Skip to Main Content
Table 7: 
Configurations for y + and y for fully supervised MT. ŷ is the highest-probability model output, ȳ is a gold standard reference. πw(y|x) is the probability of y according to the model. The arg maxy is taken over the k-best list 𝒦(x). BLEU +1 is smoothed per-sentence BLEU and α is a scaling factor.
Lossy +y
RAMP arg maxy πw (y|x) − α(1 − BLEU+1(y, ȳ)) arg maxy πw (y|x) + α(1 − BLEU+1(y, ȳ)) 
RAMP1 ŷ arg maxy πw (y|x) + α(1 − BLEU+1(y, ȳ)) 
RAMP2 arg maxy πw (y|x) − α(1 − BLEU+1(y, ȳ)) ŷ 
PERC1 ȳ ŷ 
PERC2 arg maxy BLEU+1 (y, ȳŷ 
Lossy +y
RAMP arg maxy πw (y|x) − α(1 − BLEU+1(y, ȳ)) arg maxy πw (y|x) + α(1 − BLEU+1(y, ȳ)) 
RAMP1 ŷ arg maxy πw (y|x) + α(1 − BLEU+1(y, ȳ)) 
RAMP2 arg maxy πw (y|x) − α(1 − BLEU+1(y, ȳ)) ŷ 
PERC1 ȳ ŷ 
PERC2 arg maxy BLEU+1 (y, ȳŷ 
Close Modal

or Create an Account

Close Modal
Close Modal