既有虛擬社群運作機制因資訊爆炸且群眾智慧品質控管不易落實，致使相似內容同時存在過多差異大之評論，於此，知識需求者為取得正確知識，將須投入相當之時間本於評估群眾智慧之品質，以及社群成員可參考價值，再者，由於知識正確性評估所面臨之挑戰，將致使部分知識貢獻者於未知狀況下傳遞錯誤之資訊，而當錯誤資訊持續累積至一定程度，且逐步影響整體群眾智慧之品質時，將降低知識貢獻者分享知識之意願。 有鑑於上述問題，本研究乃發展虛擬社群知識群聚度判定模式，並劃分為「虛擬社群問答契合度解析」，以及「成員參考度解析」兩核心模組，以群眾智慧之知識凝聚為切入點，發展並設計一套適用於虛擬社群之方法論。於前者中，本研究結合領域性關鍵詞彙集合概念、文章內容核心資訊萃取、文章相似度判定，以及詞彙語意分析等技術，以判定問答品質、問答內文及語意契合度；於後者中，本研究則整合頻繁項目集為基分群法、向量空間模型及語意關係距離解析等技術，以解析得成員領域知識可參考程度。 承接於本研究發展之模式，為確認本方法論於實務應用中之可行性，本研究乃建構一套以網際網路為基礎之虛擬社群知識群聚度判定系統。此外，為驗證本系統績效，本研究以「搜搜問問」虛擬社群真實問答資料作為驗證資料之樣本，並以當中之精華知識文章，作為領域訓練資料樣本，針對「問答契合度解析」與「成員參考度解析」進行獨立驗證，確保兩相議題間之驗證結果不相互影響。另一方面，本研究亦以案例為導向探討整合兩核心模組，所建構之「虛擬社群知識群聚度判定系統」，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，討論得本系統具備相當程度之實質管理效用。 整體而言，本研究發展之模式與系統，於社群管理者層面，將為其提供標準化且具公正性群眾智慧品質之管理指標，以提升群眾智慧管理之實質效益，於知識需求者層面，則降低其篩選知識時所需花費之成本，且同時提升其獲取所需且正確知識之機率，減少錯誤資訊再傳遞之行為，最終將藉以增加實具品質群眾智慧之累積，促使更多社群成員知識分享之意願，以活絡虛擬社群之利用率，進而催化虛擬社群持續性之整體發展。 In this information explosion era, the quality of collective intelligence is not easy to control. Knowledge demanders are required to invest an amount of time in evaluating the collective intelligence quality of virtual community to acquire correct knowledge. In addition, some knowledge providers might provide poor quality information in this circumstance and overall collective intelligence quality would be affected. Among that, there are two issues need to be discussed. Firstly, the question and answer suitability cannot be guarantee and secondly, member’s (knowledge provider’s) reference degree cannot be determined. Therefore, to improve the accumulation of quality of collective intelligence and encourage more members to share knowledge in virtual community, this paper develops a Virtual Community Knowledge Clustering Degree Determination (KCD) model including Question and Answer Suitability Determination (QASD) module and Member’s Reference Degree Determination (MRDD) module to enhance collective intelligence quality. In QASD module, this paper employs keyword extraction, article’s critical information extraction, article similarity determination and semantic analysis techniques to determine the suitability of Q&A content. In MRDD module, this paper integrates frequent itemsets-based clustering, vector space and semantic analysis technologies to evaluate the reference value of knowledge providers. 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 is applied to evaluate the proposed model. The verification results show that the developed system is a high-performance knowledge clustering degree determination system. As a whole, this paper provides an approach for virtual community to efficiently evaluate the clustering degree of knowledge.
