The following paper has two aims. First, it introduces a procedure of a manual annotation of selected linguistic phenomena across a large-scale dependency treebank of English.
The method was designed to provide higher consistency of annotated data, and so higher credibility of the treebank. Second, the first expert task completed by means of this method is being described ? the annotation of rhematizers and discourse connectives and their modifiers, i.e. annotation of some non-dependency relations in a dependency approach.