In this article, a new compression method is investigated using a particular chaotic modulation type. The symbolic approach presented associates to every informational sequence a trajectory in the state space of the chaotic generator. We are also introducing a new type of chaotic generator adapted to the probability distribution of the informational sequence, and we prove that using the described generator the theoretical compression performances attain the optimal entropy compression. Finally we confirm the theoretical analysis with performance tests for different sequences with different symbolic probabilities.