Skip to Main Content
Table 2: 

Exact-match span F1 on the test set; does not allow for partial matches. {Disc, Fast}Align are both word alignment models, where ours were trained for span alignment. Cue adds positional information, Mx adds max pooling of span representations, Df adds element-wise difference of span representations, and SBCE adds soft binary cross entropy.

MethodPRF1
DiscAlign (29.82) (29.82) 29.82 
FastAlign (71.02) (71.02) 71.02 
 
Cue 10.39 9.77 10.07 
Mx 80.65 77.92 79.26 
Df 87.31 85.42 86.36 
Mx+Cue 87.50 86.49 86.99 
Df+Cue 88.74 86.96 87.84 
Df+Mx 89.27 87.29 88.27 
Df+Mx+Cue 89.15 88.19 88.67 
Df+Mx+Cue+SBCE 89.14 88.99 89.06 
MethodPRF1
DiscAlign (29.82) (29.82) 29.82 
FastAlign (71.02) (71.02) 71.02 
 
Cue 10.39 9.77 10.07 
Mx 80.65 77.92 79.26 
Df 87.31 85.42 86.36 
Mx+Cue 87.50 86.49 86.99 
Df+Cue 88.74 86.96 87.84 
Df+Mx 89.27 87.29 88.27 
Df+Mx+Cue 89.15 88.19 88.67 
Df+Mx+Cue+SBCE 89.14 88.99 89.06 
Close Modal

or Create an Account

Close Modal
Close Modal