Information retrieval in a peer-to-peer environment
InfoScale '06 Proceedings of the 1st international conference on Scalable information systems
Decentralized decision making process for document server networks
GameNets'09 Proceedings of the First ICST international conference on Game Theory for Networks
| Hi-index | 0.01 |
A distributed search system consists of a large number ofautonomous search servers logically connected in a peer-to-peer network. Each search server maintains a local indexof a collection of documents available at the server oron other peer machines. When a query is received by anyserver in the network, a distributed search process determinesthe most relevant search servers and redirects thequery to them for processing.In this paper, we model the distributed search processas Markov Decision Processes (MDPs). The estimated relevanceof a server to a query is regarded as the rewardin the MDP model. Once the MDP policies representingthe global knowledge are obtained at each server throughasynchronous value iteration, the most relevant servers toa given query can be efficiently identified despite the lackof centralized control and global knowledge at each autonomousserver.We discuss the implementation and complexity of theasynchronous value iteration and how we extend the traditionalMDP to handle the multiple-access policy (i.e., morethan one optimal server is returned) and queries with multipleterms. Finally, experiments are conducted using theTREC collection. We show that the MDP-based distributedsearch can achieve results very close to that of a centralizedsearch.