隨著社群網站的崛起，在虛擬世界互動交流的人數急速成長，其中的微網誌平臺更因使用者發文方便且容易吸引關注者回應的特性，所以張貼的訊息雖然簡短，累積的資料量卻十分龐大。這些訊息含有大量的情緒成分，卻不見學者研究如何利用這些當下表達的情緒，推薦可能投合訊息張貼者所需的產品或服務。 本研究採用卓淑玲、陳學志、鄭昭明（2013）的情緒字詞庫，採擷臺灣熱門微網誌Plurk使用者張貼訊息中的情緒成分，進而透過T檢定、Anova、相關分析及決策樹，探討不同的情緒成分對產品推薦效果的影響。 本研究的實驗結果發現，隱性的情緒誘發字詞比顯性的情緒描述字詞更能提升產品的推薦效果。在各種情緒狀態中，負向情緒、強烈情緒及單一情緒會帶來較佳的推薦效果。正向情緒字詞出現的頻率和產品推薦效果之間的正向關係，也同樣具有顯著效果。這些實驗結果，應有助於運用微網誌或類似平臺的產品或服務行銷商，研判如何集中有限的財務資源於最有可能實現銷售目標的潛在線上客戶。 ;With the rise of social networking sites, the number of people participating in the virtual communities increases rapidly. Even though the messages they posted were brief, users of microblogs, a kind of social networking sites, have contributed an enormous amount of research data because of the characters of easy-posting and other unique features of the microblogging platforms. Although these messages contain a lot of emotional components, there was little, if any, work on how to use the emotion information on social media to recommend products or services to the users and their online circles. In this study, we adopt emotion word database which provided by Cho et al. (2013) to extract emotional aspects from message posted to investigate the effectiveness of different emotional component on product recommendation. The results of this study show that implicit emotion-inducing words are more effective than explicit emotion-describing words when recommending products. Besides, words with strong emotion are more effective than words with mild emotion and negative emotion are more effective than positive emotion and sentences with uniform emotion are more effective than mixed emotion. The research also finds that frequency of positive emotion words significant impact on recommendation effectiveness. The study should be helpful for product or service marketers who use Plurk or similar microbloging platforms to determine how to focus limited financial resources on potential online customers in order to achieve maximum sales revenue.
