PPM with the extended alphabet

In the following paper we propose modification of Prediction by Partial Matching (PPM)—a lossless data compression algorithm, which extends an alphabet, used in the PPM method, to long repeated strings. Usually the PPM algorithm’s alphabet consists of 256 characters only. We show, on the basis of the Calgary corpus [T.C. Bell, J. Cleary, I.H. Witten, Text compression. Advanced Reference Series, Prentice Hall, Englewood Cliffs, New Jersey, 1990], that for ordinary files such a modification improves the compression performance in lower, but not greater than 10, orders. However, for some kind of files, this modification gives much better compression performance than any known lossless data compression algorithm.