結合推文信賴及朋友興趣關係進行以信賴網路為基礎之部落格文章推薦

Abstract

部落格是Web2.0的新興模式，任何人都可以藉由部落格在網路上發表言論與分享資訊。然而隨著部落格文章數量越來越多，產生資訊過多的問題。社交書籤網站提供推文功能並顯示文章推文數，可協助解決部落格文章過載之問題，但仍未能針對使用者個人偏好進行文章推薦。如何針對使用者個人喜好推薦適合文章是一項重要的議題，目前以信賴網路為基礎的推薦系統之相關研究，多以朋友信賴關係或使用者過去評分資料進行分析預測，尚無針對考慮使用者過去推文行為，以及朋友的興趣相似度關係來建立信賴網路之推薦機制。 本研究提出一個結合使用者推文信賴及朋友間興趣相似關係之信賴網路推薦機制，所提機制進而結合使用者聲望，推薦符合使用者個人偏好之部落格文章。本研究所建置的推薦機制由funP書籤網站使用者的歷史行為紀錄進行分析，以找出使用者間的信賴及朋友間興趣相似關係。本研究實驗評估顯示，所提之信賴網路推薦機制確實可以比傳統方法能更有效推薦符合使用者個人偏好的部落格文章。

With the new generation of web-based communities, Web2.0, everyone can make a comment or share information via a blog website. However, with the rapid growth of blog articles that are produced every day, people are facing the problem of information overload. The Social bookmarking website provides users functions to post and push (recommend) articles and shows article push counts to help solve this problem, but it still cannot recommend articles to users based their personal interests. Accordingly, providing users with blog articles that suit their interests is an important issue. Existing researches on trust-network based recommendations mainly use explicitly specified relationship trust or users’ past rating data to predict trustworthiness between users and make recommendations. Very little research addresses the issues of trust propagations and recommendations based on the hybrid of trustworthiness derived from users’ past push behaviors and friend-interest relations. In this work, we propose a novel hybrid trust network, which is constructed by combining two kinds of trust link. Push-following trust links are derived based on users’ post/push-behaviors. Friend-Similarity trust links are derived from the friend relations with the consideration of user preference similarity. In addition, we also derive user reputations by considering the number of following push after users’ post/push. A novel recommendation method, which combines the hybrid trust network and user reputations with user-based collaborative filtering, is proposed to recommend desirable articles satisfying personal interests. Our proposed method uses a dataset collected from the social bookmarking website funP to derive the push-following trustworthiness and friend interest relations. The experiment results demonstrate that our proposed method performs better than conventional approaches in recommending bog articles.