Skip to Main Content
Table 14 
Accuracies of different models on test set. “GM” is short for graph merging; “TR (DO)” is short for transition-based parsing with dynamic oracle; “TR (BS)” is short for transition-based parsing with beam search. The performance of the TR (BS) parser with linear scoring model is from Zhang et al.’s paper. Gold-standard POS tags are utilized.
  UPURUFUCMLPLRLFLCM
GM Neural 88.84 88.23 88.53 29.14 86.47 85.87 86.17 24.68 
GM Linear 88.06 85.11 86.56 26.24 86.03 83.16 84.57 22.84 
TR (DO) Neural 89.56 85.75 87.61 25.11 87.41 83.69 85.51 21.86 
TR (BS) Neural 84.77 87.70 86.21 20.61 82.21 85.05 83.61 18.07 
TR (BS) Linear - - - - - - - - 82.28 83.11 82.69 - - 
  UPURUFUCMLPLRLFLCM
GM Neural 88.84 88.23 88.53 29.14 86.47 85.87 86.17 24.68 
GM Linear 88.06 85.11 86.56 26.24 86.03 83.16 84.57 22.84 
TR (DO) Neural 89.56 85.75 87.61 25.11 87.41 83.69 85.51 21.86 
TR (BS) Neural 84.77 87.70 86.21 20.61 82.21 85.05 83.61 18.07 
TR (BS) Linear - - - - - - - - 82.28 83.11 82.69 - - 
Close Modal

or Create an Account

Close Modal
Close Modal