SDEC: A P2P semantic distance embedding based on virtual coordinate system

Abstract

Large-scale P2P applications can benefit from the ability to predict semantic distances to other peers without having to contact them first. In this paper, we propose a novel semantic distance embedding approach, SDEC, in P2P network, which assigns synthetic coordinates to peers such that the distance between the coordinates of two peers approximately predicts the semantic distance between any two peers. Specifically, the semantic distance between peers is quantitatively characterized through vector space model based on peers' semantic profiles, and then, based on measured semantic distances from a peer to a handful of other peers and the current coordinates of those peers, we adopt the spring relaxation method, mimicking the physical mass-spring system, to simulate the semantic embedding procedure, which can find minimal energy configuration corresponding to relatively accurate semantic embedding. Simulation results show that a 3-dimensional Euclidean model can embed these peers with relatively high accuracy. © 2009 Springer Berlin Heidelberg.