Two-dimensional dynamic signature file method using extendible hashing and frame-slicing techniques

Abstract

The signature file method has been widely advocated as an efficient index scheme to handle applications demanding a large amount of text databases. Moreover, many recent applications in practice require a dynamic information storage structure. For this, in this paper, we propose two-dimensional dynamic signature file (TDSF) methods using multilevel extendible hashing and frame-slicing techniques in order to efficiently handle text and multiattribute data in dynamic environments. Due to the characteristics of the extendible hashing technique, the TDSF is suitable for supporting insertions, deletions, and updates. Based on the frame-slicing technique, we divide a signature into several frames, and store each frame in a separate file so that we may increase the filtering effect of the signature file, leading to a decrease in the amount of retrieval required for searching a query. We also describe the storage structure of the TDSF and outline algorithms for signature insertion, retrieval, and deletion. Finally, we derive analytic performance evaluation models of the TDSF methods, and perform extensive experiments with various frame divisions. We show from performance results that, regardless of the database size, the TDSF methods outperform the Quick Filter method in terms of both retrieval time and dynamic operating measure considering the degree of dynamic environment.