Toward Personalized Peer-to-Peer Top-K Processing
Résumé
We present the first personalized peer-to-peer top-k search protocol for a collaborative tagging system. Each peer maintains relevant personalized information about its tagging behavior as well as that of its social neighbors, and uses those to locally process its queries. Extensive experiments based on a real-world dataset crawled from del.icio.us shows that very little storage at each peer suffices to get almost the same results as a hypothetical centralized solution with infinite storage.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...