Fast construction of generalized suffix trees over a very large alphabet

Abstract

The work in this paper is motivated by the real-world problems such as mining frequent traversal path patterns from very large Web logs. Generalized suffix trees over a very large alphabet can be used to solve such problems. However, traditional algorithms such as the Weiner, Ukkonen and McCreight algorithms are not sufficient assurance of practicality because of large magnitudes of the alphabet and the set of strings in those real-world problems. Two new algorithms are designed for fast construction of generalized suffix trees over a very large alphabet, and their performance is analyzed in comparison with the well-known Ukkonen algorithm. It is shown that these two algorithms have better performance, and can deal with large alphabets and large string sets well. © Springer-Verlag Berlin Heidelberg 2003.