Termset-Based Indexing and Query Processing in P2P Search

Multi-term query is a common issue in information retrieval system. In large-scale P2P information retrieval, the method of indexing and query processing based on single-term results in large bandwidth cost. We take into account the correlation among terms and propose a termset-based indexing and query processing method suited for information retrieval in structured P2P overlay. Employing statistics, metadata and query log, we construct a dynamic termset corpus, and the index is built based on termset. When processing query, the peer extracts the termsets from the query terms, and each termset is treated as a key. Several methods are applied to reduce bandwidth consumption. We also present a method of query expansion to be a complement when there are no sufficient results. The experiments show that our method has good performance, and it is suitable for large-scale distributed information retrieval.