Dear Task Organizers,
Are we able to get hold of all model descriptions before we submit the paper? I am specifically curious on how teams performed much better on Task B than on Task A, especially since we are only given training data for Task A and none for Task B.
It would help in the paper writing if we know beforehand (A) how many of the teams used external data as additional training data in an allowed manner (e.g. machine-annotated suggestions training data for the hotels domain), (B) how complicated or simple are the team models, and so on.
Thanks!
Sincerely,
Reinald Kim