Hi! This work is excellent, and I am interested in the detailed implementations. So, I am wondering when you guys will release the code. Can you provide an approximate date?
Any ground truths of role labels are used in eval/test stage?
Thanks for this wonderful work. However, when I run the original codes provided by you, I find that ground truths of role labels are used in eval/test stage. Specifically, in transformer.py, the variable "inference" is "False" when the codes run to line 94, so that the labels of role will be used to predict. Besides, if I block the ground truths of role labels, the test performance will drop significantly. I am not sure if this is the result of the code running properly. I would appreciate it if you could address my concern.