This paper deals with an automatic part-of-speech disambiguation of Czech texts containing the word to (E. it) in fixed collocations used especially in spoken Czech.