Skip to Main Content
Table 7

English test set results (WSJ Section 23) applying sentence-level high precision and unary constraints to three parsers with parameter settings tuned on development data.

Parser
F1
Precision
Recall
Seconds
Speed-up
BUBS (2010) 88.4 88.5 88.3 586  
 + Unary(100) 88.5 88.7 88.3 486 1.2x 
 + HP(0.9) 88.7 88.9 88.6 349 1.7x 
 + HP(0.9) + Unary(100) 88.7 89.0 88.4 283 2.1x 
 
Charniak (2000) 89.7 89.7 89.6 1,116  
 + Unary(100) 89.8 89.8 89.7 900 1.2x 
 + HP(0.8) 89.8 90.0 89.6 716 1.6x 
 + HP(0.8) + Unary(100) 89.7 90.0 89.5 679 1.6x 
 
Berkeley (2007) 90.2 90.3 90.0 564  
 + Unary(125) 90.1 90.3 89.9 495 1.1x 
 + HP(0.7) 90.2 90.4 90.0 320 1.8x 
 + HP(0.7) + Unary(125) 90.2 90.4 89.9 289 2.0x 
Parser
F1
Precision
Recall
Seconds
Speed-up
BUBS (2010) 88.4 88.5 88.3 586  
 + Unary(100) 88.5 88.7 88.3 486 1.2x 
 + HP(0.9) 88.7 88.9 88.6 349 1.7x 
 + HP(0.9) + Unary(100) 88.7 89.0 88.4 283 2.1x 
 
Charniak (2000) 89.7 89.7 89.6 1,116  
 + Unary(100) 89.8 89.8 89.7 900 1.2x 
 + HP(0.8) 89.8 90.0 89.6 716 1.6x 
 + HP(0.8) + Unary(100) 89.7 90.0 89.5 679 1.6x 
 
Berkeley (2007) 90.2 90.3 90.0 564  
 + Unary(125) 90.1 90.3 89.9 495 1.1x 
 + HP(0.7) 90.2 90.4 90.0 320 1.8x 
 + HP(0.7) + Unary(125) 90.2 90.4 89.9 289 2.0x 
Close Modal

or Create an Account

Close Modal
Close Modal