TwoLayered Architecture for PeertoPeer Concept Search

Peer-to-peer search is an alternative to address the scalability concerns regarding centralized search, due to its inherent scalability in terms of computational resources. Concept search is an information retrieval approach which is based on retrieval models and data structures of syntactic search, but which searches for concepts rather than words thus addressing the ambiguity problems of syntactic search approach. In this paper, we combine peer-to-peer search with concept search and compare two di®erent approaches to peer-to-peer concept-based search. In a single layer peer-to-peer concept search, a single universal knowledge is used to map terms to concepts. In a two layered peer-to-peer concept search, the universal knowledge is used in identifying the appropriate community given query terms and the community's background knowledge is used to map query terms to appropriate concepts within selected community. Since the search is done only in relevant communities, there will be improvement in bandwidth utilization for search. Since two layered peer-to-peer concept search uses the community's background knowledge to map terms to concepts, it improves quality and e±ciency of search. We perform experiments to compare the two layered peer-to-peer concept search with single layered peer-to-peer concept search and present our results in this paper. We also show in this paper how the two layers of knowledge help in achieving scalable and interoperable semantics.