[[alternative]]An Integrated Model of Discussion Threads Emotion Category Determination and Emotional Opinion Clusters Interaction Analysis for Vi Rtual Community

系統編號: PB10608-2088計畫編號: MOST106-2221-E343-003執行機構: 南華大學資訊管理學系研究期間: 10608~10707[[abstract]]現今網際網路便利，群眾於網路社群、論壇可搜尋及閱讀感興趣之議題，便隨即表達意見或回應，並附加自身之情緒於文字，表達自身觀點，使網際網路上匯集的意見眾多時，使其中一方情緒成為群眾討論的主議題之情緒類別偏向，致使後續群眾查閱議題並給予意見時，容易受到已附有情緒類別偏向之氛圍影響，且無法客觀地評論議題。此外，由於發表意見時，由於意見發表應是互相影響，意見發表者常需自行分辨可接納之客觀及理性的意見，然而目前較無技術輔助辨別意見之情緒類別偏向與理性程度，導致其他意見發表者在無法發現非理性之意見時，進而無法針對特定議題加以客觀思考，最後發表偏頗之意見。有鑑於此，本研究乃以群眾意見之情緒詞彙為基礎，建立一套「以虛擬社群群眾意見為基之議題情緒類別判定與情緒意見群集互動解析」模式，其中包含「群眾意見相關程度判定」、「議題情緒類別判定」及「情緒意見群集互動解析」三大模組。首先，於「群眾意見相關程度判定」模組中，本研究使用「Apriori 演算法」擷取虛擬社群中意見，並尋找關聯規則與群眾意見之密切關係，再找到其中關係後，將相關性較高之意見進行彙整，進而將彙集後意見判斷其中附有之情緒。其次，於「議題情緒類別判定」模組中，本研究建立情緒詞彙之隸屬係數，並採用「餘弦函數」分析意見之情緒偏向，接著計算所有評論彙集後評論之情緒偏向比率，進而判斷事件議題之情緒類別，以推測情緒偏向議題使群眾產生情緒氛圍發生之機率，可用來即時通知或警示管理者。再者，於「情緒意見群集互動解析」模組中，本研究將議題意見以自然語言處理技術分析蘊含之情感，並以「多目標群集法」、「賽局理論」，藉以求得具影響力之目標情緒意見群集，以及群集意見之理性程度，以遏止議題往非理性、非客觀之情緒意見發展。最後，為確認本研究方法論於實務應用之可行性，本研究乃建構一套以網際網路為基之「群眾意見情緒審核系統」。此外，為驗證本系統之績效，本研究預計以YAHOO 奇摩新聞作為驗證資料來源，並針對「議題情緒類別判定」及「情緒意見群集互動解析」進行整合性驗證。另一方面，本研究亦以案例為導向探討整合三個核心模組，所建構之「群眾意見情緒審核系統」，於實務情境中之應用與管理意涵，並於最終質化探討之結果中，討論得本系統之實質管理效用。[[abstract]]With the convenient access to the Internet, people can search and read the topics which interest them in virtual communities and forums; moreover, they can immediately make comments or express opinions which indicate their emotions. If discussion threads with specific emotional category preferences are accepted by the majority of people and attract many comments and opinions, specific emotions will cumulate. Nevertheless, these emotional options may influence the judgment of other people, making them unable to make objective comments. In addition, as the opinions are expressed in virtual community, the opinion providers often need to distinguish between the objective and rational opinions that are acceptable. However, there is no technical support to distinguish the rational degree of opinions. Then, it is difficult for publishers to objectively think about specific discussion threads, and finally publishes biased opinions. For the above problems, this paper develops "An Integrated Model of Discussion Threads Emotion Category Determination and Emotional Opinion Clusters Interaction Analysis for Virtual Community", divided into three kernel modules, which are " Opinions Relevance Degree Evaluation (ORDE)", "Discussion Threads Emotion Classification (DTEC)" and "Emotional Opinion Clusters Interaction Analysis (EOCIA)" modules. Firstly, in the ORDE module, this paper adopts Apriori to establish the relevance norms and select the opinions/comments highly related to the topics; in the DTEC module, this paper analyzes the emotions of all the opinions highly related to the topics, calculates the ratios of all emotions and evaluates the probabilities of all emotional categories. After that, the administrator can evaluate the emotions for the discussion threads and topics and effectively prevent the specific emotion (e.g., negative) tendency occurrence. Secondly, in EOCIA modules, this paper uses the Natural Language Processing technique to analyze the implied emotions of opinions and uses Multi-Objective Cluster method and Game Theory to obtain an influential target comment cluster and the rationality. After that, the development of irrational and non-objective sentiment can be discouraged. Finally, in order to demonstrate applicability of the proposed methodology, a Web-based system is also established based on the proposed model. Furthermore, a real-world case (e.g., YAHOO News) is applied to evaluate the proposed model. As a whole, this paper provides an approach for virtual community to efficiently examine the emotions of published opinions and discussion threads to facilitate knowledge sharing intention