ScoreTree: A Decentralised Framework for Credibility Management of User-Generated Content

摘要

Peer-to-peer applications are used in sharing User-Generated Content (UGC) on the Internet and there is a significant need for UGC to be analysed for credibility/quality. A number of schemes have been proposed for deriving credibility of content items by analysing users' feedback, mostly using centralised computations and/or semi-decentralised approaches. In this paper, we propose our P2P schema, ScoreTree, that decentralises a relatively complex credibility management algorithm by aggregating distributed evaluations and delivering an estimate of credibility for each interested content item. Our experiments show that our schema compares favourably with existing decentralised approaches including a gossip message based implementation of ScoreFinder, and a widely adopted P2P application called Vuze.