Compact PAT trees

Abstract

Given a text string S = s 1s2s3 •.. sn, we want to preprocess S such that given a pattern P = P1P2P3···Pm, we can find {ilsi·•"i+m-1 = P} as efficiently as possible. Suffix trees are a data structure solution to this problem. Unfortunately, when n is large, the storage required by a suffix tree can be prohibitive. This thesis presents several related new representations for a close relative of the suffix tree, the PAT tree, that retain the functionality of suffix trees while requiring a fraction of the storage used by current methods.