Skip to Main Content
Table 6: 
Error analysis of unlabeled dependency parsing in the test set of different datasets.
ModelsCTB-5CTB-7CTB-9
PudepSeg-wrongHead-wrongPudepSeg-wrongHead-wrongPudepSeg-wrongHead-wrong
Pipeline 86.28% 3.49% 10.23% 80.75% 7.10% 12.15% 81.48% 6.76% 11.76% 
Joint-Multi 87.65% 3.43% 8.92% 81.81% 6.80% 11.39% 82.08% 6.55% 11.37% 
Joint-Multi-BERT 89.22% 3.2% 7.58% 85.01% 6.10% 8.89% 85.59% 5.74% 8.89% 
ModelsCTB-5CTB-7CTB-9
PudepSeg-wrongHead-wrongPudepSeg-wrongHead-wrongPudepSeg-wrongHead-wrong
Pipeline 86.28% 3.49% 10.23% 80.75% 7.10% 12.15% 81.48% 6.76% 11.76% 
Joint-Multi 87.65% 3.43% 8.92% 81.81% 6.80% 11.39% 82.08% 6.55% 11.37% 
Joint-Multi-BERT 89.22% 3.2% 7.58% 85.01% 6.10% 8.89% 85.59% 5.74% 8.89% 

The value of Pudep is the percentage that the predicted arc is correct. ‘Seg-wrong’ means that either head or dependent (or both) is wrongly segmented. ‘Head-wrong’ means that the word is correctly segmented but the predicted head word is wrong.

Close Modal

or Create an Account

Close Modal
Close Modal