Efficient Translation with Linear Bimorphisms

Christoph Teichmann, Antoine Venant, Alexander Koller

Abstract: We show that the image of a regular tree language under a linear bimorphism over binary signatures can be computed in linear time in the size of the input automaton. We do this by transformation into a novel normal form. Our result applies to the translation and parsing complexity of a wide range of grammar formalisms used in computational linguistics, which can now be shown in a uniform way.