An extension of the Burrows Wheeler transform to k words

Abstract

Summary form only given. We introduce an extension of the Burrows-Wheeler transform to a multiset of primitive words. Primitiveness is not actually a restrictive hypothesis, since in practice almost all the processed texts are primitive (or become primitive by adding an end-of-string symbol). We prove that such a transformation as the BWT is reversible. We show how to use the transformation as a preprocessing for the simultaneous compression of different texts.