Faster Compressed Dictionary Matching (extended abstract)

Abstract

The past few years have witnessed several exciting results on compressed represen- tation of a string T that supports e±cient pattern matching, and the space complexity has been reduced to jTjHk(T)+o(jTj log ¾) bits [8, 10], where Hk(T) denotes the kth- order empirical entropy of T, and ¾ is the size of the alphabet. In this paper we study compressed representation for another classical problem of string indexing, which is called dictionary matching in the literature. Precisely, a collection D of strings (called patterns) of total length n is to be indexed so that given a text T, the occurrences of the patterns in T can be found e±ciently. In this paper we show how to exploit a sampling technique to compress the existing O(n)-word index to an (nHk(D) + o(n log ¾))-bit index with only a small sacri¯ce in search time.