Studying Properties of Czech Complex Sentences from an Annotated Corpus