Visualize: Figure2 & 3.
Description: " Alignment of in- and output words via the attention for different models tested on bAbI+ data"
"A labelled example sentence to evaluate whether models have distinct representations for
reparanda, repairs, and editing terms"