A peer-to-peer hypertext categorization using directed acyclic graph support vector machines

Abstract

DAGSVM (Directed Acyclic Graph Support Vector Machines) has met with a significant success in information retrieval field, especially handling text classification tasks. This paper presents PDHCS (P2P-based Distributed Hypertext Categorization System) that classify hypertext in Peer-to-Peer networks. Distributed hypertext categorization can be easily implemented in PDHCS by combining the DAGSVM (Directed Acyclic Graph Support Vector Machines) learning architecture and Chord overlay network. Knowledge sharing among the distributed learning machines is achieved via utilizing both the special features of the DAG learning architecture and the advantages of support vector machines. The parallel structure of DAGSVM, the special features of support vector machines and decentralization of Chord overlay network lead to PDHCS being more efficient.