Intelligent Algorithms for Improving Communication Patterns in Thematic P2P Search

Abstract

The Internet is a cooperative and decentralized network built out of millions of participants that store and share large amounts of information with other users. Peer-to-peer systems go hand-in-hand with this huge decentralized network, where each individual node can serve content as well as request it. In this scenario, the analysis, development and testing of distributed search algorithms is a key research avenue. In particular, thematic search algorithms should lead to and benefit from the emergence of semantic communities that are the result of the interaction among participants. As a result, intelligent algorithms for neighbor selection should give rise to a logical network topology reflecting efficient communication patterns. This paper presents a series of algorithms which are specifically aimed at reducing the propagation of queries in the network, by applying a novel approach for learning peers´ interests. These algorithms were constructed in an incremental way, so that each new algorithm presents some improvements over the previous ones. Promising results were obtained through different simulations designed to test the reduction of query propagation as well as the maximization of the clustering coefficient of the emergent logical network.