xdqkid

Results 3 issues of xdqkid

model/reconstructor/utils/loss.py line 157 162. sentence normalization token num is not equal to amr normalization token num, they should not be shared normalization. So, in fact, ratio is uncertain.

I noticed that amr data in your experiments is stripped format, e.g. have-org-role. Can you share full format amr data, e.g. have-org-role-91, in your experiment ?