MIT Press

Skip to Main Content

Table 5:

Counts of AMR graphs with different depth for the test split and the BLEU scores of different models on these graphs.

Model	Depth
Model	1−5	6−10	11−
#count	278	828	265

S2S	37.2	21.2	19.3
GraphLSTM	+3.9	+2.1	+0.7
Graph Transformer	+6.0	+4.5	+3.7

Model	Depth
Model	1−5	6−10	11−
#count	278	828	265

S2S	37.2	21.2	19.3
GraphLSTM	+3.9	+2.1	+0.7
Graph Transformer	+6.0	+4.5	+3.7