本論文將在維基中設計一套即時推薦系統,透過條目的歷史編輯取得編輯者紀錄,經由協同過濾運算,進而預測出瀏覽此條目的使用者可能偏好的條目,並且即時地推薦相關條目讓使用者作為可編輯或瀏覽的參考依據。 根據社群90-9-1定律顯示出在社群編輯貢獻者與瀏覽者懸殊的用戶比例,如何透過推薦系統讓90%單純只瀏覽的使用者與9%只有少量貢獻的使用者,都可以很方便的加入編輯的行列將是本論文的研究目標。 本論文的主要貢獻有: (1)在維基百科上實現了即時推薦系統 (2)透過推薦系統找出使用者可能有興趣、有能力編輯的條目 (3)本系統採用非個人化的推薦,對於從無編修過條目的使用者,也可得到有效的推薦 (4)本系統為動態更新系統,使用者在瀏覽維基條目的同時,能即時運算推薦條目給使用者 (5)實驗證明經由此系統所推薦的條目有很好的推薦配對率。
This thesis proposes a real-time recommendation system on Wiki. By investigating the revision history of the article in Wiki, the contributions of the editors can be retrieved and collaboratively filtered. Then, the filtering results can be used to predict user’s preferences for articles in real-time when he/she is browsing Wiki. “90-9-1” community law shows the phenomenon of participation inequality; that is the scale between editors and viewers is extreme in the community. How to recommend wiki articles to the viewers with less contribution and make them edit articles conveniently based on what they are interested in are the objectives of this thesis. The main contributions of this thesis are as follows: (1) Propose a real-time recommendation system on Wiki. (2) Recommend the articles which users may be interested in or capable of editing. (3) As the preference of individual users is not considered (non-personalized recommendation), common articles are recommended to all users. Thus, the proposed recommendation system could be effective for users who never edit any articles in Wiki. (4) The proposed recommendation system is dynamically updated so that when users browse the Wiki article, related articles are recommended to users in real time. (5) The experiment results show that the proposed recommendation system performs well and has good accuracy of predictions.