In this paper we present a new dictionary-based preprocessing technique and its implementation called TWRT (Two-level Word Replacing Transformation). Our preprocessor uses several dictionaries and divides files into various kinds, what improves the compression performance in latter stage. TWRT also recognizes multilingual text files, but operates with almost the same speed as its predecessor ? WRT. TWRT in connection with a popular bzip2 and modern compressors PAQ6 and PPMonstr improves the compression performance by about 5?8%. Moreover, TWRT improves the compression speed with PAQ6 and on larger files with PPMonstr.