Skip to Main Content
Table 7 

Results for MWE detection on the French SPMRL corpus. Both the generation of the gold standard and the computations of the measures have been performed on this corpus.

MethodP@100P@200AP
upper baseline 1.000 0.860 1.0000 
lower baseline 0.521 0.521 0.5212 
frequency 0.500 0.480 0.4876 
  
t-test 0.500 0.485 0.4934 
C-value (pre-pruned) 0.490 0.540 0.5107 
MF (post-pruned) 0.510 0.495 0.5017 
FGM (post-pruned) 0.460 0.480 0.4703 
  
DRUID 0.790 0.690 0.7794 
log(freq) ⋅ DRUID 0.770 0.675 0.7631 
MF(post-pruned) ⋅ DRUID 0.700 0.630 0.6850 
FGM(post-pruned) ⋅ DRUID 0.600 0.570 0.5948 
MethodP@100P@200AP
upper baseline 1.000 0.860 1.0000 
lower baseline 0.521 0.521 0.5212 
frequency 0.500 0.480 0.4876 
  
t-test 0.500 0.485 0.4934 
C-value (pre-pruned) 0.490 0.540 0.5107 
MF (post-pruned) 0.510 0.495 0.5017 
FGM (post-pruned) 0.460 0.480 0.4703 
  
DRUID 0.790 0.690 0.7794 
log(freq) ⋅ DRUID 0.770 0.675 0.7631 
MF(post-pruned) ⋅ DRUID 0.700 0.630 0.6850 
FGM(post-pruned) ⋅ DRUID 0.600 0.570 0.5948 
Close Modal

or Create an Account

Close Modal
Close Modal