Pattern Matching in Hypertext

Abstract

The importance of hypertext has been steadily growing over the past decade. The Internet and other information systems use hypertext format, with data organized associatively rather than sequentially or relationally. A myriad of textual problems have been considered in the pattern matching field with many nontrivial results. Nevertheless, surprisingly little work has been done on the natural combination of pattern matching and hypertext. In contrast to regular text, hypertext has a nonlinear structure and the techniques of pattern matching for text cannot be directly applied to hypertext. Manber and Wu (1992, “IAPR Workshop on Structural and Syntactic Pattern Recognition, Bern, Switzerland”) pioneered the study of pattern matching in hypertext and defined a hypertext model for pattern matching. Akutsu (1993, “Procedures of the 4th Symposium on Combinatorial Pattern Matching, Podova, Italy,” pp. 1–10) developed an algorithm that can be used for exact pattern matching in a tree-structured hypertext. Park and Kim (1995, “6th Symposium on Combinatorial Pattern Matching, Helsinki, Finland”) considered regular pattern matching in hypertext. They developed a complex algorithm that works for hypertext with an underlying structure of a DAG. In this paper we present a much simpler algorithm achieving the same complexity which runs on any hypertext graph. We then extend the problem to approximate pattern matching in hypertext, first considering hamming distance and then edit distance. We show that in contrast to regular text, it does make a difference whether the errors occur in the hypertext or the pattern. The approximate pattern matching problem in hypertext with errors in the hypertext turns out to be NP-complete and the approximate pattern matching problem in hypertext with errors in the pattern has a polynomial time solution.