Information retrieval is a relevant topic in our days, espe-cially in distributed systems where thousands of participants store and share large amounts of information with other users. The analysis, development and testing of adaptive search algorithms is a key avenue to exploit the capacity of P2P systems to lead to the emergence of semantic communities that are the result of the interaction between participants.
In particular, intelligent algorithms for neighbor selection should lead to the emergence of efficient communication patterns. This paper presents new algorithms which are specifically aimed at reducing query propagation overload through learning peers’ interests. Promising results were obtained through different experiments designed to test the reduction of query propagation when performing thematic search in a distributed environment.