An efficient random access inverted index for information retrieval

Abstract

To improve query performance and space efficiency, an efficient random access blocked inverted index (RABI) is proposed. RABI divides an inverted list into blocks and compresses different part of each block with the corresponding encoding method to decrease space consumption. RABI can provide fast addressing and random access functions on the compressed blocked inverted index with the novel hybrid compression method, which can provide both block level and inner block level skipping function and further enhance both space and time efficiencies without inserting any additional auxiliary information. Experimental results show that RABI achieves both high space efficiency and search efficiency, and outperforms the existing approach significantly.