Table 2:
Acceptability measures for predicting the acceptability of a sentence; P(s) is the sentence probability, computed using Equation (1) or Equation (2) depending on the model; Pu(s) is the sentence probability estimated by a unigram language model; and α =0.8.
Acc. MeasureEquation
LP $logP(s)$
MeanLP $logP(s)|s|$
PenLP $logP(s)((5+|s|)/(5+1))α$
NormLP $−logP(s)logPu(s)$
SLOR $logP(s)−logPu(s)|s|$
Acc. MeasureEquation
LP $logP(s)$
MeanLP $logP(s)|s|$
PenLP $logP(s)((5+|s|)/(5+1))α$
NormLP $−logP(s)logPu(s)$
SLOR $logP(s)−logPu(s)|s|$
Close Modal