Context Model Automata for Text Compression

Finite-state automata offer an alternative approach in the implementation of context models where the states in the automata cannot in general be assigned by a single context. Despite the potential of this approach, it makes the design of the modelling more problematic because the exact behaviour of the model is not known. Here we propose a simple formalism—context model automata (CMA)— that gives an exact interpretation for the minimum context belonging to each state in the automaton. The formalism is general enough to simulate context models such as PPM and GDMC. Using the CMA formalism as our tool, we study the behaviour of the above two context models.