的在於建構出具情感的社群影響力模式,即是要協助組織找出真正具有影響力的用戶,本研究透過文獻的整理發現在所有的社群活動的行為中,以轉貼行為與回應行為最具影響性,因此採用Chen et al.,(2012)以轉貼次數與回應次數作為測量基礎的「文章品質」模型。然而,在過去的量化模型當中卻忽略了用戶間互動時背後的情緒意涵,因此本研究在把情感的觀點融入「量化」的影響力模型中。 本研究透過Twitter中挑選了政治與娛樂兩種領域共四位標的人物(歐巴馬、希拉蕊、小賈斯丁、泰勒絲)進行分析,根據研究結果所示,社群用戶不僅會依據貼文內容的不同而產生不同的文章品質分數,也會依據內容顯現出不同的情感,例如藝人的情感比率普遍高於政治人物,主因為藝人的貼文通常較無針對性的議題,不會挑起用戶與用戶間的爭端,而以政治人物來說,因其貼文通常具有針對性的議題,容易使得立場不同的追隨者產生負面的情感。以政治方面為例,在歐巴馬各主題的平均文章品質中獲得最高分的是女權議題,代表歐巴馬在該議題的發文最受群眾「關注」,但如果加入情感的構面進行比較,其實可以發現歐巴馬女權議題的貼文所獲得的情感反饋不盡理想,只排行其所有議題中的第六名,因此在情感因素調節後歐巴馬的女權議題貼文將不會是最具影響力的議題。而以藝人為例,在小賈斯丁與泰勒絲所共同具有得分類貼文如活動、節目宣傳,小賈斯丁的文章品質略高於泰勒絲,以過往量化的影響力模式中會判定是小賈斯丁於該主題獲得較高影響力,但融入情感的觀點後發現其實泰勒絲的情感比率近乎是小賈斯丁的兩倍,代表泰勒絲於貼文中所獲得正面反饋比率是小賈斯丁的兩倍,因此雖然小賈斯丁的貼文較能取得關注但在考量情感的因素後,該主題中較具影響力的用戶會是泰勒絲。 綜上所述,用戶間互動時有時具有情感表達,但過往的研究往往忽略了情感因素,將散發「負面」情感的數字與「正面」得數字視為同一類,所計算出影響力分數並不客觀,因此本研究主期望彌補此缺口,藉由融合了量化計算與質化情感分析,以建構出能夠反映出社群平台中之影響力模型,在未來提供組織判斷社群網絡中影響力者時更客觀的分析方法。

並列摘要


According to the report from PEW Research Center, 63% Americans read daily news or information via social media (e.g., Facebook or Tweet). This indicates social media plays more and more important role in propagating the information. Katz and Lazarsfeld (1955) proposed two-step flow of communication model to explain people will seek others who are influencers or opinion leaders. Our research goal is to devise a social influence model with sentiment analysis to help organization discover real influential people in the social media. Today, social media provides fans to interact with each other such as retweet, reply, or like, Existing literature shows retweetand reply are the most useful functions in social media. Therefore, this research takes into account retweet and reply as the most important metrics in the proposed social influence model, which is also considered as the quality of post. However, one important concept behind those numbers (retweet and reply) is ignored. The sentiment behind retweet and reply is more important than numbers. Our model not only measures the influence from retweet and reply but also considers the factor of sentiment. This research selected four targets (e.g., Brack Obama, Hillary Clition, Taylor Swift, Justin Bieber) from Twitter to examine the proposed model. The results revealed different content has different quality of post and sentiment ratio. The sentiment ratio of celebrities is higher than politicians because the celebrities usually posted random issues but the politicians’ tweets are easy to provoke a conflict. In summary, fans or followers sometimes may have emotional expressions but previous studies ignored the emotional factors. The calculated number from retweet or reply is not objective. Sentiment analysis can adjust it based on qualitative perspective of content. The main expectation of this study is to fill up the gap by taking into account quantitative and qualitative perspectives. The proposed model can appropriately calculate the social influence of a specific person in the social media to help companies provide adequate marketing campaigns.

參考文獻


3.邱淑華(2012)。從意見領袖到影響者:談網路時代的影響力行銷。廣告學研究,37,33-51。
6.鄭宇君、陳百齡(2013)。探索2012台灣總統大選之社交媒體浮現社群:鉅量資料分析取徑。新聞學研究,120,121-165。
5.黃泓彰(2002)。網路評價搜尋結果的正負意見分類系統。(未出版之碩士論文)。國立政治大學資訊管理研究所,台北市。
39.Pang, B., & Lee, L. (2008). Opinion mining and sentiment analysis. Foundations and trends in information retrieval, 2(1-2), 1-135.
2.Bell, D. (1975). Ethnicity and social change. In N. Glazer & D. P. Moynihan (Eds.), Ethnicity: Theory and Experience. 141-174. Boston, MA: Harvard University Press

延伸閱讀