An Investigation on Multi-token List Based Proximity Search in Multi-dimensional Massive Database

A proximity search looks for similar complex documents such as images, sounds, DNA sequences that share two or more separately matching terms within a specified distance from within a large collection. Retrieving those similar complex documents are of great importance to many applications. To achieve an efficiency query process, many different access methods have been proposed. Token list based proximity search has been proved to be a good alternative method to the LSH for a large massive database proximity search. However, single-token based method leads to a high overhead in the results refinements process to achieve a required similarity. In this paper, we investigate how the multi-token list affects the performance of database proximity search. Numerous experiments have been conducted and the results show that two-token adjacent token list can achieve the best query performance in multi-token list based proximity search.