Declustering Web Content Indices for Parallel Information Retrieval

Abstract

We consider an information retrieval (IR) system on a low-cost high-performance PC cluster environment. The IR system replicates the Web pages locally, it is indexed by the inverted-index file (IIF), and the vector space model is used as ranking strategy. In the IR system, the inverted-index file (IIF) is partitioned into pieces using the lexical and the greedy declustering methods. The lexical method assigns each of the terms in the IIF lexicographically to each of the processing nodes in turn and the greedy one is based on the probability of co-occurrence of an arbitrary pair of terms in the IIF and distributed to the cluster nodes to be stored on each node's hard disk. For each incoming user's query with multiple terms, terms are sent to the corresponding nodes that contain the relevant pieces of the IIF to be evaluated in parallel. We study how query performance is affected by two declustering methods with various-sized IIF. According to the experiments, the greedy method shows about 3.7% enhancement overall when compared with the lexical method.