Two-layered architecture for peer-to-peer concept search

The current search landscape consists of a small number of centralized search engines posing serious issues including centralized control, resource scalability, power consumption and inability to handle long tail of user interests. Since, the major search engines use syntactic search techniques, the quality of search results are also low, as the meanings of words are not considered effectively. A collaboratively managed peer-to-peer semantic search engine realized using the edge nodes of the internet could address most of the issues mentioned. We identify the issues related to knowledge management, word-to-concept mapping and efficiency in realizing a peer-to-peer concept search engine, which extends syntactic search with background knowledge of peers and searches based on concepts rather than words. We propose a two-layered architecture for peer-to-peer concept search to address the identified issues. In the two-layered approach, peers are organized into communities and background knowledge and document index are maintained at two levels. Universal knowledge is used to identify the appropriate communities for a query and search within the communities proceed based on the background knowledge developed independently by the communities. We developed proof-of-concept implementations of peer-to-peer syntactic search, straightforward single-layered and the proposed two-layered peer-to-peer concept search approaches. Our evaluation concludes that the proposed two-layered approach improves the quality and network efficiency substantially compared to a straightforward single-layered approach.