問答網站社群之知識收藏推薦機制

Abstract

隨著網路科技的蓬勃發展、Web 2.0的興起，問答網站逐漸成為重要的知識分享平台。問答網站提供知識社群的服務機制，讓擁有共同興趣或專長的使用者組成知識社群，收藏社群成員有興趣之問答知識，並分享與社群相關的知識議題。但是由於問答網站每天有大量的問答知識產生而造成資訊過量的問題，因此需要社群知識收藏之推薦機制，推薦知識社群相關有興趣之問答知識，以提升社群之知識分享效益。社群知識收藏之推薦機制有別於個人化推薦，為一群體(社群)推薦。 傳統群體推薦機制主要是以群體成員之重要性當權重，來合併各單獨成員之興趣特徵檔以產生群體興趣特徵檔，並進而以群體興趣特徵檔過濾推薦物件。目前相關文獻並未有針對問答網站社群知識收藏的群體推薦機制之研究，而傳統群體推薦機制並未考量推薦物件如問答知識之品質，以及社群成員收藏知識之聲望等因素。本研究提出問答網站社群知識收藏之群體推薦機制，以推薦社群相關有興趣的問答知識文件。所提的推薦方法主要以社群之歷史收藏知識，及社群成員之重要性包括收藏知識聲望與回答知識聲望等，來產生社群群體興趣特徵檔；並進而以社群興趣特徵檔與目標知識文件的相關性，以及知識文件的品質，來推薦社群問答知識文件。社群成員之收藏知識聲望是分析社群內成員間的知識收藏與推薦互動關係所獲得，而成員的回答知識聲望是分析成員過去代表社群回答問題之行為所獲得。此外，對於知識文件品質之評估，主要是考量目標知識文件與回答者專業知識的相關性，以及目標知識文件所獲得之評價。最後本研究以奇摩知識家問答網站做為實驗評估的資料來源，實驗結果顯示本研究所提出的方法比傳統方法能更有效的針對知識社群推薦與其興趣相關的知識文件。

With the rapid development of Internet and Web 2.0 technology, Question & Answering (Q&A) websites have become an essential knowledge sharing platform. This platform provides knowledge community service, allowing users with common interests or expertise to form a knowledge community where community members can collect and share Q &A knowledge (document) of their interest. However, due to the massive amount of Q&A documents created every day, information overload has arisen as one of the main problems. Traditional group recommender systems use member importance as weight to consolidate individual profiles and generate group profiles, which in turn are used to filter out items of recommendation. Previous researches did not investigate recommendation mechanisms for knowledge collections in communities of Question-Answering Websites. Traditional group-based recommendation mechanisms have not considered certain factors of recommended items, such as knowledge quality and the reputation of community member in terms of the popularity of his/her collected Q&A documents. In this work, a novel recommendation approach is proposed to recommend related Q&A documents for knowledge collections in communities of question-answering websites. The proposed approach mainly generates community profiles from the past collected Q&A documents by considering community members’ importance, including member reputations in both collection and answering of Q&A knowledge. The proposed approach then generates recommendations of Q&A documents based on the quality of Q&A documents and the similarity between the Q&A documents and the community profile. Experimental results show that the proposed method outperforms other conventional methods, providing a more effective manner in recommending Q&A documents to knowledge communities