The latest preliminary precision measurements (performed with SET version 0.2) show that the precision of the parser dependency output ranges between 75 and 86 percent with respect to the human-anotated Czech corpus data, depending on the particular testing set.