Data compression with factor automata

Summary form only given. A factor encoding method parses texts into words of variable length and then translates each of the obtained words with a variable length code. The factor automaton is dynamically constructed as the text is processed. A word is encoded by two integers: its length and its position in the previous part of the text. The position of the first occurrence of a factor is computed by the automaton. A special code is reserved for the first appearance of each letter. As the decipherer processes the encoded message, it rebuilds the factor automaton, so that the encipherer and the decipherer have at their disposal the same automaton when they have treated the same part of the text