An efficient method for in memory construction of suffix arrays

The suffix array is a string-indexing structure and a memory efficient alternative to the suffix tree. It has many advantages for text processing. We propose an efficient algorithm for sorting suffixes. We call this algorithm the two-stage suffix sort. One of our ideas is to exploit the specific relationships between adjacent suffixes. Our algorithm makes it possible to use the suffix array for much larger texts and suggests new areas of application. Our experiments on several text data sets (including 514-MB Japanese newspapers) demonstrate that our algorithm is 4.5 to 6.9 times faster than Quicksort, and 2.5 to 3.6 times faster than K. Sadakane's (1998) algorithm, which is considered to be the fastest algorithm in previous work.