For lossless data compression of the texts of natural language and for achieving better compression ratio we can use linguistic and grammatical properties extracted from the text analysis. This work deals with usage of word order, word categories and grammatical rules in sentences and sentence units in Czech language. Special grammatical properties of this language which are different from for example English language are used here. Further, there is an algorithm designed for searching similarities in analyzed sentence structures and its next processing to final compressed file. For analysis of the sentence units a special tool is used which allows parsing on more levels.