Computing the prefix of an automaton

¹ Institut Gaspard Monge, Université de Marne-la-Vallée, 5 boulevard Descartes, 77454 Marne-la-Vallée Cedex 2, France; (Marie-Pierre.Beal@univ-mlv.fr)
² Institut Gaspard Monge, Université de Marne-la-Vallée, 5 boulevard Descartes, 77454 Marne-la-Vallée Cedex 2, France; (Olivier.Carton@univ-mlv.f et url: http://www-igm.univ-mlv.fr/~beal/)

Received: May 2000
Accepted: February 2001

Abstract

We present an algorithm for computing the prefix of an automaton. Automata considered are non-deterministic, labelled on words, and can have ε-transitions. The prefix automaton of an automaton has the following characteristic properties. It has the same graph as . Each accepting path has the same label as in . For each state q, the longest common prefix of the labels of all paths going from q to an initial or final state is empty. The interest of the computation of the prefix of an automaton is that it is the first step of the minimization of sequential transducers. The algorithm that we describe has the same worst case time complexity as another algorithm due to Mohri but our algorithm allows automata that have empty labelled cycles. If we denote by P(q) the longest common prefix of labels of paths going from q to an initial or final state, it operates in time O((P+1) × |E|) where P is the maximal length of all P(q).