It is pointed out that the generalized thermodynamics introduced by Tsallis provides a natural frame for extension of thermodynamical formalism of symbolic dynamics. The statistical structure of 1D fractal long-range correlation sequences is investigated using n-tuple Zipf analysis. The generalized form of Zipf-Mandelbrot law can be obtained using maximum entropy principle within the frame of Tsallis thermodynamics, if the frequency of a word is associated with probability and the rank with the energy.