Enriching topic-based publish-subscribe systems with related content

This demonstration presents RMFinder (Related Messages Finder), a system that retains the simplicity and efficiency of topic-based P2P pub-sub, while providing a richer service where users can automatically receive all messages related to those in the topics to which they are subscribed. RMFinder is based on a novel, dynamic, distributed clustering algorithm, that takes advantage of similarities between topic messages to group topics together, into topic-clusters. The clusters adjust automatically to shifts in the focus of the messages published by the topics, as well as to changes in the users interest, and allow for an effective delivery of related messages with minimal overhead. Categories and Subject Descriptors C.2.1 [Network Architecture and Design]: [Distributed networks] ; C.2.4 [Distributed Systems]: [Distributed applications] ; H.3.3 [Information Search and Retrieval]: [Clustering] General Terms Algorithms, Performance, Experimentation Keywords Publish-Subscribe, P2P, Dynami...

• Type: Journal
• Year: 2008
• Where: SIGMOD
• Authors: Rubi Boim, Tova Milo
• Keywords: Database, Richer Service, SIGMOD 2008, Topic Messages, Topic-based P2p Pub-sub