Skip to Main Content
Table 9 

Results for ranking n-grams, according to their multiwordness, based on the French ERC. The candidates are extracted based on the smaller SPMRL corpus.

MethodP@100P@500AP
upper baseline 1.000 1.000 1.0000 
lower baseline 0.220 0.220 0.2201 
frequency 0.370 0.354 0.3105 
  
C-value 0.420 0.366 0.3059 
t-test 0.390 0.360 0.3134 
GM 0.010 0.052 0.1694 
MF 0.370 0.356 0.3148 
FGM 0.280 0.260 0.2405 
  
DRUID 0.700 0.568 0.3962 
log(freq) ⋅ DRUID 0.760 0.582 0.4075 
MF ⋅ DRUID 0.570 0.516 0.3776 
FGM ⋅ DRUID 0.510 0.418 0.3234 
MethodP@100P@500AP
upper baseline 1.000 1.000 1.0000 
lower baseline 0.220 0.220 0.2201 
frequency 0.370 0.354 0.3105 
  
C-value 0.420 0.366 0.3059 
t-test 0.390 0.360 0.3134 
GM 0.010 0.052 0.1694 
MF 0.370 0.356 0.3148 
FGM 0.280 0.260 0.2405 
  
DRUID 0.700 0.568 0.3962 
log(freq) ⋅ DRUID 0.760 0.582 0.4075 
MF ⋅ DRUID 0.570 0.516 0.3776 
FGM ⋅ DRUID 0.510 0.418 0.3234 
Close Modal

or Create an Account

Close Modal
Close Modal