In the paper, accepted at CONLL 🎊, we have conducted an in-depth error analysis, comparing CAIT annotations with those of the off the-shelf English Stanza parser, highlighting systematic differences in dependency predictions and the structural biases introduced by
general-purpose parsers.