Content-Based Document Recommendation in Collaborative Peer-to-Peer Network

Abstract

As the Internet infrastructure has been developed, many diverse and effective applications attempt to gain the potential of that infrastructure. Peer-to-Peer, one of the most representative systems for sharing information on a distributed environment, is a system can helps peer users to share their files with other peer users easily. But Peer-to-Peer network includes not only uncountable files but also plenty of duplicates, which is, bring about increase of network traffic. To solve this problem, we suggest an effective information sharing system supporting collaboration among distributed users with similar interests, or who are part of the same workgroup. In this paper, we exploit the techniques of association rules in deriving peer user profiles represented as a prefix tree structure called PTP-tree (Personalized Term Pattern tree). In addition, we employ content-based filtering approach to search documents that are similar to personalized term patterns. For the performance evaluation, we formed a simple Peer-to-Peer network to make experiments on real data and 10 users. Experimental results show that the proposed system helps users to reduce time for gathering documents relevant to users’ needs. In addition, PTP-tree structure of a user profile saves the memory usage.