Hello all,
It's come to our attention that Table 3 in the paper
reported the wrong results for the Directed Accuracy (with cutoff 10
instead of no cutoff).
Unfortunately this has invalidated a small part
of the evaluation. A corrected paper with updated results can be found
here.
Apologies for any problems this may have caused.
regards,
Douwe Gelling